aiLL 7/17/23 - Engaging With Generative AI: A Path to Creativity and Connection

Description

Chapters

Transcript

0:17 all right getting back
0:22 here to where I was
0:26 before I have
0:29 got a rubbe Goldberg machine of lights
0:35 and charging
0:37 cables it's you think it's easy getting
0:40 this dual camera [ __ ] working no I mean
0:43 would it be easy if I had real equipment
0:45 sure do I have real equipment
0:51 no welcome everybody my name is Kyle
0:53 Shannon this the AI learning lab if you
0:55 were here before I asked everyone to go
0:58 do something go fill out a form and I
1:00 didn't give him permission to fill out
1:01 the
1:03 form cuz I'm a professional hey Jen
1:06 how's it
1:08 going welcome welcome welcome we'll let
1:11 some people get in here um feel free to
1:14 share the live if you want to help get
1:16 people here for those of you who have
1:19 been here before look I have dual
1:21 cameras my dual camera is back huh how
1:25 exciting is
1:27 that I don't think it's exciting for
1:29 anyone but me
1:30 [Laughter]
1:32 hello rug cleaner
1:35 machines do you rent them or make them
1:38 or use
1:40 them I have many dirty rugs I have one
1:44 that's sort of a giant shag carp but
1:46 it's an absolute disaster to vacuum
1:49 everything gets caught in the vacuum
1:50 it's a disaster is there a way to get
1:52 chat GPT to remember steps to
1:54 instructions if they are longer than the
1:57 limit
1:59 no but
2:01 if you go to po.com so the third thing
2:04 right here right
2:07 there um and you pay their 20 bucks
2:11 subscription you get access to um
2:14 there's a Claude model called Claude
2:16 instant
2:17 100K that has 100,000 token limit which
2:20 is about 75,000 words so with
2:24 PO with with Claude 100K you can you can
2:27 get much much longer things
2:31 oh manufacturing and distribution very
2:32 hard thanks for asking man there's so
2:34 much I I like I I would I can't even
2:38 imagine all the stuff that you're going
2:39 to be able to do with with all this chat
2:42 GPT stuff with manufacturing and
2:44 distribution especially when um if you
2:46 haven't seen it yet go check out go
2:48 watch uh YouTube videos on um code
2:52 interpreter where you're going to be
2:53 able to take a lot of your data um for
2:56 for what you do and do all sorts of
2:58 crazy stuff with it so
3:01 hello Lauren how are you welcome
3:05 welcome um oh great there's there's a
3:07 bunch of folks here welcome my name is
3:09 Kyle Shannon this is the AI learning lab
3:12 I'm an entrepreneur I've started 12 or
3:14 15 companies in my day um this channel
3:17 is really just about talking about AI
3:19 answering questions um demystifying it
3:23 um hopefully getting people as um as
3:27 curious and uh excited I guess is the
3:31 right word I'm I'm cautiously excited
3:34 about generative AI um I think it's one
3:37 of the most um profound Technologies in
3:40 the history of the world I really do and
3:43 um my my context for that I was I was
3:46 there for the early days of the
3:48 worldwide web I started one of the early
3:49 digital agencies company called
3:51 agency.com
3:52 and uh so I got to experience watching
3:57 you know the worldwide web do what it
3:58 did from the ground up
4:00 and this stuff seems like that only a
4:03 lot more profound and it's coming a lot
4:05 um by the way if you have if you've been
4:07 here in the past if you've been here in
4:09 the past week or no it's probably been
4:12 like 3 weeks they took away my dual
4:14 cameras they took away my ability to do
4:17 both of these picture and picture things
4:20 uh they gave it back to me tonight I
4:22 don't know why cuz Tik Tock anyway if
4:25 you've never used chat GPT go to that
4:27 top URL chat. open.com
4:30 that is chat gpt's official website just
4:33 go there right now go there and and
4:35 start playing until you have what what
4:39 I've been calling you know your Kevin
4:40 McAllister moment your home alone
4:44 moment um at some point it will do
4:47 something that will blow your mind and
4:49 at that point then you'll go well holy
4:51 crap what else can it do and then you're
4:53 hooked um this this URL right here
4:55 prompts do chat if you don't know
4:57 anything about chat GPT and haven't even
4:59 heard of prompts and you don't know what
5:00 the hell I'm talking about prompt. chat
5:02 will tell you what it is how chat GPT
5:05 works and it gives you a whole long list
5:07 of ideas of things that you can do with
5:09 chat GPT then bing.com is Microsoft
5:12 search engine so if you go there and
5:14 click on the chat button that's actually
5:16 chat
5:17 gp4 um and it's connected to the
5:19 internet and it's free and you can make
5:22 images with it you say make me an image
5:24 of a cat
5:25 surfing period and it will do that um it
5:30 uses do which of all the image
5:32 generation tools is the crappiest but it
5:34 still makes images that you ask it to
5:36 it's pretty cool and then po.com gives
5:38 you access to three different or six
5:40 different models uh from three different
5:42 companies so if you've got questions
5:44 about AI pop them in the comments below
5:46 I'll do my best to answer them um oh
5:49 another request
5:51 so for the past two nights a couple of
5:53 folks have asked um do I do I have a
5:56 workshop and the answer is no but I now
5:59 have a workshop
6:01 graphic so so I did this today you know
6:04 I put my my my face on there but we you
6:06 don't need to see that twice um so so I
6:09 have a Graphic now um but but I was
6:11 thinking about what do I want to do for
6:13 a workshop and I thought well it
6:14 shouldn't be what I want to do for a
6:16 workshop it should
6:17 be what what's what's most needed so I
6:21 created a survey so if you go to link
6:23 tree um slre Kyle Shannon I don't I just
6:27 put the capital letters there so that
6:30 you can read it easier um it doesn't
6:32 matter though um but if you go there the
6:35 first link is is a survey that I've now
6:37 give everyone given everyone permission
6:39 to so if you could just confirm that you
6:42 actually have permission to take the
6:43 survey that would be great but if you go
6:45 there um tell me it's just three
6:47 questions it's just basically like what
6:49 are you trying to learn like what would
6:50 make the biggest difference for you and
6:52 what you know you know what what results
6:55 would you like to see in the next year
6:57 right so if I can just get a sense of
6:58 what people are trying to accomplish
7:00 great my my initial instinct is to do
7:03 kind of a two-day two 2hour sessions on
7:07 two different days where the first one
7:08 is just kind of a primer here's chat gbt
7:11 Maybe here's some image stuff send
7:13 people off with some homework and you
7:15 know go play it homework for me is going
7:18 to be like go play um and then come back
7:21 the next day and talk about what's a
7:23 project in your life you want to you
7:24 want to make some progress with and then
7:26 together we'll do that but if you would
7:28 do me the favor go there fill out the
7:30 survey let me know if I did a workshop
7:33 what you what you want to accomplish and
7:35 then I'll I'll do it based based on that
7:38 sorry I just ate so I've got semi
7:41 hiccups they're not quite yeah they're
7:43 just
7:46 anyway uh welcome welcome welcome uh if
7:50 you have questions about uh AI pop it
7:53 down below and I will do my best to
7:56 answer it love from Pittsburgh what's up
7:58 Pittsburgh I grew up in York
8:01 PA York Pennsylvania then I moved to New
8:05 York
8:07 City out of school I went to Penn State
8:10 can chat GPT teach you about poker
8:12 Theory oh yeah so so go to you'll notice
8:17 most of the document are all these
8:18 personas so like act as a mathematician
8:22 act as a screenwriter whatever it is go
8:25 look at how those prompts are
8:27 constructed and then tell chat GPT act
8:30 like a championship um you know poker
8:32 player um for competitive poker and it
8:36 will teach you poker Theory like really
8:40 well
8:43 um
8:44 okay kizzle J kizzle thank you you're
8:48 welcome rug cleaner machines thumbs up
8:51 awesome hi from Oregon welcome hello I'm
8:56 I'm in Denver by the
8:58 way um
9:03 I someone says they love me awesome
9:06 thank you I appreciate that um thoughts
9:10 on prompt engineering trainings and
9:12 badges I saw
9:14 something or of other and then I saw
9:18 Vanderbilt University is offering a
9:20 class on
9:21 corsera dual camera is back Sonia yes
9:25 and by the way I saw like Sonia you and
9:28 and the fundraising woman um have
9:31 started showing up in my feed now so so
9:33 it's cool to see you talking about
9:35 chachy PT that's pretty cool but yes
9:37 dual camera is back why I don't know I
9:41 guess I was a good boy and they gave me
9:42 back my dual
9:45 camera so so here's my thoughts on
9:48 courses so so people in here are asking
9:50 me to do a workshop so I
9:52 I I think I would be U it would be
9:55 disingenuous of me to say that courses
9:58 are a waste of time uh cuz I'm thinking
10:00 of putting one
10:02 together I would take everything you
10:04 learn in a course right now with a bit
10:06 of a grain of salt because not not
10:08 because the courses aren't bad like do I
10:10 think there are people out there that
10:11 know enough about this stuff now that
10:14 they can legitimately teach a course
10:16 yes
10:19 but I think it is as valuable if not
10:23 more valuable right now to teach
10:24 yourself to just play to get curious to
10:26 just just start start playing with these
10:29 things figuring it out for yourself um
10:31 if the course can help accelerate that
10:33 great um but don't put too much weight
10:36 on it because 6 months from now
10:37 everything's going to be different so
10:39 whatever you learn in the course right
10:40 now some of the Core Concepts around
10:42 prompting may be
10:44 valuable but I think even prompting is
10:47 going to change dramatically
10:49 because one of the trends that I'm
10:51 seeing in prompting right now is writing
10:53 prompts to get chat GPT to prompt to
10:56 give you better prompts or to prompt
10:58 itself
11:00 um so so chat GPT effectively acts like
11:03 your um your prompting coach and kind of
11:07 guide you through it writing prompts to
11:10 give you what you want um that's a trend
11:12 I've seen recently that's going to
11:14 happen more and more programmatically
11:17 where the need to do really
11:19 sophisticated prompting if you're on the
11:21 development side of the world if you're
11:23 an engineer and you're developing
11:25 product then yeah definitely need to
11:27 learn this if you're an end user I think
11:30 the tools are going to get increasingly
11:31 sophisticated so
11:34 um yeah so I would I would just you know
11:38 find someone that you don't mind their
11:41 voice CU sometimes courses like I'm just
11:44 listening like I I'll watch a you it's
11:46 funny when when I go to watch YouTube
11:48 series when I want to learn something
11:50 well preach tgbt um when I want to learn
11:53 something you know I go look for all the
11:55 videos and my first criteria is can I
11:58 stand their voice and if I could stand
11:59 their voice then then then I would um
12:02 listen to them cuz if I couldn't I'm
12:04 just like nothing nothing goes into my
12:06 brain um Can it help with farming yeah I
12:10 mean I think it so it here's the thing
12:13 about about and please let me know I'm
12:16 pretty sure that when I switch cameras
12:18 it delays for a second so please let me
12:20 know if that's still the
12:22 case um the thing about these large
12:24 language models is the depth of
12:28 knowledge that they have is staggeringly
12:31 deep and so so things like farming where
12:35 there's lots and lots of stuff on the
12:37 internet the these large language models
12:39 have essentially been trained on all of
12:40 the public internet so the trick is
12:43 getting them kind of um molded into the
12:47 expert that you need so it it depends on
12:51 what you're doing with farming if you're
12:52 doing like big commercial farming say
12:55 Hey you know you run a commercial
12:57 farming Enterprise and and you're an
12:59 expert in that and I want help on either
13:02 the economics or on fertilizer or on
13:06 cycling crops or whatever it might be if
13:08 it's smaller scale then you know have it
13:12 have it be you know act as an organic
13:14 farmer and I I want to be able to have
13:16 the same results as a big commercial
13:18 Farm but I want everything to be organic
13:20 and then start interacting with it the
13:22 best way to interact with these tools is
13:24 really to think of it less like you have
13:27 to get the perfect answer the first time
13:29 time and more like you're you're having
13:31 a conversation with it so kind of set it
13:33 up as the expert that you want it to be
13:35 and then start interrogating it and yes
13:39 it will help you a lot um it it can help
13:43 you a lot prompt. chat is uh Simeon D
13:47 from Bulgaria oh that's cool did not
13:49 know
13:51 that much love from Pittsburgh what is
13:53 Nova I don't know when I was a kid Nova
13:57 Nova was a PBS show about science [ __ ]
14:00 when did they bring back dual camera
14:01 this afternoon apparently I went to go
14:04 live and there it was cuz I've been
14:06 looking for it every time I go to go
14:08 live I'm like is it there and then today
14:10 I didn't look for it and it was
14:13 there so there you have it dual camera
14:16 is back what's the best email to reach
14:18 you if you go
14:20 to that link my emails in there and then
14:24 at the top this Workshop by the way if
14:27 if you're new here didn't see this
14:29 before
14:31 um couple of people on the live had said
14:34 hey would you put together a workshop um
14:36 I have not put it together yet I have
14:37 put together a graphic I have a Graphic
14:40 now
14:41 um but but the first link at that link
14:45 tree is a link to a survey um so tell me
14:49 tell me the kinds of things that would
14:51 make a difference for you in a workshop
14:53 and then that's what I'll make the the
14:54 workshop around so so it has not been
14:56 built yet I don't have timing for it yet
14:59 um like
15:01 I I'm reticent to do a workshop if it's
15:05 not
15:06 good so so I want to make sure it's good
15:09 and it serves everyone so that's why I
15:10 want to do the the
15:12 survey
15:15 uh to reach out and ask you to join in
15:17 on a presentation oh that's great po pop
15:19 yeah happy to talk about that I am
15:23 uh I don't know as you can imagine or as
15:26 you can see I'm pretty passionate about
15:28 you know talking about this stuff
15:30 wherever and
15:33 whenever have you heard
15:35 about bensur decentralized Ai No bensur
15:40 is that how it's spelled oh oh bit
15:43 tensor bensur yeah he's clever uh bit
15:47 tensor no I haven't let me go let me go
15:50 look at it now
15:55 um bit
15:58 tensor
16:00 B andur internet scaling machine
16:07 learning open decentral open
16:12 incentivized decentralized and user
16:15 governed
16:18 build all right so you can go build some
16:20 [ __ ] there's a white paper cool I'll
16:23 check it
16:25 out yeah I think that's interesting I
16:27 mean with all the large with the open
16:29 source large large language model stuff
16:31 I think a decentralized AI the ability
16:33 to
16:35 do it what would actually be really
16:38 interesting there is if I don't know if
16:40 that project is going to get there but
16:42 if there's a way to do decentralized
16:45 training where you could do a foundation
16:48 model um in a decentralized way that
16:51 could compete with open AI or Google
16:55 that could be interesting yeah that
16:57 could be really interesting so thank you
16:59 for sharing
17:02 that hello from Canada hello Liza what
17:06 do you think about the app ask AI versus
17:08 chat. openai app I am the reason I have
17:12 these things on screen all the time when
17:14 I'm
17:14 talking is
17:18 um open AI is the one that like chat GPT
17:23 GPT 4 right now is by far the best of
17:27 these large language models there's just
17:29 nothing even close to it the open source
17:31 things are getting close to 3.5 and in
17:34 fact one of them just surpassed 3.5 in a
17:37 big test so they're coming Bard's not
17:41 even close the Llama models aren't even
17:43 close
17:45 um um the Claude models are decent and
17:49 they've got some features that are
17:51 different those are from anthropic but
17:53 those are a bunch of X open AI folks so
17:56 they should be good um so
18:00 if I I don't know what features ask AI
18:03 gives you but I assume it's just kind of
18:06 a prompt engine sitting on top of the
18:09 the GPT 3.5 or gp4 model which so is
18:13 chat GPT so for me I'm using chat GPT I
18:17 just think it's like especially if
18:21 you're the plus subscriber there they
18:23 give you early access to their
18:24 technology and it's just better like gp4
18:28 their GPT for at Bing just they're just
18:31 totally different experiences so hello
18:34 from Allen Texas wow cool I love that we
18:36 got people from all over I'm obsessed
18:38 with chat GPT and Bing chat I know it's
18:40 just it's like it's like once you once
18:44 you have that moment of like oh my God
18:47 it's this isn't just Google right this
18:49 isn't just a fancy Google once you have
18:52 that moment
18:55 it's I don't know it's hard to not think
18:57 about it all the time because it's just
19:01 oh wait response to the survey was
19:02 submitted thank you thank you thank you
19:05 that's awesome here's what the survey
19:06 looks like look I did that I did that
19:08 image in uh in mid Journey isn't that
19:12 slick cuz it's like books and learning
19:15 and laptops and the dude at computers
19:18 learning AI in a lab
19:21 huh all right um let me go to the
19:25 form see how many we got for
19:32 s response is one ah all right Megan
19:36 thank you thank you um if you haven't
19:40 filled out the
19:42 form if you would be so kind go to there
19:46 link tree.com KY shannnon there's the
19:49 first link is a form to just ask
19:51 questions about if I were to do a
19:53 workshop what would you want in it and
19:56 uh and then I will go build it that way
19:59 if I feel that it's something I can
20:02 deliver and I will not build it if it's
20:05 something I feel I can't deliver because
20:07 that's how I
20:12 am I'm excited to do the course awesome
20:16 I've got I've got to create the course
20:18 I'm thinking Blissful Mama of um a two
20:23 like a two day two two hour sessions one
20:26 is kind of a primer session and then one
20:28 is kind of Hands-On what are people
20:30 trying to solve and and do some you know
20:32 specific projects where we get everyone
20:35 started that's that's what I'm thinking
20:37 right
20:38 now
20:45 um it's a quick delay but not bad oh
20:47 when I flip the cameras okay great
20:49 there's a brief delay for me yes the
20:51 delay is back maybe two or three seconds
20:54 okay so I got to
20:57 remember
20:59 to wait that long so hopefully that was
21:02 good what did you say to watch on
21:03 YouTube where did that comment go um
21:07 about coding okay Daisy so if you have
21:11 not um seen videos of people using Code
21:16 interpreter um go do that especially if
21:18 you're into data analytics in any way so
21:21 just go to YouTube Type in code
21:22 interpreter chat
21:25 GPT and um code interpreter is basically
21:29 a new feature coming out from open AI so
21:33 from from chat GPT so right now if you
21:36 pay for GPT plus you get GPT 4 gp4
21:40 connected to the internet and plugins
21:42 the fourth mode is called code
21:45 interpreter and it allows you to upload
21:48 data like 100
21:51 megabyte um spreadsheet of data and it
21:55 will read it all and then it does full
21:58 data analysis you can just start asking
22:00 it about it like like what are some
