aiLL 6/26/23 - The Fired-Up Future is Now

Description

Chapters

Transcript

0:26 hello hello hello it's Kyle Shannon with
0:28 the AI learning lab welcome welcome welc
0:31 welcome just getting started
0:34 here ah yes
0:38 happy Monday happy Monday TG
0:44 am we excited if we're if we're
0:47 reconstituting the entire Workforce with
0:49 AI maybe we
0:50 reconstitute Monday as a day we look
0:53 forward to
0:54 huh no anyone berer berer buer berer
1:01 welcome everybody Kyle Shannon with the
1:03 AI learning
1:05 lab um as people get piling on in here
1:09 I'll tell you a little bit about my
1:11 background I'm an entrepreneur of many
1:14 years uh I'm officially a member of
1:17 Elder
1:19 talk old people using new
1:24 technology um I still have my teeth I
1:26 still have some amount of
1:28 hair uh
1:33 uh let's see I've started I don't know
1:34 12 or 15 companies in my day my current
1:37 Venture is called storyvine it's a uh
1:40 automated video storytelling platform
1:42 I'm the founder and CEO and the inventor
1:44 of the tech
1:46 um I was also a co-founder of agency.com
1:49 back in the olden timey days in the mid
1:52 1990s back in the
1:54 90s when there were like six HTML tags
1:57 and um the parallels between what was
2:00 happening in the early days of the web
2:02 and what's happening with the early days
2:03 of generative AI are astoundingly
2:06 similar and yet I think the uh the the
2:11 profound impact uh that the web had is
2:15 going to pale in comparison to the
2:16 impact that generative AI is going to
2:18 have and I think generative AI is coming
2:20 a a shitload faster so uh I also by the
2:23 way swear sometimes if you don't like
2:24 the swearing thing you you don't want to
2:26 be here I'm you know I'm bitter I'm
2:29 Saucy I'll probably be here for a couple
2:31 hours and as it goes on I just you know
2:33 stuff happens anyway um if you have any
2:37 questions about AI pop them in the
2:39 comments below um happy to talk about
2:42 anything really um show you how to use
2:46 uh chaty PT talk about plugins talk
2:50 about news all the stuff that's
2:53 happening behind my head here my big
2:55 giant fat
2:56 head I got some URLs so so the reason I
3:00 put put these in here if you're new to
3:01 to generative Ai and you're curious
3:04 about it I strongly strongly encourage
3:06 you to start playing with it like now
3:07 today so that top URL is the official
3:11 chat GPT website so chat.
3:13 open.com Bing is Microsoft's search
3:16 engine but if you click on the chat
3:17 button that's also chat GPT and it's GPT
3:20 4 and it's free and then po gives you
3:23 access to six different um chat engines
3:28 uh including one that's got 100,000
3:30 token context window for those of you
3:33 that uh are are a little geeky and
3:36 understand what a context window is and
3:37 what tokens are and I'm happy to talk
3:39 about that too it basically means you
3:41 can pop 75,000 words into your chat
3:43 prompt and it doesn't explode uh which
3:46 is amazing you could pop a whole novel
3:47 in there you could pop in I put in uh
3:50 2001 of Space Odyssey and it just read
3:54 it and then I could ask it questions
3:55 about the screenplay like having written
3:58 seven screenplays the fact that you can
4:00 just pop a screenplay in and have it
4:02 read it is amazing I'm I'm uh my
4:05 beverage tonight is Dr Pepper 23 flavors
4:09 I actually had chat GPT look them up for
4:11 me and it it gave me a
4:16 bunch all right uh if you have if you
4:18 have comments questions pop them in the
4:20 comments below oh man I wore a white
4:21 shirt this is going to be a disaster
4:24 going be able to see
4:25 anything um okay neither will you but
4:28 you know I maybe I'll go my shirt this
4:31 is
4:33 hilarious okay hi I come here for your
4:37 insights on the new tech Paradigm
4:41 awesome thank you I appreciate that I am
4:43 not a technologist I'm not a machine
4:45 learning expert I'm I'm more of a once
4:49 the technologists have built the thing
4:51 and gotten it to a point that you can
4:53 put it in something like Chachi BT and
4:54 give it to the world that's where I like
4:56 to come in so the internet had been
4:58 around for decades Tim burners Lee
5:01 invents this thing called the worldwide
5:02 web when I found out about it and
5:05 figured out what it was I'm like holy
5:06 [ __ ] this is going to be big pounced on
5:08 that and then I didn't think I'd see