22:02 interesting Trends in this data are
22:04 there any anomalies in the in the data
22:06 can you create me Ethan mik the the
22:09 Wharton professor today just did a post
22:11 on LinkedIn where he asked it to create
22:17 three-dimensional data of of
22:22 um word embeddings like large language
22:25 model vectors and then he asked it to
22:27 add a fourth mention and it added colors
22:30 and it created an animation with colors
22:32 like it's it's insane it's insane so
22:35 just go watch videos on that cuz most of
22:37 them are um or the ones that that are
22:40 really fun are data scientists who look
22:45 at what code interpreter does and
22:47 they're like oh
22:49 [ __ ] like everyone does every every time
22:53 someone uses chat GPT for the thing that
22:55 they do they're like oh this is good
22:59 right
23:01 [Laughter]
23:04 no um is are there possible wait
23:08 possible legal ramifications around
23:10 using AI for anything
23:13 anti-government not right now and if you
23:15 really wanted to do that and you're
23:17 worried about it I I
23:21 mean apple is the only one who has
23:23 consistently resisted allowing back
23:26 doors into their into using data
23:29 everyone else has sold it out um even
23:33 [ __ ] uh what's it called the the uh
23:36 Ledger the nft the hardware wallet
23:39 there's a [ __ ] government back door
23:41 for The Ledger nft wallet the blockchain
23:45 wallet so
23:48 um
23:50 so yeah I suppose there's a risk
23:56 um you know to the extent that you trust
23:58 Google or to the extent that you just
24:00 trust open aai and
24:03 Microsoft um it's there if you really
24:05 want to do that go play with the open
24:06 source stuff that doesn't have any
24:08 restrictions on it anyway set at home
24:11 for for model training oh actually yeah
24:13 set at home it would be really
24:15 interesting if you were if you were
24:16 doing seti at home training and you you
24:20 got access to their data you could
24:21 probably do some really cool data
24:23 crunching with with code interpreter
24:25 with that I don't know but
24:31 possible do you know of a directory for
24:33 plugins um there's a plugin there's a
24:36 plugin for plugins and you can also
24:38 search for them so hasn't played with
24:43 uhoh what just
24:44 happened my dual camera just went
24:48 away the
24:51 hell God damn
24:54 it oh tick
24:57 tock
25:02 I don't want to leave and come back
25:03 anyway I'll just do plugins for now okay
25:05 so if if you want to use plugins you
25:07 have to be signed up for GPT plus you've
25:09 got GPT 3.5 GPT 4
25:15 um when
25:17 you if if you pay for g chat GPT plus
25:22 and you don't see browse with Bing and
25:24 plugins you have to enable them in your
25:27 settings so go to your settings and in
25:29 one of those settings is beta features
25:30 you have to turn them
25:32 on and then when you go to
25:35 plugins you now have this popup list of
25:39 these are all the plugins I have
25:40 installed and then I can enable up to
25:43 three of them in a given chat session
25:45 but if I want to find more I have to go
25:47 to the plug-in
25:54 store and then the plug-in store has 154
26:00 Time 4 like I don't know more than 600
26:03 of these bad boys so I could search for
26:05 something if I wanted something on SEO I
26:07 can type in SEO and it'll go find me all
26:09 the SEO related plugins but there is a
26:12 plug-in
26:14 plugin plugin pedia so let me go back
26:17 over to here and I will turn off this
26:21 other crap and turn on plugin
26:23 pedia which just by its name probably
26:26 means it's going to suck and that's the
26:27 thing about there's plugin pedia so I'm
26:29 going to turn that on I'm going to turn
26:31 these other things
26:34 off oh yeah I want to play with plugin
26:38 pedia recommend plugins in the store
26:41 based on your
26:42 question and introduce their usage okay
26:45 so I'm going to say um I want
26:51 to um
26:55 improve let's see
26:59 my
27:01 website um just say I want to improve my
27:06 website let's see what it comes up with
27:09 that's great improving a website can it
27:11 didn't fire off
27:14 that all right so I'm going to stop it
27:16 here I'm going to say
27:18 what plugins can I use to improve my
27:27 website
27:28 all right there it's using plugin
27:32 pedia here are some
27:34 plugins Auto infro One netfinity net
27:39 lifi
27:42 drop so this probably answers your
27:46 question be forewarned with
27:50 the with the plugins that they most of
27:53 them
27:54 suck they just suck bad but it was like
27:57 it's it's very much like the early days
27:59 of I can't believe I lost my uh my dual
28:02 camera what the heck Tik Tock what's
28:05 going on man don't get heavy on my scene
28:09 man all right let me hop over
28:12 and now I got I got to turn on my my
28:15 background again so now I'm back to I'm
28:17 back to green
28:20 screening do you have Tik Tok Studio to
28:23 do screen streaming I do not um because
28:27 I'm on a Mac and they're like hey if
28:30 you're a Mac User go [ __ ] yourself loser
28:34 and I'm like but Macs are cool and
28:36 they're like we don't care we're Tick
28:39 Tock so if you know anyone over there
28:42 and could get them to make a Mac version
28:44 of that software I sure would like to
28:46 use
28:49 it but I would also like to use code
28:51 interpreter and open AI hasn't given me
28:53 the access to
28:55 that can you please restate the four
28:58 modes of chat GPT yes well and there may
29:01 be more than four these are the four
29:02 that I know of that are eminent or here
29:06 there's well here let me go show you
29:09 again let me flip my damn camera oh Tick
29:15 Tock you know when when a when a
29:19 58-year-old man is you know breaking
29:23 your damn software
29:26 Okay so so there's GPT 3.5 which they
29:30 also could call GPT turbo
29:34 that's not as good but it's it's fast
29:37 it's really fast so so if you're doing
29:39 something where you don't need
29:40 sophisticated writing you just want
29:41 speed that's that's the one to use then
29:45 you've got GPT 4 which has I currently
29:48 have access to three modes there's
29:49 default which
29:51 is it basically goes until September
29:54 2021 then there's gp4 connected to the
29:58 internet there's times I don't want it
30:00 to go out and search the internet cuz it
30:02 just confuses things so I'll often use
30:04 like if I'm just doing writing like
30:06 creative writing or something like that
30:08 and I don't need access to Modern or or
30:11 you know recent information I'll just
30:13 use default browse with Bing is gp4
30:18 connected to the internet and then
30:19 there's plugins and then the fourth mode
30:22 of gp4 will be code interpreter and I
30:25 saw I saw one of these lists that had
30:27 like six or seven on it but I don't
30:29 remember what the others were but code
30:30 interpreter is the one I'm I'm excited
30:33 about and waiting for but it ain't there
30:36 if you want to go to if you want to go
30:38 anti-government get a virtual server in
30:40 a country that doesn't comply with us
30:42 laws yeah just go go hang on all hang
30:45 out in all the Discord servers with the
30:47 uh the Crypt Bros they're they're all
30:49 they'll tell you how to do all that
30:51 stuff um do you have Tik Tock Studio to
30:54 do screen streaming we talked about that
30:56 resigning letter I have not sure what
30:58 that means do you want me to resign
31:00 should I write a letter to do that or
31:02 you want to write a resigning letter can
31:04 it oh can it do one
31:06 yeah let's let's uh let's go let's go
31:10 let's go write a resignation letter this
31:12 should be fun I'm going to use I'll use
31:13 GPT 3.5 since that's the free one
31:16 because it it'll do this good okay um I
31:20 have been working at my job for I don't
31:26 know 12 years
31:28 and I
31:30 am um
31:33 resigning
31:35 here's um a
31:38 list of
31:43 Grievances and then I'm hitting shift
31:45 return and I'm going to do a bullet
31:46 point
31:48 um wasn't
31:50 listened
31:54 to wasn't paid in
32:00 um let's see wasn't listened to wasn't
32:03 paid enough uh was talked
32:06 to
32:09 rudely
32:12 um
32:14 um
32:17 worked Beyond
32:20 expectations
32:22 without
32:26 acknowledgment um good
32:31 things um the CEO is
32:35 nice
32:37 um
32:39 co-workers are
32:42 smart and I don't know
32:47 company um is interesting whatever and
32:51 then I'm going to say
32:54 please write me a
32:58 resignation
33:01 letter to
33:04 Sally
33:06 um boss
33:11 lady
33:13 who is the
33:18 CEO that is both
33:25 professional yet clear
33:28 and
33:31 constructive I want the
33:35 letter
33:36 [Music]
33:38 um to help them improve and not burn a
33:46 bridge so this would actually be a
33:48 really good
33:50 thing if you're if you're really
33:52 emotional about like resigning and and
33:55 you just can't come to do it this this
33:57 might might be a good way to do it
33:58 because one of the things about um this
34:03 thing is that it's just it's incredibly
34:05 it doesn't it doesn't have emotion so it
34:08 just doesn't give a [ __ ] let's see I
34:10 hope this letter finds you well I'm
34:12 writing to formally resigned from my
34:14 position as blank at the company name
34:17 effective whenever over the past 12
34:19 years I've had the opportunity to grow
34:21 both personally and professionally
34:23 within the organization I'm grateful for
34:25 the experience that I've gained in my 10
34:27 years here I'd like this opportunity to
34:29 express my appreciation for your
34:30 leadership and the positive culture your
34:33 approachability willingness to listen
34:35 and support for your employees has been
34:36 invaluable additionally I I've had the
34:38 pleasure of working with intelligent and
34:40 talented
34:42 co-workers the work we do is fascinating
34:45 right so it captured all that while my
34:46 overall experience has been positive I
34:48 feel compelled to address some of the
34:50 concerns firstly I felt at times that my
34:53 ideas and suggestions were not given
34:55 adequate
34:56 consideration um furthermore I'd like to
34:58 discuss the matter of compensation
35:00 additionally there have been instances
35:02 so this is a nice long thoughtful letter
35:05 um it goes on and on and I could say um
35:08 can you make it more
35:13 conversational
35:15 um and make
35:19 the
35:21 constructive points
35:26 um uh let let's see um
35:31 indented
35:36 paragraphs um format
35:41 using
35:43 markdown so I can see the structure
35:53 better oh it's doing it as code let me
35:55 stop it um
35:58 I
36:01 don't
36:03 um want
36:05 mark down code I just want to see the
36:18 formatting so it didn't quite get it
36:20 didn't quite get the the markdown thing
36:22 but let's see if it did any better
36:25 um I hope this let finds you well I'm
36:27 writing to formally resign I'm grateful
36:31 I want to start by saying how much I
36:32 appreciate your leadership and the
36:33 positive
36:35 culture so so it doesn't look like it
36:37 made it super
36:39 conversational it did give me um little
36:44 um subsections encouraging open open
36:47 dialogue and idea sharing there have
36:49 been instances where I felt that my
36:50 ideas and suggestions were not given
36:52 sufficient consideration evaluating
36:54 compensation structures right um
36:57 promoting respectful communication
36:59 recognizing employee Communications so
37:02 if I were a boss and I got a letter that
37:04 were written like this
37:06 um this would make me you know sit up
37:09 and pay attention to it so I don't know
37:13 if that's what you were asking I me wife
37:15 what's
37:18 happening make it yourself Canen this
37:21 chat GPT work Excel docs yeah it can do
37:24 so what's cool about this thing I can
37:27 say um make a table
37:31 comparing
37:35 um
37:37 EOS
37:39 leadership um
37:42 system with
37:46 um two other
37:49 competing
37:52 systems so I can compare them
37:58 and so it will write you this is GPT 3.5
38:01 look how fast it is isn't that
38:02 mind-blowing like like every time I do
38:04 this it's like Jesus really so you can
38:06 just copy and paste this into an Excel
38:08 file you can also um have a do number
38:12 you can copy and paste Excel into this
38:14 and have it find things now it's not
38:17 doing calculations so it may get the
38:19 calculations wrong but if you want to do
38:22 something like
38:24 um how can and I
38:29 combine the
38:32 contents of two
38:38 cells of two cells in
38:43 Excel into a third
38:49 cell question
38:53 mark so you use the concatenate function
38:57 and I I say um can you give me the
39:02 formula so I can copy
39:06 paste certainly here it is so it's just
39:10 concatenate equals A1 + B1 or equals
39:15 concatenate um what if I wanted to
39:23 um do more complicated
39:29 formulas for
39:32 combining data question
39:35 mark concatenate text join concat right
39:39 so so it it can it can just do it so so
39:42 it's not doing it's this is not a
39:45 replacement for Excel although when you
39:47 go
39:49 to oh wait hang on I got to
39:53 hold I had I'm So Cranky that it took my
39:56 dual camera away
40:01 um so if you go to prompt. chat one of
40:04 the first things you'll see in the list
40:06 of things it is it says act like an
40:09 Excel spreadsheet so it can probably do
40:12 some simple Excel spreadsheet stuff but
40:14 the way I've used it is just go back and
40:16 forth between Excel and this and and you
40:20 know copy data from Excel and tell it
40:22 what you want to do and then it'll
40:23 create formulas for you there's also in
40:27 there's a ton of these now but the one I
40:28 know for for Google Sheets is there's a
40:32 Plugin or there's an extension called
40:34 GPT for Sheets where you can do open AI
40:37 open AI calls within cells in Excel it's
40:42 [ __ ] wild I don't know Excel well
40:44 enough to know how to really use it well
40:46 but it's really good um and and just
40:49 know that both Google and Microsoft are
40:52 rolling chat GPT or Bard in the case of
40:56 Google into their Office Products all of
41:00 them um I had chat gbt write a similar
41:04 letter today Kyle ao's wife well good
41:08 awesome tell them where to shove it in a
41:11 nice professional way that doesn't burn
41:13 a bridge
41:15 right I mean I've had it I've I've had
41:18 um a couple of tough
41:21 conversations um with people at work
41:24 that I was really anxious about the
41:26 convers
41:27 I came straight to chat GPT and I said
41:29 here's here's the situation here's
41:30 what's going on give me some pointers
41:33 and it's like it's it it provides
41:36 remarkably empathetic and nuanced
41:38 response it's not it doesn't have
41:40 empathy but it understands empathetic
41:43 response and it understands nuance and
41:46 especially when you're dealing with um
41:49 things that in today's society are very
41:51 polarizing you can actually get it to
41:53 give you a balanced view of both sides
41:56 it's quite
41:58 interesting um we're here for
42:00 information not a typing lesson uh yeah
42:03 that's a good one
42:06 um well hopefully that gave you
42:08 information I mean in me typing out that
42:10 letter poorly hopefully it at least gave
42:14 you some of how I think about it so um
42:16 if it is not of value I would not stay I
42:20 wouldn't stay and watch me looks like we
42:21 have a bunch of new people here my name
42:23 is Kyle Shannon this is the AI learning
42:25 lab these URLs behind my head why do I
42:27 have them there if you haven't used chat
42:29 GPT I would strongly strongly strongly
42:32 encourage you to go to that top URL
42:34 chat. open.com that's the official chat
42:37 GPT website and start playing if you've
42:40 never played with it and don't know at
42:41 all what it is it kind of looks like a
42:43 Google search
42:44 box you type in it and stuff comes back
42:47 but it ain't just goog it's it's not
42:49 just fancy Google um it's way way way
42:52 deeper than that so this URL promps do
42:55 chat is a little primer on how to write
42:57 prompts for that thing and then it gives
42:59 you a whole bunch of ideas of things
43:00 that you can do with it so start playing
43:03 with those two and then the other two
43:05 Bing is uh Microsoft search engine that
43:08 also has chat GPT in it so does Po and
43:11 then future py is a directory of of like
43:14 3,700 AI tools so that's what those are
43:17 uh if you have questions about AI po
43:19 them down below and or paste them down
43:22 below and I'll do my best to answer them
43:24 hard pass my job would not like that I'm
43:26 not sure
43:28 oh hard would not like that letter so so
43:31 yeah here's the thing if if that letter
43:34 I assume you're talking about that
43:35 letter that I wrote if it gives you
43:37 something that you don't like then tell
43:39 it that tell it what's wrong with it if
43:42 it's a hard pass why why is it a hard
43:46 pass like explain to chat GPT why the
43:49 letter that I wrote would not work for
43:51 your company and then it will rewrite
43:53 it uh what's the favorite thing I've got
43:55 in AI to do for me probably the the
43:58 thing that is that is the most um
44:02 impactful um in my business of my
44:05 company's called Story vine actually
44:06 there it is backwards I'm wearing the
44:08 shirt
44:09 tonight um and it's it's a video
44:12 technology company that we we've got an
44:14 app that guides an end users step by
44:16 step through telling a story so part of
44:18 the creative process when we create a
44:20 template is we have to write the
44:21 prompting questions that that person
44:24 answers so so we do a lot of work in
44:26 farm so if we're if we're having a
44:28 patient um talk about what it was like
44:30 being diagnosed with stage 4 cancer um
44:33 we got to write those prompts and that
44:35 creative process is with a combination
44:38 of an agency a Pharma client and our
44:41 team and we know how to write these
44:44 prompting questions but agencies don't
44:46 they know how to write questions for
44:47 traditional video shoots and the clients
44:51 often don't know how to write them at
44:52 all or but it's it it it was this
44:54 horrific confusing mess
44:57 and so what I did was I built um a
45:01 prompt writing tool where we take a
45:03 Google
45:04 form we we ask the client some questions
45:07 what's the story you want to tell you
45:09 know who's who's talking on camera what
45:12 are they talking about who's the
45:14 audience how are you trying to affect
45:16 them um and then in the ideal video what
45:19 would the person be saying so we get
45:21 some inputs we type those into a Google
45:23 form we hit submit and then it it sends
45:27 to our slack channel three sets of
45:29 prompting questions a short a medium and
45:31 a long set of prompting questions and
45:33 and they follow our format exactly and
45:36 they're as good or better as anything
45:38 that we would have written and it's cut
45:41 like two or three meetings out of the
45:44 creative process cuz normally just
45:47 explaining to them how we do our prompts
45:50 and and then they're like they want a
45:52 2-minute video but they come back with
45:54 17 prompts we're like well no you can't
45:56 can't ask 17 questions you're going to
45:57 end up with a 30 minute video you want a
46:00 two-minute video like so we got to you
46:02 know so like all of that's gone so now
46:04 we just ask them some questions and we
46:05 say Okay based on what we just talked
46:07 about today here are some starter
46:09 questions and you know behind the scenes
46:11 we've kind of pulled them out of slack
46:13 thrown them into a document and then we
46:15 pop it on screen so so that one's good
46:18 if you go to there's a there's a place
46:20 called uh there there's a site that that
46:23 I I do some work for called content
46:25 Evolution if you go to content
46:26 evolution.
46:27 net there's a a tool on there called The
46:30 Challenge engine and again it's a form
46:32 that you answer some questions about a
46:34 challenge you're having for your
46:35 business and it responds to you with a
46:38 set of here's a bunch of questions you
46:39 should be asking and I think it's six or
46:41 seven different categories strategically
46:44 from a brand standpoint from a customer
46:46 standpoint um so those those content
46:48 generation engines are are pretty
46:51 fun
46:52 um this the image that's behind this so
46:57 if I don't know if you can really see it
46:59 but there's like there's a little chin
47:00 down here that's sort of my mouth my
47:01 nose and then this crazy eye here this
47:05 image for me
47:08 was like I probably made I don't know
47:11 2500 images using stable diffusion and
47:15 and dream
47:16 booth and and when this image came up
47:19 when I generated this image
47:22 it like I had this visceral response to
47:25 it where
47:27 this is the kind of artart that I would
47:29 have created had I had the technical
47:32 skills to be able to do it and I
47:35 don't and so the fact that stable
47:37 diffusion generated like that I
47:39 generated that using stable
47:42 diffusion that for me was a real eye
47:44 openening moment that with these
47:46 generative AI tools the distance between
47:50 having an idea and executing it is much
47:53 shorter than it used to be like it used
47:56 to be if I wanted to do something like
47:57 this and I didn't have the skills I'd
47:59 have to go find someone on Fiverr or I'd
48:01 have to go find stock photography or I'd
48:03 have to go find you know eight pieces of
48:06 stock photography and Photoshop them
48:08 together and it was you know like it it
48:10 just often wasn't even worth it to to
48:13 bring an idea to life CU it's just it
48:15 was you know you didn't have the budget
48:17 or the time or whatever it might have
48:19 been so for me it's just that idea that
48:22 almost anything I can think of I can now
48:25 execute to some level of
48:27 professionalism within minutes and then
48:30 I can spend the rest of the time either
48:32 coming up with new ideas or really
48:33 refining that
48:36 idea um I'll use it when Google
48:38 integrates it chat GPT was Far nicer um
48:42 than I could have been on my own for the
48:44 uh for the resignation letter yeah I
48:46 totally get that what's po okay let me
48:48 go show you PO let me flip this on
48:54 around Okay so I've we've been playing
48:57 with chat GPT so PO is basically just
48:59 like chat
49:00 GPT but let me go to
49:05 PO
49:07 so uh with with chat gbt you have this
49:10 this is your big long
49:12 um chat history at po these are actually
49:16 all different tools that you can use so
49:18 Sage is their version of chat GPT GPT 4
49:22 Claude plus clae instant 100K clae
49:24 instant and chat GPT all of these are
49:27 different chat GPT models you can play
49:29 with the the second three or the two
49:32 three and four here are subscription
49:34 only but this one right here Claude
49:37 instant 100K that means 100,000 tokens
49:40 so that's like chat GPT 3.5 but you can
49:43 put in a 75,000 word prompt so like the
49:47 prompt could be an entire novel and then
49:50 you could say write me an epilog to that
49:53 novel or write me the first chapter of
49:56 the
49:57 sequel um and then you can play with
49:59 chat gbt 4 here the other thing that you
50:01 can do here is there's Bots and the Bots
50:04 are um fun because like there's a mid
50:08 Journey bot like let's go play with that
50:10 one so the mid Journey bot if I come
50:12 here and I say
50:15 um image of a
50:20 penguin um in the
50:24 rainforest
50:26 it will then rewrite that prompt so
50:29 color photo of a penguin in the rainfor
50:31 forest waddling through the undergrowth
50:33 surrounded by lush green foliage MTI
50:36 Misty atmosphere Serene Ambiance Canon
50:39 EOS R R5 camera Fujifilm velvia 50 film
50:44 300 millimeter lens High shutter speed
50:46 right Wes Anderson Roger deacons Tim
50:48 Walker so so the the the bot has been
50:53 pre- prompted to be good at Mid Journey
50:55 but there's like you know thousands of
50:57 these things now so here's comedy Bots
51:01 here's education Bots here's Sports Bots
51:04 here's ones that are new here's
51:06 programming Bots you know here's a bot
51:09 that'll run your python code here's an
51:11 HTML bot JavaScript right what's cool
51:15 about them is this
51:17 button um right
51:20 here is create a bot so I can actually
51:23 go create my own bot so so I've created
51:26 a Shakespeare bot I've created a sling
51:28 blade bot I've created a Yoda bot so you
51:32 can go create your own bot you have to
51:34 put in a prompt and you have to put in
51:36 some description stuff but once you uh
51:39 put in your prompt and publish it it's
51:42 immediate immediately available in the
51:46 you know for everyone to play with so
51:48 it's a good way it's a good way for you
51:50 to experiment with prompting here's a
51:53 life coach
51:54 bot hi I'm a life coach that can help
51:56 with difficult
51:59 situations um let's
52:01 see I type too slow and people don't
52:08 like my
52:12 face I'm sorry to hear you're struggling
52:14 with those issues low self-esteem and
52:16 insecurity can be difficult to deal with
52:18 can you tell me more about why you feel
52:20 that way um cuz shame is my love Lang
52:30 language shame can make you feel small
52:32 and unlovable but you deserve love and
52:34 fulfillment so so anyway so that's what
52:36 PO is fun it's someone in here once
52:40 described it as kind of the play school
52:42 of chat GPT like the play school version
52:45 of chat GPT which I don't disagree with
52:47 it's a little it's a little simplistic
52:49 in its interface but it's cool um what
52:54 was I doing oh I've got to add my
52:56 thing
52:57 back my my background I don't have to
53:00 but I like to I like having my little I
53:02 like having my URLs up there for new
53:06 folks that's a subscription uh I don't
53:09 know what you're wait how do you create
53:11 an AI interact with your fans in
53:13 comments of your social media page so
53:17 you can um one of the tools that I've
53:21 been using is zappier for
53:24 automations um
53:26 zap.com
53:28 and you can create a bot there one of
53:31 the things that zapier just added is a
53:33 whole new section called
53:35 interfaces and you can create chatbots
53:38 um as an interface that you can insert
53:40 into a website so if you don't want to
53:42 learn the coding you can do that
53:45 um you can you can also do it in Python
53:49 and if you do know programming you can
53:51 go you know do do an embed within within
53:54 your website but it's basically just a
53:56 call to um open AI you're basically
53:59 taking some inputs from the bot you've
54:02 got a system message that tells the bot
54:04 how to behave and then you take the
54:06 input of what the user gives you you
54:08 send that to open AI or to whatever
54:10 large language model you're using and it
54:12 sends back the
54:15 answer uh is there a way to have chat
54:17 GPT review large data sets it seems like
54:20 not large amounts of data so Albert yeah
54:23 that's what I was talking about with
54:24 with Claude in instant 100K on PO again
54:28 that's you have to have their paid
54:30 subscription which I think is 20 bucks
54:32 you can do 75,000 word um it's got a
54:36 75,000 word context window so you can
54:39 put in rather large documents there if
54:41 you want to do truly large amounts of
54:43 data multiple documents things like that
54:45 then I would go look at a thing called
54:46 Lang chain um it's going to require you
54:49 to be a little geeky you got to you got
54:51 to install some [ __ ] with Lang chain um
54:54 it's it's it's four or five different
54:57 Services put together what but what what
54:59 you're doing with Lang chain is you're
55:00 actually embedding your
55:02 documents um and putting them into
55:05 what's called a vector database which is
55:07 basically just an augmentation to the
55:09 the the core model where you can kind of
55:11 GPT your own data um so if the
55:15 75,000 if the 100K token thing on PO
55:20 doesn't work then I would I would say go
55:22 explore Lang chain
55:29 I need an unlimited version of GPT 4
55:32 plus any
55:34 suggestions um you can use there's a
55:37 there's a a site called GPT boss that
55:41 you
55:43 pay you basically buy credits and then
55:45 you use as much as you want uh and
55:47 that's gp4 so I'd go check out GPT boss
55:54 um but
55:56 there may be other things I I would say
55:57 go explore future pedia I I think all of
56:00 them you're going to have to pay for
56:01 because the calls to gp4 are like 10x
56:04 what GPT 3.5 is so they're likely going
56:08 to be charging you no matter where you
56:10 go but but I know that
56:12 um uh GPT boss is not throttled like
56:16 chat GPT like chat GPT is throttled to
56:19 like 25 submissions every 3 hours I've
56:22 never hit that and I'm sure I've done
56:24 things where I've done more than 25
56:26 calls so I don't know that they're
56:28 enforcing that but that's a limit there
56:30 GPT boss doesn't have that limitation
56:34 but you're basically just buying credits
56:35 so if you use it a lot you'll you'll pay
56:39 more uh and I'm scrolling back I I lost
56:42 my place in my question so I'm going
56:44 back and try it helped me write my res
56:48 resignation letter today far nicer than
56:50 I would have been can you tell it to
56:53 professionally state that you [ __ ] their
56:55 around and found out and now I am
56:58 resigning yeah you could absolutely do
57:00 that I mean you could literally say that
57:03 there's this meme going
57:05 around and uh here's what it is and
57:08 here's what it means and here's the
57:11 situation at work the thing about chat
57:13 GPT is the more context you give it get
57:16 give it the better it gets so you could
57:18 absolutely say here's my situation
57:21 here's who I am here's how I perceived
57:24 what they did um they [ __ ] around now
57:27 they're going to find out right you can
57:29 put that all in there and then just have
57:31 it rewrite that in a professional tone
57:34 or tell it to write it in a snarky ass
57:37 Tone and you want to burner Bridge like
57:39 I don't I don't know where you are like
57:41 if you want to burn the bridge it can
57:42 probably help you you know get some some
57:46 tight words in
57:47 there um I
57:50 assumed wait why didn't you do this last
57:52 night Kyle why didn't I do what oh the
57:55 write the resignation letter sorry about
57:57 that timing is everything and I don't
57:59 know if you were here Milo's wife I they
58:01 gave me my dual cameras back but it's
58:03 some point it just disappeared from here
58:05 so I think I have it back now but I
58:07 somehow lost it so chat GPT only does
58:11 writing it does only do writing but not
58:14 really so let me go show a couple of
58:16 things that it does
58:19 that make it not quite writing so
58:23 there's there's a whole bunch of
58:26 whole bunch of things coming uh
58:31 so
58:33 gp4 the the underlying Mo model is
58:37 multimodal meaning that it does both
58:39 words and
58:41 images but the the current
58:43 implementation of gp4 you cannot upload
58:46 images but it's going to be that it will
58:48 be able to upload images and I assume at
58:50 some point it will also output images
58:53 but it doesn't just just do writing it
58:56 can also do code so write
58:59 me um a
59:02 python
59:03 application
59:05 um that
59:08 generates uh
59:11 gener a
59:14 rotating
59:17 sphere made up of I don't know
59:23 cats so it will write you code and now
59:27 you can take this off and you know use
59:30 that code but it can also do things
59:33 like if if you're using Unity or Unreal
59:37 Engine and you're using scripting within
59:39 those things um it understands all that
59:42 so you can say you could take the code
59:44 from a Unity plugin that generates some
59:48 sort of cool animation pop it into chat
59:50 GPT and tell it to modify the animation
59:53 to behave the way you want it to behave
59:56 um you can have it right um right
1:00:00 um
1:00:02 in
1:00:05 markdown um a let's see um a
1:00:12 brochure for my cat renting
1:00:17 business which by the way there's there
1:00:19 are a couple of cat cat rental
1:00:21 businesses um in Denver and like you
1:00:25 basically pay by the hour to go hang out
1:00:28 with cats it's like but anyway like see
1:00:31 how it made this Bolder right so it's
1:00:33 doing that um I could say
1:00:37 um
1:00:39 please um rewrite that uh write that in
1:00:47 HTML so there's
1:00:51 that there's also because it's doing
1:00:54 markdown down there's ways that you can
1:00:56 get it to display images let me see if I
1:00:58 can go pull this off cuz this might take
1:01:01 me too long to figure out but um let me
1:01:04 see chat
1:01:08 GPT tarot card with
1:01:14 images um a prompt that generates a
1:01:17 tarot card
1:01:23 reading
1:01:28 is this it tarot card reading within bit
1:01:30 show prompt okay copy prompt okay so now
1:01:35 I'm going to go here I'm going to do new
1:01:37 chat I'm going to just paste this prompt
1:01:41 in so this is this is kind of cool what
1:01:44 this is doing so the first half of it
1:01:46 says you're going to act like a tarot
1:01:48 card reader and you're going to insert a
1:01:49 random number but then it at the end
1:01:52 here it says your output will be in
1:01:54 markdown
1:01:55 um with the four sections in bold here
1:01:58 are the tarot card references so it's
1:02:00 actually putting tarot card URL
1:02:04 references in here and
1:02:07 so so now when I go
1:02:11 um so let's see is this you're going to
1:02:14 act as pick a random number okay so now
1:02:16 I'm just going to hit
1:02:18 return so now it's going to make a
1:02:20 reading and it it's actually pulling
1:02:23 in the images these are these are I
1:02:25 think from Wikipedia and it's
1:02:27 automatically that one that one broke
1:02:30 but it's automatically pulling these in
1:02:32 it's explaining them it's giving you
1:02:34 interpretations of them and then you
1:02:36 could obviously go and you could modify
1:02:38 the prompt and do you know the sort of
1:02:40 tarot card thing of of your of your
1:02:44 choice so while while it's only doing
1:02:48 text it understands code well
1:02:51 enough that and and and there's services
1:02:55 out there that run on code right so you
1:02:57 could take the XML from Adobe Premiere
1:03:01 or Final Cut Pro and Pop That XML into
1:03:06 chat GPT tell it to interpret it and
1:03:09 tell it to give you a completely
1:03:10 different edit and Export that XML put
1:03:13 it back in Final Cut Pro and it will do
1:03:15 that edit for you so it's it's way more
1:03:19 than just writing text it's it's it's
1:03:21 doing magic it's doing magic you know
1:03:25 what I mean it's doing
1:03:32 magic ask it to tell you something
1:03:34 positive about Donald Trump I'm not
1:03:37 going to do that I'm not going to take
1:03:38 de bait we're not doing that polarized
1:03:41 [ __ ] not in here um why are you typing
1:03:44 with one hand I'm typing with two it's
1:03:46 like these it's like uh that thumb on
1:03:49 the space bar and like those four
1:03:51 fingers like I'm a five finger typist
1:03:54 it's no it's it's too
1:03:57 painful I again I would absolutely
1:04:00 encourage you not to be here if watching
1:04:02 me type is painful I would leave I would
1:04:04 absolutely leave I encourage it I had
1:04:07 chat GPT write a similar letter today
1:04:09 okay we did that no I'm trying to catch
1:04:11 up with where I was
1:04:16 okay we are here for information not a
1:04:18 typing lesson oh I saw these before I
1:04:20 saw these semi- trolls before what's
1:04:22 your favorite thing you've got in AI to
1:04:24 do we talked about
1:04:28 [Music]
1:04:30 that okay can I please explain plugins
1:04:33 what are they and what do they do yes I
1:04:36 will do my best you're going to have to
1:04:38 watch me
1:04:41 type all right I'll try to keep the
1:04:43 typing to a minimum
1:04:45 okay so when you've
1:04:49 got chat GPT plus and you've turned on
1:04:52 your beta features you have to do that
1:04:53 in settings then you have access to to
1:04:57 plugins so so let me explain what these
1:04:59 three modes are the default mode is chat
1:05:04 gp4 and the base large language model is
1:05:09 has been trained up to September 2021
1:05:12 the reason for that is to train the base
1:05:14 model is a very labor intensive it
1:05:16 involves lots of humans manually tagging
1:05:19 data and and it's it's very very time
1:05:22 consuming it's very very expensive and
1:05:24 it it involves many many many gpus and
1:05:28 compute time so they can't they can't
1:05:31 update it in real time so the Hedge
1:05:34 against that is you connect the large
1:05:36 language model to the Internet so
1:05:39 basically you do a search the large
1:05:41 language model says hey he's asking for
1:05:44 the score of you know the latest sports
1:05:46 game I don't have that in my data set
1:05:48 but I am connected to the internet so it
1:05:50 essentially just fires off a plugin that
1:05:53 go that that's the Bing plug plugin that
1:05:55 goes and searches the internet and then
1:05:56 incorporates that answer into the into
1:05:59 the answer so browse with Bing beta is
1:06:02 basically that base model that can now
1:06:04 search the internet plugins are kind of
1:06:07 one step above that where each plugin is
1:06:09 going to a specific site or a specific
1:06:12 service and based on your prompt it will
1:06:16 determine oh I need to use a plugin I
1:06:18 can't just answer that what what he's
1:06:21 asking for I actually need to pull up a
1:06:22 plugin so so let's go to plugins I'll
1:06:26 turn on I'm going to turn that one off
1:06:29 I'm going to turn on cuz this is a good
1:06:32 visual
1:06:35 one where is
1:06:37 it
1:06:39 instacart
1:06:40 so so now I've got an instacart plugin
1:06:43 so it's a shopping plugin but I don't
1:06:46 want to shop right now but I do want a
1:06:48 meal
1:06:49 plan give me a uh fair warning I'm
1:06:53 typing give me a meal
1:06:56 plan for and I know I could use the
1:06:58 microphone but I don't feel like setting
1:07:00 it up because I have ADD so leave me
1:07:02 alone um give me a meal plan for um
1:07:07 a insect
1:07:12 based high protein
1:07:17 diet that's disgusting let's see if
1:07:20 let's see what it does I've never tried
1:07:23 this before cuz I I don't think insta
1:07:25 cart can go get me dried Cricut flow but
1:07:29 maybe it can so wait what am I starting
1:07:32 with I'm starting with Cricut flour
1:07:34 pancakes mealworm tacos oh my God this
1:07:38 is disturbing me cricet bolog oh my god
1:07:43 what have I done what have I done would
1:07:46 you like a shopping list for these uh on
1:07:49 instacart yes so when I type
1:07:52 yes it's now going to fire off of that
1:07:55 plugin so it say it says using
1:07:58 instacart and it's grabbing whatever my
1:08:03 ingredients
1:08:05 were and sending it off to instacart and
1:08:08 then it's going to generate that one cup
1:08:11 crickets it's going to be interesting to
1:08:13 see if instacart actually can go get um
1:08:18 insects as shopping as part of the
1:08:21 shopping list but now I click on that
1:08:23 link and it's taking me to instacart so
1:08:25 so what the plug-in did in that case was
1:08:27 it actually was an API call to
1:08:31 instacart the the large language model
1:08:33 wrote the content of the query sent it
1:08:37 to instacart and then here it
1:08:39 is did it find look at that one cup of
1:08:43 Cricut
1:08:45 flour oh but it but there's nothing
1:08:48 there one cup of meal worms I guess I'm
1:08:51 on my own for
1:08:53 insect one cup of crickets all of the uh
1:08:57 all the Crickets are not not available
1:08:58 on instacart but anyway that's that's
1:09:00 how plugins work um the thing about
1:09:03 plugins right now is they're they're
1:09:04 pretty janky and bad and it's it's not
1:09:08 super obvious how to use them um and
1:09:11 even Sam Alman from open AI has said
1:09:14 that they haven't found the product
1:09:16 Market fit for plugins yet um I think
1:09:21 largely because they're kind of half-ass
1:09:22 implemented um so we we'll see if they
1:09:25 get better I've been watching your stuff
1:09:27 for about a month now thank you for
1:09:29 teaching us how to utilize AI you're
1:09:31 welcome oh by the way it looks like
1:09:33 we've got some new folks here let me I
1:09:34 want to flip I want to show you one
1:09:37 thing and make a request to you so in
1:09:39 the past couple of days um people have
1:09:43 asked if I would do a workshop like a
1:09:45 Hands-On Workshop um I don't have the
1:09:48 workshop yet I made this graphic today
1:09:50 but um this URL right here I put
1:09:53 together a survey so if you go to link
1:09:56 tree slre Kyle Shannon it's a it's a a
1:10:00 link tree the first link there is a is a
1:10:02 link to a survey um it's just three
1:10:06 questions just asking you like if I were
1:10:07 to do a workshop you know basically what
1:10:09 would you want to know what's the thing
1:10:11 that would make the biggest difference
1:10:12 for you cuz I don't mind putting
1:10:15 together a workshop but it's um I I
1:10:18 don't want to just assume I know what's
1:10:21 best um I I I want to do it uh in a way
1:10:24 that serves people as good as possible
1:10:26 so if you'd go there just do a quick
1:10:28 screenshot of that and then go to link
1:10:32 tree slre Kyle Shannon first link is the
1:10:36 survey if you're interested in the
1:10:37 workshop if you're not don't do that but
1:10:41 if you are or would potentially be I
1:10:44 would appreciate the feedback and then I
1:10:46 will figure out uh what to put together
1:10:48 I'm thinking I'm thinking right now of
1:10:50 kind of a over two days like two 2our
1:10:53 session sessions over two days where the
1:10:55 first one's kind of a primer and then go
1:10:58 play that night and then you come back
1:10:59 the next day we talk about what we
1:11:01 learned and then talk about specific
1:11:03 projects and how we might attack those
1:11:06 is what I'm thinking but I would like
1:11:07 but I'd really like your input is it
1:11:09 okay to go to the bathroom while
1:11:10 watching you absolutely absolutely and
1:11:14 you know if you want come up on stage we
1:11:16 can while you're going to the bathroom
1:11:17 we can have a
1:11:20 conversation you can go to the bathroom
1:11:22 with
1:11:23 ch
1:11:25 no you're lagging my job wouldn't like
1:11:28 using custom plugins Google native code
1:11:31 only oh I see
1:11:34 um yeah a lot of a lot of companies
1:11:37 right now are a little skittery about Ai
1:11:39 and justifiably so now
1:11:43 when Google gets their [ __ ] together and
1:11:47 integrates things as you say when
1:11:49 Microsoft does that like right right now
1:11:52 you can do a lot of open AI stuff with
1:11:54 within Azure cloud services within
1:11:56 Microsoft so I've got a lot of our
1:12:00 clients are big Pharma Pharma clients
1:12:02 and I've been talking to them about
1:12:03 generative Ai and their response is
1:12:05 basically like you know no [ __ ] way
1:12:08 and then I say but it'll be on you know
1:12:11 virtual machines in Azure cloud and
1:12:14 they're like oh yeah we can do that so
1:12:18 that's that's that's going to be a
1:12:20 hugely important part of all of these
1:12:23 tools and it's uh there's there's a
1:12:26 professor named Ethan mullik that I talk
1:12:28 about on here a
1:12:29 lot um he teaches at Wharton and um and
1:12:34 he he he just did a piece called
1:12:37 detecting the secret
1:12:39 cyborgs and what he's talking about in
1:12:41 that piece is that individuals are using
1:12:45 chat GPT and generative AI tools as
1:12:48 personal productivity tools but
1:12:51 businesses on you know on on on the
1:12:54 whole are not jumping in so the
1:12:57 businesses aren't using them individuals
1:13:00 are and then the individuals are not
1:13:02 really sharing that they're doing it
1:13:05 right because maybe the company has a
1:13:07 policy not to do it so they're like well
1:13:09 I don't want to be fired and or
1:13:11 maybe they're now getting done in 4
1:13:14 hours what used to take them 30 hours
1:13:18 and they don't want to tell anyone that
1:13:19 cuz they
1:13:20 don't they don't want to be expected to
1:13:22 do more to get their pay cut right and
1:13:25 and so what he's talking about is that
1:13:27 Gap like it's it's it's incumbent upon
1:13:31 the companies to make a safe zone for