5:11 another one of those in my lifetime and
5:12 sure enough uh last November 30th chat
5:15 GPT comes out and I about about two
5:19 weeks into playing with it I had my
5:20 Kevin mallister moment
5:24 the holy [ __ ] the world just
5:28 changed any tips on keeping my mid
5:31 Journey question in one place instead of
5:34 coll scrolling with others yeah you
5:36 create your
5:39 own you create your own Discord Channel
5:42 hang
5:43 on um just just go Google just go Google
5:48 I I
5:49 don't it it's too much for me I don't I
5:53 don't know I haven't done it myself I've
5:55 just been um complaining that all of my
5:58 prompts are comingled with everyone
5:59 else's but basically you create your own
6:01 channel and then you put a mid the mid
6:03 Journey bot in it you add a mid mid
6:05 Journey bot to your channel um so just
6:07 go Google that and and that's how you
6:09 can do
6:10 that um I keep losing while it's
6:13 generating you can also search
6:16 for um messages sent to you in Discord
6:19 here let me show
6:28 you so if we go over to
6:31 Discord and you go in the uh little
6:35 search window in the upper right right
6:37 here you can search for mentions and
6:40 then put in
6:43 your your thing and it should show up in
6:47 there yes maybe uh maybe not
6:52 huh yeah oh no that's Roger dabling an
6:55 art I didn't do those no never mind go
6:59 start your own channel Channel I don't
7:00 know I don't know what I'm talking about
7:02 I do not know what I'm talking about
7:06 yeah welcome to the AI learning lab
7:09 where we're all learning all the
7:12 time and where cluelessness is a virtue
7:20 welcome all right um what else is worth
7:24 looking at here um on these fancy URLs
7:27 behind my head there's also prompt. chat
7:29 so if you're new to chat GPT and you
7:31 can't if you've heard people talk about
7:32 prompting and you don't know what the
7:34 hell they're talking about go there it's
7:36 a document that will teach you um about
7:40 prompting and then it's a whole list of
7:41 different personas you can pop into chat
7:43 GPT it's really interesting and then
7:45 futur
7:58 pedia.com is this sugar no it's high
8:02 fructose corn syrup welcome to America
8:05 where we've perfected the art of
8:10 diabetes okay CET gbt has been most
8:14 successful for me generating email
8:16 subject lines it's awesome yeah that's
8:17 great that's it's good for that what did
8:19 I do today oh I did this cool thing
8:22 where I had a client that was like oh
8:25 you'd be really cool if you wrote me an
8:26 article on blah blah blah right so I
8:29 this article and months have gone by and
8:32 so I talked to her today and she's like
8:34 what did I have you write that for she
8:36 didn't even remember and she had since
8:38 come up with this form for these
8:40 guidelines for different concepts and
8:42 she said oh well why don't you send me
8:43 the article you wrote and then maybe we
8:45 can figure out if there's Concepts in
8:47 there we can put into these
8:49 guidelines and so I thought hey let me
8:52 try that so I took her her template for
8:55 the guidelines and I took this long
8:57 article that I wrote and I said pull out
9:00 concept A and populate the guidelines
9:03 document and then pull out concept B and
9:05 pop and it just [ __ ] did it it was
9:07 brilliant it still took me about two
9:09 hours to kind of sift through the
9:10 answers and refine the prompts and
9:12 things like that but chat jpt is very
9:14 very good at uh parsing other documents
9:17 and helping you do things any idea when
9:19 co-pilot will push to all users of 365
9:23 not soon
9:24 enough I do I have no idea I haven't
9:27 even heard any rumor
9:30 you saw it again with smartphones I
9:33 didn't really see smartphones as big as
9:36 the worldwide web You could argue that
9:38 but I see it as kind of a
9:40 miniaturization of the personal computer
9:43 so you could argue that smartphones were
9:47 were another uh technological Revolution
9:50 and then social media was another one
9:52 but I see social media just as a a
9:54 simplified version of a website and a
9:57 smartphone just a a miniature ized
9:59 version of a PC um but the worldwide web
10:03 like fundamentally changed how we
10:05 communicate so I guess I guess the phone
10:07 did I don't know it these are all
10:09 they're all kind of semantic arguments
10:10 it doesn't really matter um I still
10:14 believe though that generative AI is
10:16 going to live in the pantheon of big
10:19 Tech revolutions more like the steam
10:21 engine so like big massive profound
10:25 change so the steam steam engine you
10:27 know uh brings on the industrial
10:29 Revolution I think generative AI like it
10:31 just changes the way you even think
10:33 about computers and programming and
10:35 things like that it's just absolutely
10:37 wild all right so if you've got uh
10:39 questions about AI pop them down below
10:42 I'm going to go change my shirt I think
10:43 cuz I can't see and my eyes are bad
10:47 enough already
10:50 um is there a paid version of chat GPT
10:53 that allows you to work with info after
10:55 September 2021 yes there is and and also
10:58 bing.com
11:00 if you click on the chat button that's
11:02 chat GPT 4 and it's connected to the
11:04 internet so so bing.com chat GPT um goes
11:08 beyond September 2021 and then if you do
11:11 the paid version of of chat GPT to get
11:15 access to GPT 4 you have to enable the
11:18 beta features um but then one of them is
11:21 search with Bing which does that what do
11:24 you think about one shot learning models
11:27 um I think that there so so basically
11:30 one shot means you you you know give it
11:32 an example and it it follows that it's
11:34 really really good at that
11:37 um so I don't quite know what your
11:41 question's asking I mean all of these
11:43 models do you know some version of zero
11:45 shot one shot few shot
11:48 learning I I would say the the broader
11:50 my broader comment on that is
11:53 that the big learn
11:57 learn the voice ain't working so good
11:59 tonight
12:01 boys the hell is going
12:06 on the big learning I had about chat GPT
12:09 and these large language models was
12:13 um I feel like for two decades Google
12:17 has trained us to make your prompt as
12:20 concise as possible just put put in the
12:23 minimum amount of words necessary right
12:26 cuz if if you make them too complicated
12:28 in Google it's just you get [ __ ]
12:29 everywhere right it's kind of the
12:32 opposite with chat GPT and and it's it's
12:34 much less about a single prompt giving
12:37 you the perfect answer that's I think I
12:39 think that's one of the big mistakes and
12:40 why people get disillusioned and and
12:43 disappointed early on with chat gbt
12:46 because they put in a shitty little
12:47 prompt and it gives them a shitty little
12:49 answer and they're like see I told you
12:50 it was stupid it's just like Google
12:53 except it only goes until September 2021
12:56 you loser and that's unfortunate because
13:00 I personally feel that you owe it to
13:02 yourself to go there and play long
13:04 enough until you have that moment cuz it
13:06 will come um and and for me it was when
13:10 I started interacting with it less like
13:13 a search engine and more like a um
13:17 collaborator like I almost thought of it
13:19 in in in the early days like like we
13:22 just hired an
13:23 inter that needs a lot of
13:26 Direction and lies but is the smartest
13:30 [ __ ] intern you've ever worked with
13:32 he knows everything that's what it's
13:35 like she knows everything Sydney's her
13:36 name actually I'm pretty sure
13:42 um the other thing I do here which
13:45 annoys the [ __ ] out of people is I try
13:47 to answer all the questions so as the
13:50 live goes longer and longer I get
13:52 farther and farther behind on my
13:54 comments and so the new people come in
13:56 are like he's not paying attention to me
13:59 I have a
14:00 complex but I'll get to you all
14:11 right what about AI for trading stocks I
14:14 think it is going to be remarkable for
14:17 that there's I don't know there was some
14:19 bot out of the University of Austin or
14:21 somewhere in Texas that that the bot had
14:23 some sort of 500% returns but I don't
14:26 know over what time frame so you know
14:29 any bot can have 500% returns in short
14:32 time frames right I think in the end
14:34 they all kind of normalize to
14:36 the to the uh to the average but um but
14:40 it's really good at that stuff it's
14:42 really good it can write you bots it can
14:44 merge to different Bots you can give it
14:46 one bot and tell it to write you a new
14:48 one with these updated triggers and
14:52 mechanisms I don't know [ __ ] about
14:53 trading so U but it's good it's good
14:56 with that stuff I you know so what I
14:58 would do is I would go to to chat
15:00 GPT and tell it to act like a an expert