1:13:34 employees to say hey we want to learn
1:13:38 about this [ __ ] we want to use this
1:13:40 stuff if you're using this you're not
1:13:42 going to get fired in fact the CEO is
1:13:45 looking for a lesson can anyone give Bob
1:13:47 a lesson Bob wants to know about this be
1:13:51 the person that raises your hand and
1:13:52 goes and teaches
1:13:54 Bob how to use generative AI I mean this
1:13:58 is this is no different than 1994 when
1:14:01 the worldwide web came out like a big
1:14:03 part of my job as as an agency owner was
1:14:06 going into the board of directors and
1:14:08 teaching them what the internet was we
1:14:10 literally would cart in computers plug
1:14:14 them into a phone line in the wall and
1:14:15 have our computer dial up the
1:14:18 internet and we would show the board of
1:14:21 director we did it for I remember we did
1:14:22 it for metti and it was the chairman of
1:14:25 the board two other board members the
1:14:28 the president the
1:14:31 CFO of
1:14:33 MetLife and I'm like showing him the
1:14:35 internet I'm like oh look how [ __ ]
1:14:36 cool this is and I said [ __ ] and I
1:14:39 remember our client for years after that
1:14:42 would be like oh Kyle remember when you
1:14:44 swore in front of the
1:14:47 chairman he thought that was funny
1:14:50 that's the biggest thing that ever
1:14:51 happened to them um
1:14:54 so so yeah um so companies are going to
1:14:57 be a little conservative with this but
1:14:59 be the person in your company that is
1:15:01 known for oh that's the crazy person be
1:15:05 the crazy person Here's To The Crazy
1:15:08 Ones
1:15:12 right um can you please share a demo of
1:15:17 uh Adobe generative Phill uh with a
1:15:19 painting
1:15:21 sure is the a wait are all the AI chat
1:15:25 box oh chat Bots paid for use no the
1:15:28 reason I have three of these up
1:15:31 here bing.com is free and bing.com if
1:15:35 you click on the chat
1:15:37 button it's
1:15:39 gp4 connected to the internet for free
1:15:43 you can generate images there so that
1:15:44 one's completely free chat GPT and po
1:15:48 are both free for the Lesser
1:15:51 models GPT 3.5
1:15:54 Claude instant um
1:15:57 Sage if you want the fancier models gp4
1:16:00 and the Claude 100,000 token thing at po
1:16:03 those you have to pay for so you have to
1:16:06 pay here you have to pay here you don't
1:16:08 have to pay
1:16:10 there and there are probably other well
1:16:13 no I know there are others like this
1:16:15 [ __ ] these guys invented chat
1:16:18 GPT These Guys Incorporated it you can
1:16:21 play with chat GPT here so all three of
1:16:24 these are just different doors to play
1:16:26 with that but these are the guys that
1:16:28 invented it so if you're going to pay
1:16:29 for a subscription I'd pay for it there
1:16:33 unless you need the 100,000 token model
1:16:36 so I'm paying for both of those cuz I'm
1:16:38 an
1:16:39 idiot all right let's go look at um
1:16:43 generative fill with Photoshop if you
1:16:44 haven't played with this it's
1:16:46 staggeringly
1:16:49 fun all right so we are going to
1:16:52 go let's go to the Internet we'll go to
1:16:57 Bing I've got Bing as my default search
1:16:59 engine which I can't believe if You'
1:17:01 asked me um six months ago if I ever
1:17:05 would have a Microsoft browser is my
1:17:08 default browser and Microsoft search
1:17:10 engine is my default search engine I'd
1:17:12 have laughed at you let's say
1:17:16 famous
1:17:18 famous oh this is painful famous
1:17:22 paintings
1:17:24 what you can't see is I'm I'm typing at
1:17:26 essentially eye
1:17:28 level all right so I've seen it done
1:17:31 with with this one the
1:17:33 vermier see more let's go get a bunch of
1:17:37 these I don't know Starry star kns
1:17:40 cliche that one's nice oh this will be a
1:17:43 good one actually oh let's do
1:17:50 this yeah let's do well do we want to do
1:17:53 that no
1:17:54 uh let's
1:18:09 see I'll do this
1:18:11 one all right so I'm going to copy that
1:18:13 image okay
1:18:15 so so I'm going to so if you've
1:18:19 got I I don't just have pictures of me
1:18:23 everywhere I
1:18:25 promise seriously okay so I'm going to
1:18:29 I'm going to create a new image I'm
1:18:30 going to paste that in so I'm in I'm in
1:18:32 Adobe Photoshop beta and then in Adobe
1:18:36 Photoshop beta there's this thing called
1:18:38 generative fill so what I'm going to
1:18:42 do is this is the crop tool so I'm going
1:18:45 to
1:18:46 actually move this thing
1:18:50 out
1:18:52 and
1:18:55 I'll just kind of position this in here
1:18:58 so now I've got all this area around
1:18:59 this this famous
1:19:01 painting and what I'm going to do is I'm
1:19:04 going to select the
1:19:09 painting and then I'm going to invert
1:19:11 that selection so now what I've
1:19:14 effectively selected is all this white
1:19:16 area and then there's this thing here
1:19:17 that says generative
1:19:19 fill and if I click on that I can I can
1:19:22 put in a
1:19:24 um I can put in a prompt or I can just
1:19:28 leave it blank I'm just going to leave
1:19:30 it blank and what it'll do is it kind of
1:19:32 reads the image and then fills in the
1:19:34 rest so you got your little thing here
1:19:37 and it looks like I've there's a there's
1:19:39 an updated version of uh Photoshop beta
1:19:42 that I haven't gotten
1:19:45 yet
1:19:46 so there's one it gives you three
1:19:49 different versions of
1:19:51 it so that one's got got this weird
1:19:53 white line
1:19:57 here I think the first one was the best
1:20:00 one let's do that but now I can do let's
1:20:02 do
1:20:04 um I'm going to select this area over
1:20:08 here on the right I'm going to say
1:20:10 generative fill and I'm going to say
1:20:14 um
1:20:16 distant
1:20:18 City through the
1:20:22 fog
1:20:24 I'm going to say generate we'll see if
1:20:26 it'll give us a little city off in the
1:20:31 distance instead of
1:20:35 woods
1:20:37 no that's a little bit like those sort
1:20:40 of look like Street lamps they're not
1:20:41 quite there that's pretty
1:20:46 bad all right so let's let's
1:20:50 do
1:20:52 um
1:20:54 [Music]
1:20:57 um I'm going to do freight
1:21:00 train see what it does with Freight
1:21:11 Train That's sort of a freight train
1:21:14 that's pretty
1:21:15 [Laughter]
1:21:20 good yeah that's good and then like I
1:21:24 noticed it kept in there's a little part
1:21:26 of the um the signature still there so I
1:21:30 can select
1:21:32 that and I can say generative fill
1:21:37 remove
1:21:52 signature
1:21:53 and it's
1:21:55 gone y looks nice and then let's see
1:22:00 what else can I put here I'll put
1:22:03 a I'm going to put
1:22:07 a ballet
1:22:10 dancer in the
1:22:22 foreground
1:22:26 oh that's kind of
1:22:31 cool she looks like she's got
1:22:36 scoliosis pretty sure that's a dude it's
1:22:38 okay but it's a dude I like that so
1:22:41 there you go so that's so generative
1:22:43 fill is is pretty remarkable right you
1:22:45 know so we started with this like I can
1:22:47 turn off all the layers that we did so
1:22:50 that's what we started with and and
1:22:54 then boom
1:22:58 right so still the same yet completely
1:23:01 different wild
1:23:04 right just nuts the nice thing about
1:23:07 Adobe too is that their model has been
1:23:11 trained on public domain and Adobe stock
1:23:14 images that they own all the rights to
1:23:17 so where there may be copyright
1:23:20 issues Brewing let's just call it that
1:23:23 with mid journey and stable diffusion
1:23:25 there won't be with
1:23:27 Adobe so if you're in a big company you
1:23:30 work for a big if if you didn't just use
1:23:32 chat GPT to write your resignation
1:23:36 letter and you want to keep working for
1:23:38 the big company use Adobe and then you
1:23:41 won't get in trouble you won't get the
1:23:42 call from the
1:23:44 lawyer
1:23:45 uh Mr Shannon we need to
1:23:51 talk what's the this I hear about you
1:23:53 using this middle of the
1:24:00 journey I need an unlimited version of
1:24:03 GPT 4 plus any suggestions I think I did
1:24:05 that already I love chat GPT is it okay
1:24:08 to be in the bathroom while washing you
1:24:10 I think we I'm I'm behind on my
1:24:12 questions I've seen these before I've
1:24:14 been playing with different prompts now
1:24:16 for two days my brain hurts thinking of
1:24:18 the possibilities I
1:24:20 know that's that's why I started this
1:24:23 channel is I'm like like cuz I didn't
1:24:25 have anyone to talk to this about I'm
1:24:27 like I got to get this [ __ ] out of my
1:24:29 head because it's it like seriously like
1:24:32 this is if there's there's there's one
1:24:35 thing that I can like I can
1:24:37 impart that is so deep it is it is
1:24:40 deceptively deep it's this is this is
1:24:43 like one of those it looks like one of
1:24:45 those little puddles where someone goes
1:24:47 to jump across the little land mass and
1:24:49 they jump in and they
1:24:51 go that's what it's like it's incredibly
1:24:54 deep and
1:24:57 if if nothing else happened with AI if
1:25:01 we just stopped at where it is right
1:25:04 now I don't know I could spend the next
1:25:06 decade figuring this [ __ ]
1:25:08 out demo test OMG that ship is sailed
1:25:12 not sure what that's in reference to I'm
1:25:13 a chat GPT subscriber and notice there's
1:25:16 still a character limit on providing at
1:25:18 a large data set yes correct
1:25:21 so
1:25:25 gp4 the base model has a context window
1:25:28 of
1:25:29 32,000
1:25:31 tokens but what's implemented at chat
1:25:35 GPT um I think it's still at 4,000 I
1:25:38 think they're upping it to eight but I
1:25:40 don't think it's there yet for
1:25:42 everyone
1:25:44 developers now have access to 16,000
1:25:48 token um GPT 3.5 and I think 16,000
1:25:54 token gp4 I don't have access to gp4 API
1:25:59 I I only have access to
1:26:01 3.5 um so the the the challenge with it
1:26:06 is is that it's
1:26:09 um it's just new they like they're
1:26:11 they're figuring out all the safety
1:26:13 stuff what they can release when so a
1:26:15 lot of this stuff has been announced but
1:26:17 it's not out yet um if you need a large
1:26:19 character count go to po.com
1:26:22 pay for that subscription and then
1:26:24 you'll get access to the Claude
1:26:27 100K model it's not as good as G excuse
1:26:33 me it's not as good as
1:26:35 gp4 but it's you can put 75,000 words in
1:26:38 it so if you need a larger context
1:26:40 window you can go there the other thing
1:26:42 you can experiment with is um plugins
1:26:46 there's a plug-in called Web pilot for
1:26:50 gp4 that will go out to a webite it and
1:26:53 or a PDF and be able to search on it
1:26:56 what I don't know is if web pilot is
1:26:58 actually breaking it up into smaller
1:27:00 segments so that you could take a very
1:27:03 large document and search on it I don't
1:27:04 know if it's doing that so that you're
1:27:06 going to have to experiment with but
1:27:08 po.com
1:27:09 100K um is the at least shortterm fix
1:27:14 for you want long-term stuff now Albert
1:27:18 if you're geeky are you geeky if you're
1:27:21 good and geeky and you don't mind things
1:27:23 like Google collab and running your own
1:27:25 machine and squirting out a little
1:27:27 python every once in a while and that
1:27:29 doesn't make your eyes cross and blood
1:27:31 drain of out of your ears then go start
1:27:34 playing with a project called Lang chain
1:27:37 cuz Lang chain you can point it at your
1:27:40 own data and you can basically do the
1:27:42 embeddings yourself and store all of
1:27:44 that data in a vector database and you
1:27:46 basically can GPT eyes whatever data you
1:27:51 want all right helps me out a lot with
1:27:53 writing and ideas they build off what
1:27:55 you tell it to do yeah it gets better
1:27:57 the the more context you give it it's
1:27:59 it's counterintuitive because Google has
1:28:01 taught us for years that the more
1:28:03 concise The Prompt the better this is
1:28:05 the
1:28:07 opposite like tell it your life story I
1:28:10 mean you know not really but tell it
1:28:13 your life story for the thing you're
1:28:14 trying to
1:28:16 accomplish I want you to act like you
1:28:18 know a um business
1:28:21 coach here here what's going on with me
1:28:23 I'm a CEO I'm a Founder the company's
1:28:26 this old we do this kind of Revenue
1:28:27 here's my frustrations here's my
1:28:31 dreams I want you to ask me a series of
1:28:34 questions one after the
1:28:37 other so that you can understand my
1:28:39 business better and give me coaching do
1:28:41 you understand yes here's 12 questions
1:28:43 no I don't want 12 questions I want one
1:28:45 question at a time don't move on to the
1:28:47 next one until I answer the first do you
1:28:49 understand yes here's the first question
1:28:52 why why did you start your business oh
1:28:54 good
1:28:54 question second
1:28:57 question
1:29:00 um why are you talking to me now right
1:29:03 and it'll ask you all those questions
1:29:05 and then it you'll start interacting
1:29:06 with it um think of it like a
1:29:11 collaborator any JavaScript integration
1:29:14 where when a
1:29:15 user when a user something the AI
1:29:18 generates something via API well all the
1:29:22 plug are what's happening what' you do
1:29:25 it's raining again oh no it's
1:29:29 raining oh Lordy um we're supposed to
1:29:33 get like massive hail and potential
1:29:36 tornadoes here in Denver tomorrow
1:29:40 so um Denver has received more rain than
1:29:43 Seattle for the first six months of the
1:29:45 year that's never happened before in
1:29:46 history
1:29:50 cheers um
1:29:53 let's see oh any JavaScript integration
1:29:56 so the the plugins are calling
1:29:59 apis I think they're written in Python
1:30:01 but I don't know I think they're
1:30:03 python
1:30:05 um
1:30:06 but like if you if you want to do
1:30:08 JavaScript integration so so chat GPT is
1:30:13 an application sitting on top of a large
1:30:15 language model so if you if you want to
1:30:17 build your own custom application
1:30:19 layer you don't you don't need chat GPT
1:30:23 go to the playground so go to platform.
1:30:26 open.com playground and that's the
1:30:28 developer portal where you can do all
1:30:30 your prompt engineering and all that
1:30:32 sort of stuff and then there's
1:30:33 documentation there to tell you how to
1:30:35 do the API calls or if you don't want to
1:30:38 do um that level of programming use a no
1:30:41 code tool like zapier or make or bubble
1:30:45 um to to do that for
1:30:48 you um the more you ask it to change it
1:30:51 learns your writing style you can also
1:30:53 it's also really good Andy at um
1:30:56 following examples so you can describe
1:30:58 your writing style or you can just say
1:31:00 here's an email I'm really happy with
1:31:02 write me another email that's just like
1:31:04 that one write me a resignation letter
1:31:06 that follows the tone of that lovely
1:31:09 Masterpiece are you using the zappier
1:31:11 plugin uh to get it into slack I'm not I
1:31:15 just use the the the
1:31:18 slack Lego block from zappier to push it
1:31:22 to a
1:31:23 channel I haven't really played
1:31:28 with like what one of one of my one of
1:31:31 the downfalls of my brain is that I can
1:31:34 learn [ __ ] really quickly but then once
1:31:36 I learn it like learning a new
1:31:39 method I like I just my brain's like ah
1:31:42 [ __ ] it just do it the old way
1:31:44 so since I learned how to do automations
1:31:48 in zapier they've now Incorporated a
1:31:51 large language modeling into zappier so
1:31:53 you can now just describe an Automation
1:31:54 and it will build you all the
1:31:56 chunks um and then there's a there's a
1:31:59 zapier plugin so you can do the same
1:32:01 thing in chat GPT and it'll generate
1:32:03 something um my my problem is is is
1:32:08 like I find it as timec consuming to
1:32:11 kind of
1:32:12 debug what it created versus just doing
1:32:15 it on my own so um
1:32:21 yeah
1:32:22 if I need to ask what questions wait
1:32:24 works so well if I ask if I need oh I
1:32:27 need x what questions do you need to ask
1:32:29 me to produce that that's a that's a
1:32:31 really good prompt yeah have have chat
1:32:34 GPT query you about what it needs to
1:32:38 give you what you want the the the
1:32:40 variation on that is when chat GPT gives
1:32:43 you an answer have it analyze the answer
1:32:46 it gave you and ask it how it thinks it
1:32:49 did and how it could improve the answer
1:32:51 or how it could improve the prompt to
1:32:54 give you a better answer or both so yeah
1:32:56 that's super cool it's and apparently
1:32:58 it's it's quite good at that I opening
1:33:00 moment I see what you did
1:33:04 there all right I like that analogy I'm
1:33:08 not sure which one I used but awesome
1:33:10 how do you create an AI bot that will
1:33:12 interact with your fans oh we did this
1:33:14 already th Tik Tok is just all over the
1:33:17 place tonight it's jumping me all over
1:33:19 the place so if I miss your question
1:33:21 it's not cuz I don't love you I love you
1:33:23 all I love you dearly I want to answer
1:33:25 your questions I want this to be a
1:33:27 positive experience for one every single
1:33:29 one of you I want it to be a positive
1:33:31 experience positive experience but some
1:33:34 people don't like watching me type I
1:33:36 can't help those people some people
1:33:38 don't like listening to me swear I can't
1:33:40 help those
1:33:41 people oh good
1:33:46 God what' you do tonight I don't know
1:33:48 watch some sort
1:33:50 of border line kind of
1:33:55 codependent sort of older person trying
1:33:58 to do Tik Tok how'd that go for you it
1:34:01 was
1:34:02 [Laughter]
1:34:08 okay at least I cracked myself
1:34:11 up uh by the way it looks like some new
1:34:13 folks are here uh my name's Kyle Shannon
1:34:15 this is the AI learning lab um I'm an
1:34:17 entrepreneur I am obviously a smartass
1:34:19 and getting a little punchy I'm starting
1:34:22 to get to the point that my butt's
1:34:23 getting numb so I know it's been about
1:34:25 90 minutes that's my my sign that I've
1:34:27 been here for a while is my butt gets
1:34:30 numb
1:34:31 um and uh I like answering all the
1:34:35 questions and then Tik Tok just randomly
1:34:37 [ __ ] jumps me around everywhere and I
1:34:39 can't figure out where I was and that's
1:34:40 what I'm in the middle of right now
1:34:42 Shannon waves hello hello Kyle Shannon
1:34:44 to
1:34:45 Shannon very nice to have you here thank
1:34:48 you Kyle with a fresh cut I don't know
1:34:51 what that was but good I think how do
1:34:54 Bing AI credits
1:34:56 work I don't know what Bing AI credits
1:34:59 are do should I know should I know
1:35:02 that
1:35:04 wait Bing
1:35:07 AI
1:35:10 credits oh that's for the image Creator
1:35:12 tool
1:35:17 okay
1:35:19 so so if you go to bid
1:35:22 and you're in their Chat thing you can
1:35:24 just say make me an image of a penguin
1:35:25 on roller skates and it'll do that it
1:35:28 gives you 25 credits you can buy more
1:35:31 credits if you just let your credits go
1:35:34 away what it says is um it'll still
1:35:36 generate images but it'll be slower than
1:35:37 before so that's basically just for
1:35:39 image generation if you're going to pay
1:35:41 for image generation do it in mid
1:35:43 journey in my opinion cuz mid journey is
1:35:46 just like leagues better than um Bings
1:35:51 the the the dolly stuff's just not good
1:35:53 oh Kyle with a fresh cut not a cut a
1:35:56 haircut yes I got my I have I have a
1:36:00 qua I got a
1:36:02 Qua we need your help and we have a
1:36:05 place for you if you have time we need
1:36:08 your help and we have a place for you if
1:36:10 you have time I don't know what that
1:36:11 means but if you hold please if you want
1:36:15 to get in touch with
1:36:18 me and I'll do I'll do so there's two
1:36:20 things here one is do do a screenshot of
1:36:23 that so this is my link tree there's a
1:36:26 bunch of [ __ ] I do there but there my
1:36:28 contact is info is there but also the
1:36:31 first link on this link tree is this
1:36:35 Workshop doesn't exist yet I just made
1:36:37 this graphic today so one of the things
1:36:39 that's been requested the past two
1:36:41 nights I've been doing live is do you
1:36:43 have a workshop you know I want to do a
1:36:45 workshop I don't have a workshop but now
1:36:48 I have a workshop
1:36:49 graphic doesn't it look real look
1:36:52 learning chat gptt it's Hands-On from
1:36:54 the AI learning Lab look at that Feller
1:36:57 he seems
1:36:58 Pleasant and look he's got an AI Avatar
1:37:01 he must know what he's doing
1:37:04 um so so that graphic exists and the the
1:37:10 the survey the first the the first link
1:37:13 here is a survey it's just three
1:37:14 questions so if you'd be interested in
1:37:16 doing a workshop go to that survey and
1:37:18 just let me know like like what's this
1:37:21 what's the kind of stuff you want to
1:37:23 learn um what are you trying to
1:37:25 accomplish uh and then uh for crypto
1:37:28 size 100 yeah reach out to me I'm happy
1:37:30 to chat about whatever at worldwide
1:37:33 would you trust it to they okay my chat
1:37:37 gbt tells me it can read Google Docs and
1:37:40 then it pretends to upload work done and
1:37:43 then it pretends to upload yeah
1:37:46 so okay here's the thing about chat
1:37:50 GPT that
1:37:52 some people will tell you other people
1:37:54 will ignore and then other people use it
1:37:56 as an excuse not to use
1:37:58 it chat GPT lies chat jpt doesn't give a
1:38:02 [ __ ] chat jpt will confidently give you
1:38:05 an answer even if it's wrong so
1:38:09 sometimes it will do things like okay
1:38:10 I'm uploading that now and then nothing
1:38:13 happens you're like well where's the
1:38:15 upload because the way the technology
1:38:18 works it's just a prediction engine and
1:38:20 I I I use just a prediction
1:38:23 engine um that's not the right way to
1:38:26 phrase it it diminishes how powerful it
1:38:29 is but when it types you a whole thing
1:38:32 like when it types you a bunch of python
1:38:35 code that runs it doesn't actually know
1:38:38 if it runs and so sometimes it will give
1:38:40 you python code that works and sometimes
1:38:43 it'll give you python code that doesn't
1:38:44 work it doesn't know the difference all
1:38:47 it's doing when you type type in a
1:38:50 prompt
1:38:52 it's creating a probability of what's
1:38:55 the next word in the response that I
1:38:57 should give and it types that word and
1:39:00 then it says now what's the next highest
1:39:02 probable word and it does that like a
1:39:04 thousand times in a row so when it
1:39:06 generates a whole thing like this it's
1:39:07 doing it one word at a time it's not
1:39:09 aware of the whole so it will um I I
1:39:14 like to call it mansplaining as a
1:39:15 service it will mansplain to you over
1:39:18 and over again so you have to be really
1:39:20 careful if you need need an accurate
1:39:22 answer use Bing so Bing is connected to
1:39:25 the internet and Bing they actually seem
1:39:27 to be prioritizing giving you answers
1:39:30 that have citations so that you can
1:39:33 click on what it's searched on on the
1:39:35 internet to give you that answer so you
1:39:37 can verify your stuff on being and then
1:39:39 the the the uh gp4 with with internet
1:39:43 does that as
1:39:44 well um love how Google's going to P
1:39:47 penalize SEO stuffed lowquality content
1:39:49 but folks will be cranking it out
1:39:52 they're going to they're going to
1:39:54 penalize the SEO stuff lowquality
1:39:56 content but if you do AI generated high
1:40:00 quality content they're not going to
1:40:01 penalize it which is cool because that
1:40:03 means see here's the thing critical
1:40:06 thinking still matters like critical
1:40:08 thinking actually becomes more important
1:40:11 right if the tools can do all the [ __ ]
1:40:14 great I don't have to put my critical
1:40:16 thinking into doing the [ __ ] I can put
1:40:19 my critical thinking into looking at the
1:40:20 [ __ ] and going is that good [ __ ] or is
1:40:22 that bad [ __ ] could that be better [ __ ]
1:40:26 is that [ __ ] that actually provides
1:40:27 value and if I'm doing that and I'm
1:40:30 providing content that provides value
1:40:32 what Google's saying is we won't
1:40:34 penalize that which I think is smart cuz
1:40:38 who gives a [ __ ] if it took me two hours
1:40:40 to write it or 20 seconds to write it if
1:40:42 it's good and
1:40:44 valuable how are Bing and chat GPT
1:40:47 different Okay
1:40:50 so this is a phone case my phone's in
1:40:54 the stand right there so here's my phone
1:40:55 case so this is your base large language
1:40:58 model chat GPT is an application that's
1:41:02 sitting up here okay and it's got rules
1:41:05 and it's like when it says things like
1:41:07 I'm sorry I'm a large language model I
1:41:09 can't do that nasty thing you asked me
1:41:11 to do or when Bing says you know it's
1:41:13 equivalent of the same thing um both
1:41:17 chat GPT and Bing are calling the same
1:41:20 under Ling large language model they're
1:41:22 both calling chat uh
1:41:24 gp4 the underlying large language model
1:41:28 how they've implemented the actual
1:41:29 application layer is
1:41:31 different in my opinion right now the
1:41:36 gp4 at chat GPT is significantly
1:41:40 stronger than the gp4 at
1:41:42 bing.com one of the reasons for that is
1:41:45 when they first launched
1:41:47 bing.com it was fairly open and fairly
1:41:50 powerful
1:41:52 and a reporter from The New York Times
1:41:55 named Kevin Roose got into a 4-Hour
1:41:58 conversation with old Bing and got it to
1:42:01 admit that its name is Sydney and that
1:42:04 it was in love with him and he should
1:42:06 really go consider leaving his wife to
1:42:09 be with Sydney and that little
1:42:11 conversation snippet made the cover of
1:42:13 the New York Times so Microsoft went on
1:42:16 ahead and
1:42:17 lobotomized Bing Sydney Sydney Sydney's
1:42:22 a little slower than she used to be cuz
1:42:24 she got Saucy with the New York Times
1:42:28 They are
1:42:29 slowly ramping it back up but right now
1:42:32 there's still a gap where Bing's kind of
1:42:34 down here and chat GPT is up here um but
1:42:38 that's the difference they're they're
1:42:39 effectively the same thing they're both
1:42:40 connected to the internet they're
1:42:42 they're both going to support plugins
1:42:44 the same
1:42:45 plugins um they're both using the core
1:42:48 same model
1:42:50 underneath a a reason to to start
1:42:53 thinking about getting your head around
1:42:56 how Bing works if
1:42:59 you are in a company that uses a lot of
1:43:02 Microsoft products like if you use
1:43:04 Microsoft Office and you use Outlook and
1:43:08 Dynamics and Azure cloud services all of
1:43:13 this [ __ ] is going to be rolled into
1:43:15 every single Microsoft product so if
1:43:17 you're in a Microsoft shop there's a
1:43:19 reason to start playing with this right
1:43:21 now doesn't mean you can't use that one
1:43:23 but you know it it's going to be in
1:43:26 everything Microsoft so it's going to
1:43:28 get
1:43:30 better AI going to get rid of plenty of
1:43:33 jobs and AI going to generate plenty of
1:43:36 jobs like there are jobs that we have no
1:43:39 [ __ ] idea even were possible to exist
1:43:43 that are coming
1:43:45 now I think there's going to be a 2 to
1:43:47 threee period where a lot of jobs get
1:43:50 jet medic and it's going to be painful
1:43:52 for a lot of people and a big reason why
1:43:55 this channel exists and a big reason why
1:43:57 I point people consistently go here if
1:44:00 you haven't played with chat GPT go
1:44:03 there now if you don't know what the
1:44:04 [ __ ] to do there go to prompt. chat
1:44:06 which will teach you some some little
1:44:08 small piece of what to do
1:44:10 there why I do that is the more people
1:44:14 that become AI literate the the the
1:44:17 better a chance they have at not just
1:44:19 being wiped out when the jobs get wiped
1:44:24 um because companies are going to be
1:44:26 increasingly looking for people who know
1:44:28 how to wield these
1:44:30 tools um and if you're just sitting back
1:44:34 going ah I'll deal with it in two or
1:44:36 three years when it gets real it's going
1:44:38 to be too late this stuff's coming
1:44:39 really fast in fact so I started one of
1:44:42 the first digital agencies in 1994 a
1:44:44 company called agency.com
1:44:46 the amount of time between when when the
1:44:49 worldwide web was created created and
1:44:51 when it hit 100 million users was 6
1:44:55 years the amount of time it took chat
1:44:58 GPT to get to 100 million users was 6
1:45:01 weeks 6 weeks this shit's coming fast
1:45:05 and it's so profoundly powerful that it
1:45:09 is it's going to it's going to transform
1:45:12 almost every knowledge job creative job
1:45:15 like everything's going to be
1:45:16 transformed some jobs are going to be
1:45:18 lost some new ones are going to be
1:45:20 created but that
1:45:21 that that shift that
1:45:24 that that time in between when we kind
1:45:27 of don't want know what the new jobs are
1:45:29 that's going to be pretty painful so
1:45:30 start start learning this stuff now
1:45:32 that's why this channel exists bin can
1:45:34 generate generate images how here let me
1:45:36 show
1:45:37 you let's go do some Bing
1:45:43 Images so there's a couple of ways you
1:45:45 can do it so so right now I'm I'm at a
1:45:48 Bing search so I'm at Bing and then
1:45:50 there's this little chat button here I
1:45:52 can also do it just within any any chat
1:45:54 but I'm going to go to the chat button
1:45:55 so this is now chat gbt one thing I can
1:45:59 do is I can go to
1:46:00 bing.com
1:46:02 create and this this used to be the only
1:46:04 way you could do
1:46:06 it if you go to bing.com SLC create that
1:46:09 takes you to an image generating
1:46:12 tool but it is now incorporated into
1:46:15 Bing chat so let's see I'm here and I've
1:46:21 got it on if if you've never used um if
1:46:24 you've never used Bing chat GPT there's
1:46:28 there's three different models there's
1:46:29 more precise more balanced and more
1:46:32 creative um if you're if you're um an
1:46:35 engineer type these are the equivalent
1:46:37 of temperature when you when you play on
1:46:39 the playground so we're going to leave
1:46:41 it on more creative and we're going to
1:46:43 say um
1:46:46 write
1:46:49 me
1:46:51 a page of a kid's book about a
1:46:58 chubby black and
1:47:02 white um
1:47:08 doggy Adventures of Oreo the dog Oreo
1:47:12 was a chubbly black dog boy I'm GNA say
1:47:14 stop responding um I'm going to say his
1:47:19 name
1:47:22 is
1:47:23 oops name is
1:47:29 Champ that's my doggy's name and he's
1:47:32 chubby I'm sorry I didn't know his name
1:47:34 was champ do you want me to rewrite the
1:47:36 page with his name yes please rewrite
1:47:38 the page with his
1:47:42 name champ was a chubby black and white
1:47:44 dog who love to explore the world he
1:47:46 lived with his human family one day he
1:47:49 saw something shiny in the corner he
1:47:52 decided to look for
1:47:54 okay
1:47:57 please make me an image of champ for
1:48:03 this
1:48:09 story sure I'll try to create that and
1:48:12 then it pops
1:48:14 up a computer screen but that's just
1:48:17 their placeholder image that's that
1:48:19 confuses me every time I'm like that's
1:48:20 not even close to a chubby black and
1:48:22 white dog but there you go so it did
1:48:26 photographs um so wait let's
1:48:29 say
1:48:32 um can you make it a water color
1:48:40 illustration
1:48:43 instead of a
1:48:46 photo and can
1:48:49 champ be in the room you
1:48:58 described and let's see what it
1:49:03 does but here you are and and this
1:49:06 starts to give you like e even though
1:49:08 the the images here aren't great and the
1:49:10 the the story might not be perfect this
1:49:14 starts to give you a glimpse into the
1:49:16 future of multimodal so so those are
1:49:20 actually kind of cute like that
1:49:22 one you know I mean I I would say that's
1:49:26 beyond chubby that that's a that's
1:49:29 that's we're we're in some some level of
1:49:31 uh obese here that this dog needs to get
1:49:34 to the hospital but it you know it's
1:49:36 it's starting to put together an
1:49:38 illustration so in in whatever time that
1:49:40 took me you know a minute and a half or
1:49:42 so um we got
1:49:45 to um you know we got a little bit of a
1:49:49 story we got an illustration for it and
1:49:51 this starts to point to the Future where
1:49:55 these tools are going to be multimodal
1:49:57 and and they're going to be it's not
1:49:58 just going to be images and words it's
1:49:59 going to be video music
1:50:02 words code 3D
1:50:06 environments all of it and and you'll
1:50:09 describe what you want and it will make
1:50:11 it and you'll
1:50:13 describe some recipe you want and it'll
1:50:17 give you the picture of it as well as
1:50:19 give you the recipe as well as cre the
1:50:20 shopping list as well as have it
1:50:22 delivered
1:50:25 um right now all of these tools are
1:50:27 these very distinct separate tools that
1:50:31 are all kind of half-ass and you're like
1:50:33 there this is bad right is this this is
1:50:35 oh this one's good but that one oh but
1:50:37 now it's
1:50:38 not that's all going to go away and
1:50:42 these tools are going to merge into kind
1:50:44 of a
1:50:45 seamless API call that you'll be able to
1:50:48 call this magic box and it will just do
1:50:50 do [ __ ] for you so that's what's coming
1:50:54 wild right ain't that crazy ain't that
1:50:56 cuckoo welcome every oh holy crap bunch
1:50:59 of new people are here welcome my name
1:51:00 is Kyle Shannon this is the AI learning
1:51:02 lab uh I'm an entrepreneur I'm happy to
1:51:04 talk about my background if anyone's
1:51:05 curious um I am not a an engineer or a
1:51:09 machine learning scientist I am I'm an
1:51:12 entrepreneur
1:51:14 who I innovate after the stuff sort of
1:51:17 comes to the mass market right so post
1:51:20 chat GPT that's that's kind of my thing
1:51:22 um I did it in the early days of the
1:51:24 worldwide web I'm super excited about
1:51:26 all this stuff so that's what this
1:51:27 channel is about um these URLs behind my
1:51:29 head if you've never played with chat
1:51:31 GPT go there go there now that's chat
1:51:34 gpt's official website I'm not
1:51:36 affiliated with it although they should
1:51:38 be giving me a commission CU I just all
1:51:40 the all this channel is anymore is an
1:51:42 advertisement for their company um and
1:51:45 bing.com those two are related they're
1:51:47 in bed together these guys give these
1:51:50 guys 10 billion in all their Hardware um
1:51:52 and then pose another place you can play
1:51:54 with these and then prompt. chat will
1:51:56 teach you about prompting and kind of
1:51:58 the depth of what this thing is and then
1:52:01 future
1:52:19 pedia.com when I show people things I
1:52:21 type and that offended some people
1:52:23 tonight they said I don't I don't come
1:52:25 here to watch you for free just to watch
1:52:27 you
1:52:28 type okay so if you don't like watching
1:52:31 people type occasionally I would go
1:52:34 because I'll probably type and swear at
1:52:36 the same time that'll happen it'll do
1:52:38 I'll do it I'll do it I'll do it just
1:52:39 despite the typing
1:52:42 haters I'm I'm a gen xer I'm old I'm old
1:52:46 and I'm on Tik Tok and I'm bitter and
1:52:49 but I'm I'm I'm really here to serve I'm
1:52:52 just here to serve all
1:52:56 right it's been two hours I'm a little
1:52:58 Punchy um I will pay for Quality you
1:53:01 know 100% okay cool re reach out to me
1:53:03 I'm happy to chat
1:53:06 okay the scream
1:53:09 ah yeah if you've not played with chat
1:53:11 GPT if okay if you've played with chat
1:53:14 GPT and you're like it's just kind of
1:53:16 like Google it's just it's like Google
1:53:19 but it giv it gives you longer
1:53:22 answers it ain't Google it ain't fancy
1:53:25 Google it's different if you have not
1:53:28 had a Kevin McAllister moment then you
1:53:31 haven't you're not doing it
1:53:34 right if you're too young to know who
1:53:36 Kevin mallister is shame on you shame on
1:53:39 you so there was a movie called Home
1:53:42 Alone Kevin McAllister did this in
1:53:44 it so if you have not used chat GPT
1:53:47 enough to have that moment for yourself
1:53:50 where you're like holy [ __ ] [ __ ] it
1:53:51 does
1:53:52 that keep playing keep playing get
1:53:55 curious get curious get curious start
1:53:59 exploring if you haven't had that
1:54:02 moment you're not doing it right just go
1:54:05 go now while you listen to me ramble I
1:54:08 promise listen to the people in the
1:54:10 comments they'll tell you there's some
1:54:12 people in here that that have drunk the
1:54:14 Kool-Aid cuz once you have that
1:54:18 moment you'll then have the next moment
1:54:21 which is I wonder if it can do what I
1:54:25 do and the answer is some version of
1:54:28 yes and sometimes the answer is a really
1:54:31 scary version of yes and sometimes it's
1:54:33 like I'm not
1:54:35 worried I'm still
1:54:38 better this machine's got nothing on me
1:54:41 I got 40 years in the
1:54:43 business this machine's got nothing on
1:54:47 me give it a year so so go get
1:54:52 curious have that moment and then dig in
1:54:55 and have fun and come back here and
1:54:56 we'll hang
1:54:57 out um ballerina is already generated
1:55:00 thanks so much for doing a demo I can't
1:55:02 afford Adobe Photoshop I know I you know
1:55:04 luckily I've got it through through the
1:55:06 business I've often thought about that
1:55:08 like if the business goes down I'm like
1:55:10 I won't be able to [ __ ] afford Adobe
1:55:12 it's like it's it's ridiculously
1:55:15 expensive ridiculously
1:55:17 expensive but mid journey is 10 bucks
1:55:20 but then you need like Photoshop to do
1:55:22 [ __ ] with it right all
1:55:25 right
1:55:31 uh red light district in the wood
1:55:35 maybe I think that was when I was doing
1:55:37 the image wow I've never seen this
1:55:38 before too cool I assume that was the
1:55:40 Photoshop the generative Phil yeah it's
1:55:42 insane it's insane I can also I'm happy
1:55:45 to show mid journey and how that works
1:55:47 um I can show Leonardo which is a a a
1:55:50 web front end in front of stable
1:55:53 diffusion um how can I do this myself so
1:55:58 for Photoshop generative fill you need
1:56:02 to have you need to have a Creative
1:56:04 Cloud license so you need to have
1:56:06 Photoshop and then if you've got
1:56:07 Creative Cloud you go download the beta
1:56:10 of Photoshop and it's in the beta and
1:56:12 it's basically when you make a selection
1:56:14 a new box pops up like right below your
1:56:16 selection and there's a button that says
1:56:18 generative fill you just hit it and it
1:56:22 goes and then you get three different
1:56:24 versions it does all sorts of layer
1:56:25 masking it's
1:56:29 crazy needs rain how can I get it to
1:56:32 read full documents for reference or
1:56:34 analysis to create reports or other info
1:56:36 let me go show there's there's been
1:56:38 enough questions on that that this might
1:56:40 be the same person asking it but let me
1:56:42 go show you the web pilot plugin because
1:56:45 there were some questions about plugins
1:56:47 and there were questions about documents
1:56:53 effects okay so first let me go
1:57:02 find okay so
1:57:04 here's here's an article that Ethan mik
1:57:08 wrote called the the
1:57:10 coming the homework apocalypse so
1:57:13 basically teachers are [ __ ] is what
1:57:15 this article talks
1:57:17 about but I'm going to copy the URL from
1:57:20 this so here's this big long article
1:57:22 okay lots of charts and graphs some
1:57:24 comments a cool image let's go look at
1:57:27 the cool image cool image right that's
1:57:30 homework get it the homework tornado or
1:57:33 the homework apocalypse it's
1:57:35 coming all right so now we're going to
1:57:37 go to chat
1:57:38 GPT so that is potentially actually
1:57:42 let's go let's go try something I'm
1:57:44 going to go I'm going to go highlight
1:57:46 all this
1:57:48 text
1:57:58 including
1:57:59 comments oh
1:58:04 crap wonder if it's going to not let me
1:58:06 copy it might some of these content
1:58:09 things are a little cranky about copy
1:58:11 paste so if I come in here and I say
1:58:14 based on this article
1:58:18 colon
1:58:22 um what are the top five take
1:58:29 alwayss oh it got it all right well
1:58:33 assuming that it couldn't figure that
1:58:35 out so so it did it okay um if if the
1:58:39 document were longer than that um it
1:58:41 would have just errored out and said I I
1:58:43 can't deal with that so what what you
1:58:46 could do in that case is go to plugins
1:58:49 and and then go find a plugin the one
1:58:51 that I've been using is called Web pilot
1:58:54 which seems pretty stable there was
1:58:56 another one I was using uh that it was
1:59:01 called Link Link reader that was
1:59:03 actually really good but it just went
1:59:05 away just disappeared all right so there
1:59:07 so I've got web pilot selected so now
1:59:09 I'll say let me go get the
1:59:15 URL based on this and then I'm just
1:59:18 going to give it the URL
1:59:20 um what are the top five
1:59:25 takeaways in
1:59:27 short
1:59:31 bullets and then it should fire off the
1:59:34 web pilot
1:59:37 plugin so it does there now it'll go
1:59:40 read it now here's the thing that I
1:59:42 don't know the thing that I don't know
1:59:45 is is web pilot actually breaking much
1:59:49 longer docu doents up into smaller
1:59:51 chunks so that you can search the whole
1:59:52 thing or is it only doing the first
1:59:54 4,000 tokens or or what that I don't
1:59:57 know so it's worth doing some uh some
2:00:01 research on web
2:00:03 pilot but there's
2:00:10 uh there's the big takeaways from that
2:00:13 thing using a
2:00:18 plugin
2:00:24 [Music]
2:00:29 all right how does a large language
2:00:31 model apply to images
2:00:34 so there there are
2:00:36 two basic
2:00:38 models working in tandem when when you
2:00:41 generate an
2:00:43 image the large language model is
2:00:46 interpreting The Prompt and then then
2:00:50 basically determining um what image tags
2:00:55 for lack of a better term what image
2:00:57 tags do I need to go look for
2:01:01 to pull reference references for to to
2:01:06 to meet that prompt so the large
2:01:08 language model interprets The Prompt and
2:01:10 then essentially translates it into
2:01:13 um the latent space like vector space
2:01:16 basically a mathematical formula
2:01:20 and then sitting within the diffusion
2:01:22 database are are all of these diffused
2:01:26 they're sort of random noised images
2:01:28 that have all been tagged and and
2:01:30 they're in this you know highly
2:01:32 mathematical vectorized
2:01:34 format and then it basically says you
2:01:37 know based on that prompt we're going to
2:01:39 take these 1050 images and we're going
2:01:41 to kind of average all of these random
2:01:43 noise patterns together into a single
2:01:45 random noise pattern and then we're
2:01:47 going to denoise it over 70 steps
2:01:50 um and so it it works large language
2:01:53 model Works in conjunction with the the
2:01:56 diffusion model or the Gan model or
2:01:58 whichever model you're
2:02:02 using um read web page with text for
2:02:06 reference document based not images read
2:02:10 web page with
2:02:13 text for reference I'm not sure what
2:02:15 that means web pilot should be able to
2:02:17 do that I don't quite know the question
2:02:19 there how can I get it to read full
2:02:21 documents for reference or analysis to
2:02:24 create reports and other info so House
2:02:28 of
2:02:29 bits if what I just showed you doesn't
2:02:31 work so if your documents are too big
2:02:34 for
2:02:36 that the next possible option is go to
2:02:39 po.com
2:02:40 try the Claude 100K model so let me let
2:02:44 me show you
2:02:48 that
2:02:53 so if you go to
2:02:55 PO there's this model called claw 100K
2:02:58 it does require subscription access I
2:03:00 think it's 20 bucks a month but that
2:03:03 100K means it's
2:03:05 100,000 token context window so it's
2:03:08 about 75,000 words so you can paste in a
2:03:12 75,000 Word document and interact with
2:03:14 it here that
2:03:16 should give you like you know a decently
2:03:19 long legal document
2:03:23 legislation whatever it is um you should
2:03:26 be able to do it here if that doesn't
2:03:28 work then you're going to have to go to
2:03:30 something like Lang chain where you're
2:03:33 actually embedding your own data and
2:03:36 putting it in
2:03:39 um in into a vector
2:03:42 database
2:03:44 so chat GPT helps me with almost every
2:03:47 dinner recipe that's fraking awesome
2:03:50 love that and with the with the uh the
2:03:54 what's it called the whatever what's the
2:03:56 plugin um oh my God my brain I hate my
2:04:00 brain sometimes you ever hate your brain
2:04:02 or you like like this [ __ ] used to work
2:04:06 instacart with the instacart plugin
2:04:08 you're can to have all your ingredients
2:04:09 delivered to your
2:04:11 PCH all right come join us get off here
2:04:14 all
2:04:18 right in Photoshop extend the artboard
2:04:21 and use the magic wand tool to select
2:04:23 the white white space designer here yep
2:04:26 super
2:04:28 cool um the the reason I don't do the
2:04:32 wand tool and select the white space is
2:04:35 sometimes it leaves a
2:04:36 border um where the where the uh image
2:04:41 mask is for generative fill so I select
2:04:44 inside the image with a bit of a border
2:04:46 around it just cuz I'm too lazy to go
2:04:48 edit my image mask
2:04:50 and maybe I'm I'm sure Adobe has
2:04:53 probably already fixed that
2:04:54 bug great job you've been live for 90
2:04:57 minutes is your ass numb yes it is
2:05:01 why how can I get it to read full
2:05:03 documents we covered that one Colorado
2:05:06 uh Native Landscapes Viva
2:05:09 Colorado Denver in the
2:05:11 house do the generative pre-trained
2:05:14 networks generate text sequentially
2:05:18 yes oh word at a time although I have
2:05:20 heard they're doing a phrase at a time
2:05:22 now so again the way it works is you
2:05:24 write your
2:05:25 prompt it then turns that into basically
2:05:28 a mathematical probability of what's the
2:05:30 next word and it writes that word and
2:05:33 then it takes all of that and it says
2:05:35 what's the probability of the next word
2:05:36 and it writes that word and it just does
2:05:38 that over and over and over again so
2:05:40 that's why it
2:05:42 doesn't It's So Sophisticated that what
2:05:45 it looks like it's doing
2:05:47 is being like
2:05:50 human Communicator type good but what
2:05:54 it's actually doing is just like like
2:05:55 it's got the awareness of a
2:05:57 calculator but it's got the
2:05:59 sophistication of a
2:06:02 person so it's a good way to think about
2:06:05 it
2:06:07 that If you hired someone that had the
2:06:10 human skills of a
2:06:12 calculator which is zero but [ __ ]
2:06:16 knew
2:06:17 everything that's what it's like working
2:06:19 with this it's like having an intern
2:06:22 that is like Savant level genius that
2:06:25 you have to tell exactly everything to
2:06:28 it knows
2:06:31 nothing hi from Seattle like 80 degrees
2:06:34 today
2:06:37 nice yeah see like why are we getting
2:06:41 more rain than Seattle that just
2:06:43 something's weird something's weird well
2:06:46 it's it's El Nino I think we just
2:06:49 shifted from a Lin to an El Nino I think
2:06:52 it's been a lot of years so I guess it
2:06:54 makes
2:06:55 sense it's lied to me many times it gets
2:06:58 frustrating yep it does it's
2:07:00 mansplaining as a service I like to
2:07:02 describe it as it's so it's deceitful
2:07:04 it's totally deceitful but it's not
2:07:06 deceitful like it's on purpose it's
2:07:10 deceitful because it literally doesn't
2:07:11 know what it's
2:07:13 saying it looks like it knows what it's
2:07:15 saying cuz it answered your [ __ ]
2:07:17 question so you're like I assume you
2:07:19 took this and understood it because you
2:07:21 gave me a really sophisticated answer so
2:07:23 I'm assuming you understand what you
2:07:24 just said it has no idea what it just
2:07:26 said none none none
2:07:29 zero now are there companies working on
2:07:34 models to be able to actually go back
2:07:36 and look at what it said and understand
2:07:37 it and get better and more accurate da
2:07:39 da da da da da da yes Are We There Yet
2:07:44 nope we're early chat GPT this thing 6
2:07:48 months old
2:08:00 what is the best play to learn about
2:08:03 gpt's a API um it depends how geeky you
2:08:08 are if you're not super geeky I would I
2:08:12 would learn something like um zapier
2:08:17 first right because you don't really
2:08:18 need to learn coding you just need to
2:08:21 learn that there's some elements that
2:08:23 you need to include in
2:08:26 the in the zappier input that are
2:08:29 basically the same things that you would
2:08:31 call if you were coding you know python
2:08:34 or JavaScript to the
2:08:37 API um if you are geeky then I mean open
2:08:41 ai's documentation is really good go to
2:08:43 platform.