15:05 um day trader or whatever kind of
15:07 trading you want to do derivatives
15:08 whatever it is and then have it start
15:10 teaching you how to create Bots or how
15:13 to create trading systems and then just
15:16 give it you know here's what I know
15:18 here's what I don't know here's my goals
15:19 here's what I want you to do and let it
15:22 start solving [ __ ] for you and then have
15:24 it write you a bot like figure out a
15:26 strategy together together it's like
15:29 your little buddy say hey little buddy I
15:31 want to come up with a little stock
15:32 trading strategy here he'll be like sure
15:35 that's really exciting stocks can be
15:38 fun at some point it'll probably say you
15:42 know I'm just a large language model I
15:43 really can't give Financial advice but
15:45 uh if I were to give you Financial
15:47 advice
15:49 here I really can't write a bot for you
15:51 but if I could here it
15:54 is so it's
15:56 fun what is one shot so so one shot few
16:00 shot and zero so zero shot is just write
16:03 me a poem one shot is here's a poem I
16:06 like write me a poem that's like that
16:09 about subject X and then F shot is you
16:12 can give it multiple examples the the
16:14 other thing that that is important about
16:16 um prompting that I found is that it can
16:19 be very tempting once you sort of get
16:21 that it can do complicated kind of
16:26 responses I initially found myself right
16:29 in increasingly complicated prompts that
16:31 were this a large single prompt and and
16:34 what I found is better is break your
16:36 prompt into little subsections so rather
16:39 than say write me the Great American
16:41 novel and you write this you know you
16:43 essentially write the first chapter of a
16:45 novel version of a prompt to try to get
16:48 it to write the Great American novel
16:50 start with like you know I want you to
16:52 be a writing coach and I'm going to
16:55 write a novel and and first let's start
16:57 with story structure and and give me
16:59 three different story structure um
17:03 approaches that that um might help me
17:06 get there and it'll talk about three
17:08 different you know here's the hero's
17:09 journey and here's save the cat and
17:11 here's Sid fi's screenplay thing and
17:15 then you can say oh that that saved the
17:17 cat one's interesting can you tell me
17:18 more about that sure here you go well I
17:20 have some ideas here about the novel and
17:23 here they are you know how would that
17:24 fit in that structure oh here you go you
17:26 know what don't you know I don't know
17:28 this and if you break it into little
17:30 segments like that the the more context
17:34 that you give cat GPT it gets better and
17:37 better and better so it's almost like
17:38 the harder you push it the better it
17:40 gets Bing is beyond useless rather use
17:44 perplexity a AI I do not disagree with
17:46 you black I so the reason Bing sucks
17:49 right now and you know may for a
17:52 while when they launched it it was very
17:54 very capable and it it it had less guard
17:57 rails on it than GP T4 at at open AI
18:01 does today or or they were different
18:03 guard rails but it was it was pretty
18:06 open it was pretty adventurous um like
18:08 you could do unlimited back and forths
18:11 and so this reporter from The New York
18:13 Times times named Kevin Rose spent four
18:16 hours with with uh chat GPT at Bing and
18:20 he got it to admit that its name was
18:22 Sydney and then at some point it [ __ ]
18:25 snapped and it told Kevin it was in love
18:28 with him and he should leave his wife
18:30 and
18:32 so it cracks me up every time I say that
18:36 so Microsoft had to lobotomize Sydney so
18:39 so Sydney got a little snip to
18:44 to that's why it ain't good so cuz they
18:48 got they got many many more safety guard
18:50 rails on it right now and you can only
18:52 do like whatever 15 or 20 back and forth
18:55 before it shuts you
18:57 down in the end I suppose it's good they
19:00 want they want safety have I used mid
19:03 Journey 5.2 yet yeah mid Journey 5.2 is
19:05 great if you want I can go play with
19:07 something
19:08 there I'll show you one of the cool new
19:11 features is the zoom out feature oh I'll
19:13 got to show you two new things so po.com
19:16 one of the sites up
19:18 there lets you play with Bots and we're
19:22 going to go play with a bot on
19:24 PO so this is
19:27 Po Let Me Clear My context so when you
19:31 go to PO so PO is just like a