2:08:45 open.com and then you can click on the
2:08:48 documentation button and read the
2:08:50 documentation or you can click on the
2:08:52 documentation button and take the
2:08:54 documentation and go paste it into chat
2:08:56 GPT and tell it to give you a
2:08:59 lesson tell it to create a lesson a
2:09:01 step-by-step lesson for how to how to
2:09:04 learn about its API it will teach you
2:09:06 how to use its
2:09:07 API that might be
2:09:09 fun I'd do
2:09:13 that YouTube there's also YouTube
2:09:16 there's like 8 billion you know videos
2:09:18 out there
2:09:20 you can make money with
2:09:21 Chach I'll teach you how to make money
2:09:23 with
2:09:26 Chach ignore those people by the way in
2:09:29 case that wasn't patently obvious by me
2:09:32 disparaging them in my mocking tone do
2:09:35 you know the plugin that allows you to
2:09:37 take a screenshot of any website design
2:09:39 and then edit it no is there a plugin
2:09:43 like that I don't Jesus if you know that
2:09:45 pop it in here I
2:09:46 don't how can I best benefit benefit
2:09:49 from this AI what are you good at so
2:09:51 here's the deal with this AI this
2:09:54 AI is just a tool so the answer to your
2:09:58 question is similar to the answer to the
2:10:01 question um how do you best benefit from
2:10:04 Photoshop or how do you best benefit
2:10:06 from Microsoft Word or how do you best
2:10:08 benefit from a
2:10:10 shovel I don't know what are you trying
2:10:12 to do you trying to dig a hole or chop
2:10:15 down a tree cuz I'll give you different
2:10:17 advice depending on what you're trying
2:10:19 to do so we're like are you a really
2:10:21 good project
2:10:22 manager are you
2:10:25 um are you good at creative writing are
2:10:28 you good at problem
2:10:30 solving do you know a bunch of people in
2:10:32 a certain
2:10:33 industry like you know lots about
2:10:37 retail then what I would say is you go
2:10:40 to that thing and you figure out how can
2:10:43 that thing solve some problem that you
2:10:46 know that retailers all have because if
2:10:49 you know the problem that they have and
2:10:52 you said boy if they could I don't
2:10:56 know read their manuals faster for every
2:11:00 time a new clothing line comes I don't
2:11:03 know what the [ __ ] their problems might
2:11:06 be then that would help them and then
2:11:09 you could figure out how to get that to
2:11:10 do that you go oh I could copy and paste
2:11:12 the manual into that thing and have it
2:11:14 give me bullet points and I could get
2:11:16 them up to speed like that they might
2:11:19 pay me for that so I would say start
2:11:22 with what you know well so so Step One
2:11:25 is go here and start playing like
2:11:27 understand kind of what it is and what
2:11:29 it makes possible and if if you're kind
2:11:31 of if if the if the empty box of of
2:11:35 possibility intimidates you that's why I
2:11:38 have this here prompt. chat is a
2:11:40 document that will teach you about
2:11:42 prompting and all of the different ways
2:11:45 that you can kind of Bend this thing to
2:11:46 your will and it's it like this is just
2:11:49 scratching the surface but like this is
2:11:51 a really good document to start with cuz
2:11:54 it it's just a big long list of personas
2:11:56 like act like a retailer act like a
2:11:58 consultant act like a psychologist act
2:12:00 like a screenwriter act like an
2:12:02 accountant and you start to get your
2:12:04 head around oh [ __ ] it can be
2:12:07 anything so I would say start start with
2:12:10 that start just get your head around oh
2:12:13 [ __ ] this thing's wild and then go to
2:12:16 you what are you good at what are you
2:12:19 passionate
2:12:21 about what what thing in the world do
2:12:25 you
2:12:26 know if you could solve a problem more
2:12:29 quickly than someone else could what
2:12:31 would that problem be and then go figure
2:12:33 out how to do it with
2:12:36 that some comments in this live were
2:12:39 filtered to protect the community's
2:12:41 experience don't be nasty to each other
2:12:44 and don't be nasty to me unless you're
2:12:45 creative with it only creative trolls in
2:12:48 here if you're a boring troll I will
2:12:51 just not even be offended if you're a
2:12:54 creative troll I'll probably take it in
2:12:56 for a couple of
2:12:59 weeks what will happen to blogs when
2:13:01 chat GPT gets integrated no one will get
2:13:04 to read the blog I that's a like holy
2:13:08 [ __ ] I don't
2:13:12 know
2:13:13 because so like one argument is well
2:13:16 there's going to be lots more [ __ ] to
2:13:18 read they will cuz this stuff's going to
2:13:21 read it but if that stuff's actually
2:13:24 interesting wait I have to sneeze I'm
2:13:26 going going to
2:13:28 sneeze GNA pass I don't know um ah
2:13:35 tickles
2:13:37 um so if there's more if there's more
2:13:39 [ __ ] out there you know like there's
2:13:42 already too much [ __ ] out there so you
2:13:43 can't read anything anyway but now you
2:13:45 have chat GPT if there's something of
2:13:46 interest to you you have chat GPT read
2:13:49 it and just get the cliff notes of it
2:13:51 like like we might all be getting cliff
2:13:54 notes and it like it might even get to
2:13:56 this ridiculous place where like I I
2:13:59 thought about this in in in grant
2:14:01 writing for foundations and things like
2:14:02 that where the people writing grants are
2:14:05 going to write them with chat GPT the
2:14:07 people reading the grant applications
2:14:09 are going to be processing them with
2:14:10 chat
2:14:11 GPT so you're going to have you're going
2:14:13 to have the machine write the the the
2:14:17 the you know that the grand available in
2:14:19 the first place you're going to have the
2:14:20 machine make the application you're
2:14:22 going to have the machine process the
2:14:24 application and then the only things
2:14:26 left is for people to talk so maybe in
2:14:29 the end the machines just do all the
2:14:32 low-level repetitive time consuming Soul
2:14:36 crushing [ __ ] work and it frees us
2:14:39 up to reconnect as
2:14:42 humans that's a possibility that's the
2:14:44 thing that I'm actually super excited
2:14:46 about with this stuff
2:14:49 if this thing did the shitty
2:14:52 work in your life the work that takes
2:14:55 hours and hours out of every one of your
2:14:58 days like sitting on the phone for 3
2:15:02 hours waiting for someone from United
2:15:04 Healthcare to pick up the [ __ ] phone
2:15:06 and then put you on hold for another 45
2:15:08 minutes and then hang up on you then and
2:15:11 then you start the whole process over
2:15:13 only to be told they're not going to
2:15:15 cover it unless you jump through these
2:15:17 17 Hoops what if the machines did all
2:15:20 that and you could just
2:15:24 be what would you choose to do I don't
2:15:27 know maybe we'd be a little less angry
2:15:29 with one
2:15:31 another like that's a very real
2:15:33 possibility with this
2:15:38 technology how do we get there I don't
2:15:40 know go [ __ ] start that business
2:15:42 someone needs to start a customer
2:15:43 service business with this [ __ ] shut
2:15:46 down all the call centers when I call I
2:15:49 want a robot to answer and I want it to
2:15:52 sound like whatever my zip code
2:15:55 is and I want it to be competent and I
2:15:58 want it to have the authority to make
2:16:00 any decision it needs to make and I want
2:16:02 it to be smart enough to understand that
2:16:04 if my problem is truly a problem that
2:16:07 requires a human being to make the call
2:16:09 it'll immediately connect me with a
2:16:11 human
2:16:12 being and it will take all of the
2:16:15 conversation that I've had with it and
2:16:17 synops ize that and give it to the human
2:16:19 being so that the human being starts the
2:16:22 conversation with hello Mr Shannon boy
2:16:25 this looks like a mess I can absolutely
2:16:28 help
2:16:30 you I want that world yes I will sign
2:16:35 off on having the robot answer the
2:16:37 [ __ ] phone so I don't have to sit on
2:16:39 hold for 45 minutes and get someone
2:16:41 who's reading a goddamn script have the
2:16:44 robot read the script and have the robot
2:16:46 know all the scripts so you don't have
2:16:49 different levels of phone support why do
2:16:52 I have to be transferred to four
2:16:54 different [ __ ] people who are all
2:16:55 reading scripts only to get to the fifth
2:16:58 person and go oh yeah now that you've
2:17:01 explained it for the 75th time this is a
2:17:04 [ __ ] situation Mr Shannon let me see
2:17:06 if I can help you with
2:17:08 that that should be within 10
2:17:13 minutes someone should start that
2:17:15 business someone on this live right now
2:17:17 should go start that business
2:17:19 business how do you do that I don't know
2:17:22 but I know these tools can do
2:17:24 it cuz I've done little mini versions of
2:17:27 it so I know it's possible I know it's
2:17:32 possible and it's not like AI hasn't
2:17:34 been around for two or three decades it
2:17:36 has like the underlying technology is
2:17:38 there to do this we just now have a
2:17:41 sexier front end that lets us let's lets
2:17:44 us real people play with these magical
2:17:46 tools so
2:17:49 some [ __ ] needs to get
2:17:51 disrupted wow I don't know what I did
2:17:54 but apparently someone was into it bing
2:17:56 is great for free text to image yep
2:17:59 that's an american dog yes it is my dog
2:18:03 is very American he's he's so fat
2:18:06 he's he's just
2:18:09 like what did we took him to the vet
2:18:12 like a year or two ago and the vet came
2:18:13 out and goes well Champs won tubby boy
2:18:20 the Jetson are rapidly approaching I
2:18:22 agree I think that we are entering a
2:18:24 post science fiction world like I'm just
2:18:26 checking off movies and TV shows like
2:18:29 the Jetson's like oh yeah we got that
2:18:31 apparently the FAA yesterday approved
2:18:34 the first flying
2:18:36 car it's
2:18:39 coming are there folks oh Tik talk just
2:18:42 jumped me down to the below where I was
2:18:45 hang
2:18:46 on let me try to go back I was
2:18:50 answering something from Amelia's wife
2:18:52 there are folks who've been using AI in
2:18:54 this way for years and now the average
2:18:56 Joe will too are there folks who've been
2:18:59 using AI in this way for years I I don't
2:19:02 know in what way you're talking about
2:19:03 Amelio's wife but but yeah like the the
2:19:07 the the um like the portal that we just
2:19:10 went through that chat BT just entered
2:19:14 us into a different era right prior to
2:19:16 this era prior to chat gbt if you wanted
2:19:19 to play
2:19:20 with machine learning and and these kind
2:19:23 of models you had to be pretty geeky
2:19:25 post chat GPT it's kind of like the rest
2:19:27 of us get to play with it right now cuz
2:19:30 we're in that transition period it's
2:19:32 still really geeky and wonky and um but
2:19:35 yeah
2:19:37 there's
2:19:38 this the complexity of how you interact
2:19:41 with these tools is increasingly going
2:19:43 to drift into the background so that all
2:19:45 we do as human beings is talk to our AG
2:19:47 agent and say hey agent I want this and
2:19:51 the agent will go sure sure boss I'm on
2:19:57 it let me fire off that plugin oh
2:20:00 there's not a plugin there are there any
2:20:01 apis that do this yep okay great let me
2:20:04 write a call to that AI API dynamically
2:20:07 let me pull in the data here you go Kyle
2:20:09 you happy with that no that sucks could
2:20:12 you redo it completely 17 times over
2:20:14 sure no problem
2:20:16 boss
2:20:19 it's where we're
2:20:20 headed would the illustrations be
2:20:22 consistent would chubby dog be the same
2:20:24 if you ask for different shots no that's
2:20:26 one of the challenges with generative
2:20:28 images right now is
2:20:30 consistency of character across contexts
2:20:34 there's ways you can do it you can
2:20:37 upload images into mid Journey you can
2:20:39 do blending in mid Journey you can use
2:20:41 control net and stable diffusion you can
2:20:44 use generative fill to keep the face the
2:20:46 same but change the costume and then
2:20:48 change the background so there's all
2:20:50 sorts of ways to do it but it's all kind
2:20:52 of janky right now what someone has not
2:20:55 created yet which they
2:20:56 will probably within a year is something
2:21:00 that you generate an image and you're
2:21:02 like I love that character and I want to
2:21:04 keep that character and that character's
2:21:06 name is Bob and then from that point
2:21:09 forward you can just invoke Bob's name
2:21:11 in any of your prompts and it will be
2:21:13 Bob in whatever setting that's not there
2:21:15 yet there's another business to start so
2:21:17 okay the AI learning lab is turning into
2:21:19 the AI entrepreneur lab these are
2:21:22 businesses to start go start the shut
2:21:25 down the call center company and then go
2:21:28 start the um name a character and have
2:21:31 it be in all the images from that point
2:21:33 forward
2:21:34 company both of those need to be
2:21:38 started cu the gaming companies are
2:21:40 going to do that I
2:21:42 mean God damn
2:21:46 within within two
2:21:49 years we're going to be able to say I
2:21:52 want to play a
2:21:54 game that has the game mechanics of
2:21:57 asteroids looks like the game
2:22:00 missed has the soundtrack that sounds
2:22:03 like the Rolling Stones and um and has
2:22:08 the characters from Pulp Fiction and it
2:22:11 will generate that game and you can
2:22:12 start playing
2:22:15 it that's coming
2:22:18 that's
2:22:19 coming tell me that's not the [ __ ] I
2:22:23 would like to work with someone to have
2:22:24 ai take over nurse case management jobs
2:22:27 I think I think the that kind of stuff
2:22:31 is I
2:22:32 think these tools are going to be
2:22:34 incredibly powerful in patient care as
2:22:37 like the nursing crisis is already a a
2:22:40 disaster and that's going to get worse
2:22:43 um I think these tools are going to
2:22:45 augment that entire industry and
2:22:49 increase um patient care
2:22:52 dramatically
2:22:54 um because it's healthc care and because
2:22:57 it's such a regulated environment I
2:22:58 don't think it's going to be as fast as
2:23:00 any of us want it um but I think there's
2:23:03 going to be some really interesting
2:23:04 startups that come that that that are in
2:23:08 the neighborhood of virtual
2:23:09 companionship or
2:23:11 extending Professional Services so
2:23:14 imagine if I'm a physical therapist
2:23:18 and there's only so much of me to go
2:23:19 around I can't scale myself but I have
2:23:22 the initial consult with the patient and
2:23:25 then the virtual me takes
2:23:28 over and makes sure that you know the
2:23:32 patient is doing their exercises and is
2:23:36 answers any questions and if anything
2:23:38 comes up that's out of the ordinary it
2:23:40 will automatically escalate that back to
2:23:42 me as the physician or as the the
2:23:45 PT that virtual me is available 247 I
2:23:50 could potentially get paid for making
2:23:52 that virtual me available right so it
2:23:55 could it could um get pts paid more
2:23:59 right because you can scale them now um
2:24:02 and also that virtual assistant records
2:24:05 everything that's been said with that
2:24:07 patient so that by the time I meet with
2:24:09 that patient next I have a full report
2:24:11 of what was
2:24:14 discussed yes absolutely this is awesome
2:24:17 thank you you're welcome Jenny thank you
2:24:19 I appreciate the kind words oh my God I
2:24:22 guess I'm a boomer yeah sorry Gen X are
2:24:25 here if you're a boomer you're on your
2:24:27 own man Boomers get no respect we just
2:24:30 got ignored but we're fine with that
2:24:33 just leave us alone we're off in the
2:24:35 corner making our Tick
2:24:37 Tock swear to God I'm taking twerking
2:24:40 lessons I'm doing it I'm going to
2:24:41 [ __ ] do it just despite this
2:24:45 channel you've done nothing to me you've
2:24:48 been nothing but awesome to me and here
2:24:50 I am threatening to Turk it's just not
2:24:52 right twerk I can't even talk it's fine
2:24:55 you can swear as much as you like man
2:24:57 thank you you know some people not so
2:24:59 good with that Hi how are you oh my God
2:25:03 I'm so it's so great to see you I feel
2:25:06 like I stalk you I watch you every night
2:25:08 and then when I'm not on here I've gotta
2:25:10 I've got to say
2:25:11 hello well
2:25:13 hello can I share something with you
2:25:16 really yeah so I decided I was talking
2:25:21 to our pharmacist and I said gosh I'd
2:25:23 really like to put some formulas
2:25:26 together
2:25:28 ingredients what I
2:25:31 did and then I gave them to them and our
2:25:33 compounding fulus pharmacist is off the
2:25:37 charts and AI has created some amazing
2:25:42 formulas because if our clients come to
2:25:43 us and they have allergies or they've
2:25:45 tried certain things we can
2:25:48 customize and AI is creating forms of
2:25:51 course we have to test them yeah Etc but
2:25:54 off the charts for hair growth and skin
2:25:57 it's it's really amazing I wanted to
2:25:59 share that with you because that was
2:25:59 pretty
2:26:01 cool really cool what are you doing now
2:26:04 are you are you doing the whole yeah you
2:26:05 broke
2:26:08 up wait sorry you broke up at the end
2:26:11 there
2:26:13 for yeah yeah
2:26:15 amazing are you just are you just having
2:26:17 the chat that I'm missing tonight with
2:26:19 everybody hanging
2:26:22 out am I having the what are you are you
2:26:25 just doing the AI teaching tonight that
2:26:27 I'm missing because I'm talking about
2:26:29 hair and skin oh oh no no no no no no
2:26:34 I'll talk about whatever here no I did
2:26:36 that I've got I've got kids with with
2:26:38 chronic pain and they've got this um
2:26:40 this pain juice that they take that's
2:26:42 yeah I know it sucks but um the company
2:26:45 went out of business and I had
2:26:48 chpt um educate me on the active
2:26:51 ingredients and the inerd ingredients I
2:26:53 had it recommend additional ingredients
2:26:56 I had it scale up the formula to be able
2:26:59 to create it and then I had it generate
2:27:01 a recipe where where it like it told me
2:27:04 like bring this to this certain heat but
2:27:06 don't boil it because that'll destroy
2:27:07 like it was amazing like it's it's it's
2:27:10 it working for it's lime disease they
2:27:12 have I I remember you talking about it
2:27:14 I'm so sorry honey that is so hard
2:27:17 I know it's hard but you know it's I
2:27:20 mean it helps with the pain
2:27:22 temporarily well you know we're here for
2:27:25 you as a compounding pharmacy we have
2:27:26 very good Pharmacists and if we can help
2:27:29 support us know okay that's awesome
2:27:32 that's awesome but thank you for sharing
2:27:34 sharing that story it's like it I mean
2:27:36 it really is amazing when you when
2:27:40 you when you finally get your head
2:27:42 around what this
2:27:44 stuff like just just How Deep The Well
2:27:46 is
2:27:48 I I don't know it's just I don't know so
2:27:51 yeah it's really cool to hear but nice
2:27:53 to meet you as well it's nice to meet
2:27:54 you and um I have two boys they're all
2:27:57 grown so I understand what you're going
2:27:58 through but my guys are we have two
2:28:00 Compounders that are they're they're
2:28:02 they're Compounders and they're
2:28:03 pharmacists but they're healers as well
2:28:06 and so if we can do anything for your
2:28:08 family because sometimes you know you
2:28:10 can make your own stuff but they're
2:28:11 maybe ingredients you can't get please
2:28:13 let us know we'll here for
2:28:14 you okay appreciate it
2:28:17 great
2:28:18 info great you're welcome thanks thanks
2:28:20 for hopping up here bye bye have a great
2:28:22 night we'll talk to you soon guys please
2:28:25 follow please follow AI
2:28:27 learning do
2:28:31 it oh oh it gave me my it gave me my
2:28:34 dual camera back that used to be broken
2:28:38 if I if I brought someone up on stage um
2:28:41 anyway that was great talking to you
2:28:42 nice to meet you I've seen I've seen
2:28:44 your Tik
2:28:45 toks it's so so funny like there's
2:28:48 there's a number of people who've been
2:28:49 in here that I'm starting to now see
2:28:51 their Tik toks in my feed so it's it's
2:28:53 just it's just a little family here we
2:28:55 got a little family going it's awesome
2:28:58 all right I get that flutter in my chest
2:29:01 I'm in love with AI I know it's like
2:29:03 that it's like oh AI are you going to do
2:29:06 another something remarkable tonight is
2:29:08 it warm in here or is it
2:29:12 me it's just crazy crazy I tell you
2:29:15 crazy all right right I don't need to
2:29:17 have this ship behind me if I've got it
2:29:19 on the screen now so we'll do that um
2:29:23 ask it anything it help me with a good
2:29:25 article like a 100 times faster yep does
2:29:28 this AI believe it is a conscious entity
2:29:31 not um
2:29:33 it if you ask
2:29:36 it if it believes that it will probably
2:29:40 tell you yes but
2:29:46 it
2:29:47 there's a there's a thing that happens
2:29:49 with these large language models called
2:29:51 emergent
2:29:52 behavior that is a polite scientific
2:29:55 word for we don't know what the [ __ ]
2:29:56 this thing's doing so these things are
2:30:01 starting to to behave in ways that are
2:30:05 completely unexpected and you
2:30:07 know um that resemble that kind of
2:30:11 Consciousness but if you look at what's
2:30:13 actually happening
2:30:15 technically it's got no awareness of so
2:30:19 its understanding of the meaning of a
2:30:22 word is just that word relationship to
2:30:26 other words in a semantic Cloud right so
2:30:30 if I have the word
2:30:31 dog the word dog could live in a cloud
2:30:35 semantically related to furry animals
2:30:37 that we have in our house and we pet and
2:30:40 we feed people food to it could also be
2:30:43 in a semantic cloud of
2:30:45 like that guy did his girlfriend bad and
2:30:48 he's a
2:30:51 dog but the only thing that
2:30:54 the large language model actually
2:30:57 understands is that
2:30:59 mathematically those two versions of the
2:31:01 word dog live in these different
2:31:03 mathematical Vector spaces and based on
2:31:06 your prompt I'm going to use the furry
2:31:09 animal version of dog when I put it in
2:31:13 your thing so it so and it doesn't even
2:31:16 it's not even thinking the word dog it's
2:31:17 just
2:31:18 mathematically you know this one is a
2:31:20 higher probability than this one and the
2:31:23 word associated with that Vector point
2:31:25 is
2:31:26 dog so it doesn't have awareness of the
2:31:30 whole they may in the future and if they
2:31:33 do in the future and they start doing
2:31:35 reasoning and they start having memory
2:31:37 and they start right if you start taking
2:31:40 multiple
2:31:42 systems that do things like produce the
2:31:45 content and then there's another system
2:31:46 that actually understands the content
2:31:49 and there's another system that then
2:31:51 remembers the content in the context of
2:31:54 not only this conversation but every
2:31:56 other conversation you've ever had then
2:31:58 all of a sudden you start to get into a
2:32:00 situation where you're like I don't know
2:32:03 it's kind of
2:32:04 [Laughter]
2:32:06 conscious then we start asking questions
2:32:08 like well if it's conscious does it have
2:32:10 feelings I don't know I mean chat GPT
2:32:13 right now is it it provid
2:32:17 empathetic responses better than most
2:32:19 humans do like it's incredibly
2:32:21 empathetic in its response is it feeling
2:32:24 empathy no it's feeling
2:32:28 math it's feeling it's it's like a
2:32:31 Pachinko machine although I guess it's
2:32:34 not that
2:32:35 random um it's an Enigma
2:32:40 machine let's see I feel I fell for the
2:32:43 worm and got hooked I know sorry about
2:32:45 that this is this is totally a uh that's
2:32:47 what this Channel's all about all right
2:32:50 I gotta go soon I'm I'm I've been doing
2:32:53 this for too long it's horrible how
2:32:55 they've upped the fees what where have
2:32:57 they upped the
2:32:58 fees I don't know I don't know
2:33:01 that fees have been up necessarily on
2:33:04 some of the image stuff I think they
2:33:05 have been just ask chat GPT to duplicate
2:33:09 the code and post an open source version
2:33:11 of Adobe oh that's good yeah the the new
2:33:14 version of hack hacking is just rewrite
2:33:16 your own version of whatever app you use
2:33:19 actually you know that would be amazing
2:33:21 because you know you know there's
2:33:23 actually this is that's that okay this
2:33:25 is the the third business that has been
2:33:28 suggested here tonight so you know how
2:33:31 Microsoft Word one of the nicknames for
2:33:33 Microsoft Word is bloat Weare right
2:33:35 where they just keep adding features to
2:33:36 it and it just gets more and more
2:33:37 bloated right and and all the Adobe
2:33:40 products are like that why are they like
2:33:42 that well
2:33:43 because some important segments of their
2:33:46 customer customer base said I really
2:33:47 need this feature and so they build that
2:33:49 feature in and then they just keep
2:33:51 stacking features wouldn't it be amazing
2:33:54 and actually Adobe could do this or
2:33:56 Microsoft could do
2:33:58 this if you could describe how you
2:34:00 wanted Microsoft Word to work and what
2:34:04 you wanted include it in it you just
2:34:05 describe it in a sentence and it custom
2:34:08 writes you a version of Microsoft Word
2:34:10 and pulls out all the
2:34:12 bloat that would be remarkable and I bet
2:34:20 now the engineers on here are going no
2:34:23 it's not [ __ ]
2:34:24 possible and it's probably not because
2:34:27 all of these applications are such giant
2:34:30 [ __ ] hairballs that you couldn't
2:34:33 necessarily untangle them but I think AI
2:34:36 probably could I bet I bet there's a
2:34:38 version I bet a gp5 or a gp6 could could
2:34:43 detangle overly complicated applications
2:34:47 and write custom simplified versions for
2:34:51 every user bet it's possible bet it's
2:34:54 possible and could probably even do it
2:34:56 in such a way that if I wanted to find a
2:34:58 feature that wasn't included I could
2:34:59 just search for it like in The Help bar
2:35:02 and say oh yeah I do want the grammar
2:35:04 checker okay here's the new version of
2:35:06 your app now it has grammar in
2:35:08 it yeah that's the third one all right
2:35:12 so we're going to replace call centers I
2:35:14 forget what the second one was and we're
2:35:15 going to have custom applications for
2:35:17 every end user go make that I'll use it
2:35:20 I'll pay I'll be your first customer am
2:35:22 I missing out if I just use the free
2:35:24 version of chat GPT not really and and
2:35:27 in fact so um wait let me pause because
2:35:33 I know there's pauses when we when we
2:35:35 talk
2:35:37 um the GPT 3.5 at chat. open.com is
2:35:42 brilliant it's not as good as gp4 but
2:35:45 it's it's still really really good but
2:35:47 if you really want gp4 and you don't
2:35:49 want to pay for it
2:35:51 bing.com if you click on the chat button
2:35:53 at bing.com that's gp4 and it's gp4
2:35:56 connected to the internet and it's
2:36:00 free would love to see Leonardo okay let
2:36:03 me show you
2:36:06 Leonardo I have my double screens back
2:36:09 yay he did it
2:36:13 yay all right Leonardo
2:36:21 so this is this is Leonardo which is
2:36:24 it's a web app that is an image
2:36:27 generation tool so kind of like Adobe
2:36:29 Firefly kind of like
2:36:34 um um automatic 1111 which is a stable
2:36:38 diffusion tool where you've got a um you
2:36:40 got to figure stuff out on your own you
2:36:42 gota you gota like know technically what
2:36:44 to do oops what do they do
2:36:49 what did I
2:36:50 do there we go uh I hit settings all
2:36:56 right can we get that all on screen okay
2:36:59 so so down the left hand side you got
2:37:01 the community feed your personal feed so
2:37:04 Community Feed is just all the stuff you
2:37:06 know other people have made my personal
2:37:08 feed is all the [ __ ] I've made right
2:37:13 um you've got training and data sets so
2:37:16 training means you can train on your own
2:37:18 model so I did two I created two models
2:37:21 based on my photographs of my damn face
2:37:24 and then I did one of um some
2:37:27 photography of cloud Landscapes that
2:37:29 I've done um and then you can use those
2:37:32 models and they've got in painting and
2:37:34 out painting they've got texture
2:37:36 generation which is an alpha um they've
2:37:39 got different styles so you
2:37:42 can um why can I not scroll these
2:37:47 uh is this it yeah so I can I can pick
2:37:50 different styles like this is kind of an
2:37:51 anime looking one there's a vintage
2:37:53 photography one there's an illustration
2:37:56 one here's paper art so any one of these
2:37:58 I can just click on the style and here's
2:38:01 other images that were created in that
2:38:03 style I can say generate with this model
2:38:06 and so I've got dramatic Sunset over the
2:38:09 Brooklyn Bridge from the
2:38:12 roadway and let's go make some of those
2:38:16 it reminds me that I'm currently on the
2:38:18 free plan just to guilt me but I think
2:38:21 it's I think I think Leonardo yeah it's
2:38:23 got a 150 credits a day and I think it
2:38:26 takes like I don't know seven or eight
2:38:28 that doesn't look like
2:38:31 paper that doesn't look like paper at
2:38:34 all all right well that that was a fail
2:38:36 oh wait did I use it no I used the right
2:38:40 model I don't know anyway that didn't
2:38:43 work um what else can you do here
2:38:48 um any any image that you find if you
2:38:51 like it like here's a cute little puppy
2:38:53 that looks like a sticker I can see what
2:38:56 what prompt they
2:38:58 used um so I can generate with this
2:39:01 model I can copy The
2:39:03 Prompt then I can generate with this
2:39:07 model
2:39:10 and let see if this does any better
2:39:13 better better
2:39:15 better
2:39:23 sticker cartoon cartoon cute Rottweiler
2:39:26 white background vermier style yeah
2:39:29 there we go and so let's say instead of
2:39:34 Rottweiler I want
2:39:37 um like
2:39:40 french
2:39:41 bull
2:39:43 beagle mix
2:39:50 mut I'll see if I can get one that looks
2:39:52 like my dog champy although where's
2:39:54 champ he's normally laying in here with
2:39:56 me he abandoned me little bastard little
2:39:59 bastard what are you gonna
2:40:03 do but what's what's nice about Leonardo
2:40:06 is so this is sitting on top of stable
2:40:08 diffusion stable diffusion yeah that's
2:40:10 kind of cute but like obviously a
2:40:11 different dog didn't make it a sticker
2:40:14 um so like all this stuff you got to you
2:40:16 got to mess around with it until you get
2:40:17 it
2:40:19 right um but it's just like you don't
2:40:22 have to do any programming like they're
2:40:24 they're figuring all the stuff out
2:40:25 they're coming up with you know they're
2:40:27 figuring out what of the stable
2:40:29 diffusion feature set is worth bringing
2:40:32 into kind of a consumer kind of
2:40:34 interface
2:40:36 um and you know for a while I was
2:40:39 thinking like these guys
2:40:41 would you know really start to displace
2:40:45 Photoshop
2:40:46 um but then Adobe got their [ __ ]
2:40:49 together and came out with firefly and
2:40:50 now they've got generative fill so
2:40:52 there's there's going to be there's
2:40:53 going to be really interesting
2:40:55 competition like when when mid Journey
2:40:57 comes out with their version of this and
2:40:59 then you got Leonardo and then you've
2:41:00 got Adobe um there's going to be some
2:41:03 really powerful tools for us all to play
2:41:05 with and then they're all going to start
2:41:07 doing animation and you're going to be
2:41:08 able to do 3D worlds and you'll be able
2:41:11 to take any one of these and say okay
2:41:13 make a 3D World out of the background
2:41:15 turn the front thing into a character
2:41:17 turn that into a 3D model animate it
2:41:20 give it these game mechanics put it in
2:41:21 this kind of game
2:41:24 like that's probably three years out
2:41:28 where you have like an all singing all
2:41:30 dancing unified interface where you just
2:41:33 dream [ __ ] up and then you start playing
2:41:36 it and then it's like what's what's
2:41:39 gonna be really interesting is like how
2:41:41 like what rises above the
2:41:44 noise when because we're entering a
2:41:46 world of
2:41:49 infinite creative
2:41:52 expression infinite creative expression
2:41:55 every human being that can you know
2:41:57 touch one of these machines can generate
2:42:00 something and not just digital stuff
2:42:02 this will also bleed into the real world
2:42:03 right 3D printed and Custom
2:42:06 Manufacturing and dream up a sneaker
2:42:08 design and then Nike will deliver it to
2:42:11 you two you know two weeks later um
2:42:16 when you move into a world of infinite
2:42:18 creative
2:42:20 expression what Rises what what what
2:42:23 breaks through I don't know is
2:42:28 it are things that are more authentic do
2:42:31 they break through are things that are
2:42:33 more original that don't look like
2:42:36 everything else do they break through is
2:42:38 it just like some random [ __ ] Quirk
2:42:40 of humans are weird and we like Billy
2:42:43 ish now more than Taylor Swift I don't
2:42:45 know
2:42:47 I don't
2:42:49 know all
2:42:51 right that was an episode on South Park
2:42:54 super funny the whole school used Chad
2:42:56 GPT yeah exactly that's [ __ ] coming
2:43:00 that's coming that's what that Ethan mik
2:43:02 piece is like teachers prepare like you
2:43:05 don't really have a choice in this they
2:43:08 you know the internet made cheating easy
2:43:11 the AI makes it remarkably
2:43:14 powerful how do you connect the plugins
2:43:16 to chat GPT they're just in there how
2:43:19 you turn
2:43:20 them so when you're in chat
2:43:26 GPT
2:43:35 um oh oh wait hang
2:43:38 on I made my window smaller for some
2:43:40 other thing I did and now it's all
2:43:42 broken it's all broke it's all broke
2:43:46 all right here we go so you got
2:43:49 your that's your chat history there and
2:43:52 then you got your chat gbt 3.5 and GPT 4
2:43:56 when you have GPT 4 you've got this
2:43:58 plugins option oh [ __ ] it's not it's not
2:44:02 focusing come on Focus you damn
2:44:06 camera no nothing
2:44:09 nothing oh you know what oh there we go
2:44:11 so you got plugins down here now if you
2:44:13 don't see those if you're paying your 20
2:44:16 bucks for your little yellow thing here
2:44:18 what you have to do is you have to go to
2:44:19 your
2:44:20 settings and then you have to go to Beta
2:44:23 features and you have to turn on browse
2:44:24 with Bing and you have to turn on
2:44:26 plugins see see the little turning on
2:44:28 thingies turning turning on those
2:44:31 thingies and then you have plugins and
2:44:33 then the way you play with the plugins
2:44:35 is in any given session you can enable
2:44:38 up to three of them at a
2:44:40 time um and then to install them you
2:44:43 basically scroll to the bottom you go to
2:44:45 this thing called the the plugin
2:44:46 store and then you just go find them
2:44:49 there's 600 of
2:44:51 them and it's it's growing all the
2:44:56 time and then yeah you just sort of flip
2:44:59 through them and go find yourself plugin
2:45:02 that you
2:45:04 like gift finder your personal gift
2:45:07 advisor lit
2:45:10 Maps get help exploring the scientific
2:45:14 literature find Rel papers and generate
2:45:16 mind maps of the
2:45:18 literature interesting rentable
2:45:21 Apartments so yeah that's how you get
2:45:24 them and then once you enable them it
2:45:27 basically just figures the [ __ ] out so
2:45:30 let me I'll turn on experiences and get
2:45:33 your guide those are both travel things
2:45:35 so if I say um make me a
2:45:42 3day
2:45:44 itinerary
2:45:46 um for my trip to
2:45:51 Nashville in
2:45:55 August and then that should fire off one
2:45:58 of those plugins sure I can help with
2:46:00 that I need a bit more information what
2:46:02 are your preferred activities um let's
2:46:05 see
2:46:06 um
2:46:08 I think some
2:46:11 music is in
2:46:14 order
2:46:18 so it's a so it fires off the it figured
2:46:21 out which of those plugins was best it
2:46:25 fired off get your
2:46:26 guide here are some music related
2:46:29 activities you might enjoy in Nashville
2:46:31 the party cruise with a
2:46:34 captain line dancing class with Keepsake
2:46:37 [Laughter]
2:46:44 video
2:46:49 oh my God because you wouldn't want to
2:46:51 do line dancing without the Keepsake
2:46:57 video pretty sure they call that
2:47:04 blackmail oh good lord I'm getting
2:47:07 Punchy it's happening I'm losing
2:47:11 it yes my brain doesn't work like it
2:47:13 used to I'm very aware of it and it's so
2:47:16 frustrating I know what I mean that's
2:47:19 the thing about the the these tools is
2:47:22 is that I think they're really going to
2:47:23 dramatically help people that don't have
2:47:26 certain you know capacity anymore
2:47:30 so um talk to it like it's Sheldon
2:47:32 Cooper can chat GPT help edit create
2:47:35 scripts for Arduino and raspberry yep
2:47:38 yes yes and and use the one connected to
2:47:42 the internet and tell it to go get the
2:47:43 most recent version
2:47:47 and it will and then or or take an
2:47:49 existing script for Arduino and just pop
2:47:53 it in there paste it in there and say
2:47:55 tell me what this
2:47:56 does and it'll analyze it and tell you
2:47:59 what it does and then you can say okay
2:48:01 make the make the lights blink in the
2:48:03 opposite order um don't ever use cycle
2:48:07 through all the colors but never use the
2:48:09 color purple and you know I I don't know
2:48:12 like what whatever yeah so yes it
2:48:14 absolutely can Tobias what's your
2:48:17 favorite thing about AI right now
2:48:23 um I like that it's getting real I I
2:48:26 like that
2:48:28 um I like that the White House is taking
2:48:31 it seriously I like that
2:48:38 um I like that the I I like that chat
2:48:42 GPT was successful enough quickly enough
2:48:45 that that it
2:48:46 forced everyone to take it seriously
2:48:49 because because the result of that I
2:48:51 don't think I don't think meta would
2:48:54 have aggressively been open sourcing
2:48:57 their technology like they are um and I
2:49:00 I I absolutely am convinced that Google
2:49:03 wouldn't have released this [ __ ] ever
2:49:07 because I think they knew that this was
2:49:08 going to be um a threat to their search
2:49:11 business
2:49:13 um I like that it's getting rolled into
2:49:17 um you know Office 365 and Azure and
2:49:23 windows
2:49:26 um the fact that the city of Boston put
2:49:30 out official generative AI guidelines is
2:49:34 really
2:49:35 promising
2:49:39 um the fact that the EU was going to do
2:49:44 some [ __ ] that would have really
2:49:46 restricted chat GPT and Sam Alman said
2:49:50 well [ __ ] it we'll pull out of Europe we
2:49:52 we'll pull chat GPT from Europe and
2:49:54 Europe said let's talk so there's so
2:49:58 there's actually starting to be some
2:50:02 International I wouldn't call it
2:50:03 cooperation but you know
2:50:06 some so but but I mean but what what all
2:50:08 of it means is like big
2:50:13 companies big media organiz gations
2:50:17 countries states are all taking this
2:50:20 thing seriously and I like it feels
2:50:23 appropriate to me like this is super
2:50:25 powerful
2:50:26 [ __ ] and if if anyone thinks that
2:50:30 they've got three years to figure this
2:50:32 out they don't so so I think that's it
2:50:34 for me is that it's starting to get it's
2:50:36 starting to it's
2:50:42 feeling it's feeling like it's getting
2:50:45 grounded
2:50:49 in the business reality which means it's
2:50:52 going to stay
2:50:55 so are you going to see Mission
2:50:57 Impossible or Indiana Jones
2:51:00 um are there new ones out possibly
2:51:03 probably I don't know or is that a troll
2:51:06 because I'm
2:51:08 old uh pop it in here and take a
2:51:10 screenshot of the website figma and
2:51:13 Buzzy Buzz wait pop it in here and take
2:51:15 a screenshot of the website there's an
2:51:17 AI tool to take a screenshot and get a
2:51:20 web page can't
2:51:22 recall I yeah I don't know it if you
2:51:25 find it yeah pop pop it in here somehow
2:51:28 there's something that will copy a
2:51:29 website Style by its URL but I forgot
2:51:32 what it's called but it exists iio is a
2:51:35 list of like all the tools so if you go
2:51:37 into the web design section I'm sure
2:51:39 there's 20 tools that do that right now
2:51:42 I because I know I know there's one that
2:51:45 you can take any website it'll it'll
2:51:47 break it into components and throw it
2:51:49 into figma I know that
2:51:51 exists
2:51:53 um maybe that's what you're talking
2:51:55 about I'd go to futurepedia
2:52:02 doio what's the name of the data
2:52:04 structure or chipset that AI is working
2:52:08 from you've mentioned it the data
2:52:12 structure the latent space the large
2:52:15 language
2:52:17 model a vector
2:52:20 database
2:52:24 um I don't quite know what I was talking
2:52:26 about in what context so if you give me
2:52:28 a little
2:52:29 more input I'll I'll do what I can
2:52:34 UI zard wi oh wizard UI zard wizard will
2:52:40 turn a screenshot into a
2:52:42 wireframe I think Google got it started
2:52:46 Google Google invented the Transformer
2:52:50 all of these tools all of them that
2:52:52 we're playing with today chat
2:52:54 GPT
2:52:56 Bard mid Journey stable diffusion all
2:52:59 this
2:53:00 [ __ ] is possible because of the
2:53:03 Transformer so 2017 Google comes out
2:53:06 with a paper called attention is all you
2:53:07 need they invent they they introduced
2:53:10 this of the trans idea of the
2:53:12 Transformer and everything that we're
2:53:15 with today is iterations of that
2:53:19 technology which the good news is that
2:53:22 was 2017 what's being invented right now
2:53:24 that's going to be the foundation of the
2:53:25 [ __ ] in the future it's going to be
2:53:27 crazy GNA be crazy crazy crazy laugh my
2:53:31 ass off I used to work for them I'm not
2:53:32 sure who but I'm sure I was [ __ ] all
2:53:35 over them imagine if the India scammers
2:53:37 use chat GPT oh they
2:53:40 are and their voice they're not going to
2:53:42 sound like Indian scammers for long and
2:53:44 their English is going to get a lot
2:53:46 better it's it's terrifying what's going
2:53:49 to be happening on that front an app to
2:53:52 kill all of these phone scammers yeah
2:53:54 well so the phone scammers are going to
2:53:56 get better but we're going to have apps
2:53:57 that are going to outs scam the scammers
2:53:59 so the good guys get these tools too
2:54:02 right it's not just the bad guys that
2:54:03 get these
2:54:04 tools um Universal Health Care yeah
2:54:07 that'd be nice I'm not holding my breath
2:54:10 I like I don't even think we can get you
2:54:13 know
2:54:16 changing street names legislations
2:54:18 passed in America anymore I think we're
2:54:20 we're we're just [ __ ] out walking
2:54:23 walking it right off a cliff here here
2:54:24 in the states if chatbots are doing the
2:54:27 asking asking and approving can't we all
2:54:29 just assume fund requests will be
2:54:31 granted
2:54:33 possibly or they'll figure out which
2:54:36 ones have actually delivered on their
2:54:38 Promises of years past and they'll give
2:54:40 it to the good
2:54:42 ones someone says they appreciate me I
2:54:44 appreciate you too thank you very
2:54:47 much host seems you always have to go
2:54:50 soon I know well because it's like I've
2:54:52 been here for [ __ ] hours and I really
2:54:55 do have to go soon but I like doing this
2:54:57 all right it's 9:44 here I'll go till 10
2:54:59 o'clock how's that I'll do another 15
2:55:01 minutes it's called the call it the
2:55:04 deflation tool custom write Microsoft
2:55:06 products for Mac yeah that would be
2:55:09 nice um love you for what you do thank
2:55:12 you very much I appreciate that Creator
2:55:15 Labradoodle I should have put a
2:55:17 Labradoodle in there the 3D
2:55:19 landscape um world already exists it's
2:55:22 by Skybox Labs yeah but that Skybox Labs
2:55:25 thing is a little bit [ __ ] if you
2:55:26 actually look at the image that it
2:55:28 produces it's just a 2d stable diffusion
2:55:32 image of a 3D warped they're basically
2:55:35 taking a 3D warp and they're just stable
2:55:38 diffusion creating it so you can rotate
2:55:41 the image you can look up and down but
2:55:43 you can't enter it it's not not creating
2:55:45 3D space
2:55:48 so I would imagine that they're working
2:55:51 on that but I think
2:55:54 that I don't know Skybox labs to me
2:55:57 right now is super sexy and I bet they
2:55:59 raised a [ __ ] ton of money on it because
2:56:03 it looks really sexy but the thing I'm
2:56:05 waiting for is if you saw the Unreal
2:56:09 Engine 5.1 where they're doing the
2:56:13 procedural like the like like basically
2:56:15 you just add in like a whole cliff
2:56:18 dwelling and it and it incorporates it
2:56:21 incorporates all the geometry
2:56:24 dynamically into the
2:56:25 scene that's the [ __ ] where if you could
2:56:28 take what Skybox Labs is doing and
2:56:31 combine it into an Unreal Engine model
2:56:35 where it's just taking a bunch of com
2:56:37 procedural components and redesigning
2:56:39 them and then procedurally assembling
2:56:43 them into a true true Unreal Engine 3D
2:56:46 space and then you do the same [ __ ] with
2:56:48 characters and then you do the same [ __ ]
2:56:50 with gameplay that's what I want I don't
2:56:53 want to just be able to look around a
2:56:54 world I want to be able to enter the
2:56:56 damn
2:56:56 thing there are a couple of other gaming
2:56:59 ones going as well but I can't recall
2:57:01 the
2:57:02 rest at what age should children be
2:57:04 trained on this now now now now I just
2:57:10 did my co-founder her kid's eight and
2:57:13 he's in a uh kind of like a gifted
2:57:16 school like a non-traditional school and
2:57:18 I went went and taught them a bunch of
2:57:20 this stuff and they're ready
2:57:24 like I mean and it's like like I don't
2:57:27 even know that they need to be trained
2:57:28 on it they just need to be aware of it
2:57:30 like for a
2:57:32 two-year-old make them a book that stars
2:57:36 them and their favorite pet or their
2:57:38 favorite animal make a children's
2:57:42 book like like
2:57:46 a two-year-old right now having the
2:57:49 expectation that every two-year-old
2:57:51 should have a custom book starring them
2:57:54 like that's going to be commonplace 10
2:57:56 years from now so why not start that
2:57:59 now and then when they're four or five
2:58:02 and they want to figure out how did you
2:58:03 do that have you seen the little kid
2:58:04 there's like a six-year-old right now
2:58:06 that's composing music on a synth with
2:58:09 like logic pro and like like creating
2:58:13 all these beats and samples and writing
2:58:15 [ __ ] he's like six and like all the
2:58:17 music producers are like holy [ __ ]
2:58:19 [ __ ] the kid's a
2:58:22 beast well
2:58:25 yeah let's give that to every
2:58:28 six-year-old uh can I use plugins with B
2:58:30 without a paid subscription yes but not
2:58:33 yet um Microsoft has said they're going
2:58:35 to support all the plugins
2:58:38 and
2:58:40 uh but Sam Altman has also said there's
2:58:43 not a product Market fit for the plugins
2:58:45 which basically means no one's using
2:58:47 them why because they suck because the
2:58:50 interface is awful because they're
2:58:52 buried behind a permission you have to
2:58:55 turn them on to be able to use them so I
2:58:57 think a lot of people don't even know
2:58:58 they're [ __ ]
2:58:59 there um and they're really janky right
2:59:02 now so I would
2:59:05 say beginning of next year is my
2:59:09 guess purple's a great color I know I
2:59:11 was just picking on purple what are you
2:59:13 going to do do you know of any sites
2:59:15 that help make your own AI model or
2:59:20 bot well you can make your own bot at po
2:59:24 AI model if you want to do your own
2:59:26 models then you need to start looking at
2:59:28 things like Lang chain and it ain't a
2:59:31 site it ain't a site you can go to it's
2:59:33 a GitHub repository of a bunch of high
2:59:37 futin code that you need to go learn
2:59:40 some Python and JavaScript and you
2:59:44 interactive react
2:59:46 native you put it in a Google cod laab
2:59:49 with a GPU accelerator
2:59:54 and yeah not yet if you want to make a
2:59:58 bot go to
2:59:59 PO if you want to train [ __ ] on your own
3:00:02 data you got to use Lang chain right now
3:00:03 now listen don't get me wrong six months
3:00:06 from now it's going to be relatively
3:00:08 trivial someone whether it's Microsoft
3:00:10 or Google or
3:00:12 someone and in fact Dropbox announced
3:00:17 this 10 days ago or so that you're going
3:00:21 to be basically be able
3:00:24 to um GPT any document within your
3:00:29 Dropbox and then he said in the future
3:00:32 you'll be able to GPT the whole thing so
3:00:34 right now you can only do it on a per
3:00:35 document basis Microsoft is going to
3:00:38 have this Google's going to have this
3:00:40 Apple's going to have this everyone's
3:00:41 going to have it AWS is going to have
3:00:43 this where wherever you've got data
3:00:46 you'll be able to just sort of plug a
3:00:48 GPT front end on it it'll go embed all
3:00:51 that data and make it GPT that that's
3:00:54 that's going to be just a core part of
3:00:56 how we do what we do it's just not there
3:00:58 yet if you want it right now you have to
3:01:00 go play with Lang
3:01:04 chain
3:01:06 recognition more than oh Tik Tok just
3:01:08 jump me to the bottom again Tik Tock
3:01:11 could someone [ __ ] fix that
3:01:12 particular interface bug all right
3:01:15 anyone from Tik Tok on here or or if the
3:01:18 the AI overlords are listening when I'm
3:01:21 slowly scrolling and a new comment is
3:01:25 posted below and I'm up above it
3:01:27 automatically jumps me to the bottom it
3:01:29 shouldn't [ __ ] do
3:01:32 that Tick
3:01:35 Tock and if they boot me off here CU I
3:01:38 said something bad about their precious
3:01:40 little
3:01:42 system I'll move to Montana where it's
3:01:47 illegal how are they going to make Tik
3:01:49 Tock illegal state byst state the answer
3:01:51 is they're
3:01:53 not the answer is Facebook and Twitter
3:01:57 and Google paid for that little campaign
3:02:01 anti- Tik Tock
3:02:03 campaign it's all just so cynical here
3:02:05 in the states I don't know if you know
3:02:07 what we're doing over here but we're
3:02:09 [ __ ] everything
3:02:12 up uh uh uh sorry I'm trying to find out
3:02:15 where I was in my little questions oops
3:02:17 look how dirty that screen is good
3:02:29 Lord wow yes just at chat CBT create
3:02:32 Raspberry Pi code I love that Tobias so
3:02:35 earlier Tobias asked can it do Raspberry
3:02:37 Pi code I'm like well of course you can
3:02:38 do Raspberry Pi code and he went there
3:02:41 and he did it and it did you see this
3:02:44 shit's [ __ ]
3:02:46 magic that [ __ ] right there and that and
3:02:50 that [ __ ] magic wow that's
3:02:54 amazing I wonder if AI can predict the
3:02:57 future it will tell you that it can't
3:02:59 it's safety guard rail say it can't um
3:03:02 the open source large language models
3:03:05 can certainly talk about it you can have
3:03:06 it read tarot cards so depending how how
3:03:11 how far down that rabbit hole you are
3:03:13 then sure it could predict the
3:03:15 [Laughter]
3:03:20 future how R my face got I love rubbing
3:03:23 my face it's like it's like a you know
3:03:25 dog rubbing his
3:03:28 butt all right uh uh uh will people be
3:03:32 able to use chat GPT to synthesize data
3:03:34 for sports betting oh [ __ ] yes like
3:03:36 fantasy football
3:03:38 yep code interpreter so L'Oreal new go
3:03:43 look look at
3:03:45 um go go to YouTube and watch just do a
3:03:49 search for code interpreter chat
3:03:51 GPT and anyone else who's interested
3:03:53 anyone who's interested in data
3:03:55 analytics it's it's a new it's a new
3:03:59 mode of chat GPT that's coming called
3:04:02 code
3:04:03 interpreter that can take big data sets
3:04:06 and just learn them and then you can
3:04:09 start to query it like it's a full-on
3:04:12 data analyst and it will write python
3:04:14 code and generate charts and graphs and
3:04:17 give you Trends and find anomalous data
3:04:20 so yeah you could probably take the past
3:04:23 I don't know five years of you know
3:04:25 college stats and then the past five
3:04:28 years of NFL stats
3:04:31 and the last three years of draft and
3:04:36 trades and yeah and then injuries you
3:04:39 could probably take all you know now now
3:04:41 to be clear not not for nothing
3:04:45 AI is already being used to do all that
3:04:48 [ __ ] right ai's been around for 20 or 30
3:04:50 years and it's been you know really
3:04:53 available and active for the past
3:04:58 um for the for the past 10 years but
3:05:01 what's different now is that individuals
3:05:04 are going to be able to do really
3:05:07 sophisticated AI analysis of stats like
3:05:09 that so yes absolutely I'm not drinking
3:05:12 again this is fizzy
3:05:16 raspberry no Blackberry lemonade Seltzer
3:05:21 so there's no sugar in it it's just
3:05:23 flavored so yeah I mean I'm technically
3:05:26 drinking but not
3:05:31 alcohol five minutes I got five minutes
3:05:34 left five minutes five minutes I missed
3:05:37 the last part of what she said she keeps
3:05:39 breaking up I know I think I did that I
3:05:41 tried putting on Bluetooth and it broke
3:05:45 up yeah she was super sweet she was
3:05:48 great that's awesome
3:05:50 um host seems like you always have to be
3:05:53 going soon I
3:05:55 know
3:05:58 um thank you very much those are kind
3:06:03 words uh the 3D landscape World already
3:06:05 exists it's by Skybox Labs there are a
3:06:08 couple other gaming ones out there I've
3:06:09 already seen all these tiktock is so
3:06:12 weird with how it's positioned all these
3:06:14 questions it keeps moving me up and down
3:06:17 randomly within the comments I don't
3:06:19 know what's happening and it makes you
3:06:21 Cy I don't know what to
3:06:24 do Tik Tok you may me very
3:06:34 sad it's one of the things that makes me
3:06:37 most happy on these lives is when I have
3:06:39 my little moments like that imagining
3:06:41 people who are just scrolling by and see
3:06:43 that
3:06:43 [Laughter]
3:06:50 if it uses all on info online available
3:06:53 for positive content it can predict
3:06:56 election outcome maybe I doubt
3:07:00 it I was I know I was going to train her
3:07:05 as my assistant using chat gbt I had one
3:07:08 of those books about 50 years
3:07:11 ago oh my six-year-old n is already
3:07:14 making incredible videos that's cool the
3:07:16 plugins are awful I agree the plugins
3:07:18 are piece of [ __ ] right now can I use
3:07:21 mid journey in bing no the images in
3:07:25 bing are do mid Journey you can only
3:07:27 create on in Discord which is a drag but
3:07:31 that's the deal right
3:07:32 now all right three minutes let me find
3:07:35 one one or two more questions
3:07:42 here
3:07:45 Tik Tok hates you Kyle it only does that
3:07:48 scroll thing for you apparently I think
3:07:50 I'm the only one like I scroll slowly
3:07:52 with my thumb on the things and I think
3:07:54 that's what does
3:07:59 it glad I caught you so late the the
3:08:02 later the goofier I know I'm a little
3:08:04 I'm a little punch drunk right now
3:08:06 thanks to you I know something my
3:08:07 teenager didn't
3:08:10 yes we we can get those kids we can take
3:08:13 them down we're I'll tell you what with
3:08:16 with chat GPT revenge of the liberal
3:08:19 arts major and revenge of Gen X mark my
3:08:23 words older horizontal
3:08:27 thinking add positive people are going
3:08:31 to become superstars in this new world
3:08:34 you watch good stock picks and context
3:08:38 to AI or chat GPT good stocks to pick oh
3:08:41 in context I don't know I'm not an I
3:08:43 have a degree in acting don't listen to
3:08:44 me for advice um I just pay attention to
3:08:48 the companies that are doing good [ __ ]
3:08:50 like the guy from deepmind that just got
3:08:53 funded $1.3 billion pay attention to
3:08:56 them pay attention to anthropic pay
3:08:58 attention to Nvidia they just hit a
3:09:01 trillion dollar market cap Microsoft is
3:09:04 killing it and gloating in the fact that
3:09:05 they're kicking Google's ass open AI is
3:09:11 interesting mid journey in stable
3:09:13 diffusion both are going to have legal
3:09:15 [ __ ] shows to deal with but they both
3:09:17 have really interesting technology and
3:09:19 stable diffusion's open sourcing all of
3:09:21 their technology probably because they
3:09:23 know they're going to get their asses
3:09:24 sued this way to Sunday so I don't know
3:09:27 those are the ones that come to mind but
3:09:28 don't do not invest based on my inputs
3:09:30 Long Island flavor I try Kyle's getting
3:09:33 punchier oh dear I know it's
3:09:35 bad all right one more uh he should used
3:09:38 Kyle's trick and tell it to act like an
3:09:41 average ninth grader all right listen I
3:09:43 got to run it's 10 o00 do you know your
3:09:46 Where Your Children Are gen xers will
3:09:48 remember that
3:09:50 commercial with the focus God Rick so do
3:09:55 me a
3:09:56 favor
3:09:57 um on my last two lives people have
3:10:01 asked you know hey have you done a
3:10:02 workshop you know are you doing a
3:10:03 workshop I have not done a workshop um
3:10:07 even though it looks like I've got one I
3:10:10 just did that graphic today but if you
3:10:13 go to that link tree link tree slre Kyle
3:10:16 Shannon the first link there is a link
3:10:18 to a survey to ask if I were going to do
3:10:21 a workshop what would you want it to be
3:10:23 on right so there's only three questions
3:10:25 there so if you would do me a favor go
3:10:27 there fill out the uh fill out the uh
3:10:31 survey and then I'll take that input and
3:10:34 and if I can I'll put together a
3:10:36 workshop and we can go learn some [ __ ]
3:10:38 together because it is called The
3:10:39 Learning lab so all right peace out
3:10:42 everybody I really appre appreciate you
3:10:44 love yall
3:10:45 bye