19:33 normal um what's it called chat chat
19:37 engine so so you've got access to Sage
19:40 which is which is their model you can
19:42 you can pay for subscription to gp4
19:45 Claude plus Claude instant 100K I pay
19:47 for those because of the 100K one that's
19:49 the one that you can put 75,000 words in
19:52 um you can play you can use clot instant
19:54 chat GPT but there's also bots so here's
19:56 a Shakespeare oh wait sorry um um sorry
19:59 I was talking before about all these
20:01 different models here here are the the
20:02 top six are the different models you can
20:04 play with and then the things below it
20:06 are Bots and like I created this one
20:08 called I am Shakespeare so if you go to
20:10 po.com am Shakespeare you can interact
20:13 with Shakespeare and it'll he'll he'll
20:15 write you some I Amic pentameter for all
20:18 of your answers it's really annoying but
20:21 but if you like reading stilted English
20:23 that doesn't rhyme it's perfect for you
20:26 um but there's also there's there's a
20:28 bot called Journey this one's called MJ
20:30 captain that one of the guys on my on
20:32 one of these lives uh showed me he he
20:35 created this um and what this is is you
20:39 can write a crappy prompt for Mid
20:41 Journey so we'll do let's see what are
20:44 we going to do um I'm in Denver we'll do
20:48 Denver Skyline Sunset big
20:53 storm Over the
20:57 Rockies okay so that's a crappy prompt
21:00 and then what it does is it rewrites it
21:02 a photograph of the Denver Skyline
21:04 during sunset with a stormy sky over the
21:06 Rocky Mountains in the background using
21:08 a wide angle lens in the style of anel
21:10 adoms with cool blue and warm orange
21:13 tones capturing Dynamic uh I don't think
21:17 anel Adams ever did color photographs
21:19 did he but whatever we'll just take it
21:22 wholesale and see what we get so okay so
21:24 now so now we're going to pop over to
21:26 Discord why are you going Discord Kyle
21:29 isn't that just for shitty kids to play
21:31 games on yep
21:34 but now it's for old people too okay so
21:38 Discord if you come here these things
21:40 down the Le hand side are servers so
21:42 those are basically different
21:43 communities like that's an adobe beta
21:45 testing one that's Runway ml that's a
21:47 thing called neotokyo that's a thing
21:49 called Legion this top one that I'm in
21:52 is the AI Salon that's that's a salon
21:54 that I created that's a community of
21:56 people into AI so that's that's the one
21:58 we're in and then this next column is
22:01 these are the different conversations so
22:03 we talk about things like you know tools
22:05 and things like that but mid Journey
22:08 this channel lets you go make stuff with
22:11 mid
22:12 Journey so oh that's cool who who did
22:15 that oh that's Ally Allie's one of the
22:18 organizers of the salon that's really
22:19 cool that's actually really beautiful
22:22 dag um by the way I'm a huge fan of mid
22:25 Journey so you know right now if you
22:28 want high quality images mid journey is
22:30 it okay so what we're going to do the
22:33 way you make a an image in mid journey
22:35 is you go slash
22:37 imagine and then it'll give you an area
22:39 for a prompt and since we had that bot
22:41 just write as a good prompt we're just
22:43 going to pop it in
22:44 there um it gave us an aspect ratio of
22:47 3x two I'd rather have 16 by9 cuz that's
22:51 how I roll and da da it's using 5.1 I'm
22:56 going to flip that over to 5.2 that's
22:58 the latest model and then you hit return
23:01 and it will send that off to the mid
23:04 Journey servers and then what you'll see
23:06 here shortly once it unless it's kind of
23:09 hung sometimes it gets busy people
23:11 people have been loving 5.2 okay this is
23:13 perfect so you see how it's starting to
23:16 it's it starts as just these blurs and
23:18 blobs the way the diffusion models work
23:21 is when you do your prompt it basically
23:24 goes and finds a bunch of sort of
23:27 mathematical probability ities
23:31 of noisy images that that you know were
23:35 tagged with words similar to what you
23:37 put in there it kind of then averages
23:38 them together and every image starts as
23:41 this sort of random noise image that
23:43 over a number of Cycles gets denoised
23:46 and denoised and denoised um to to come
23:49 up with the final image so each image is
23:51 actually being generated from scratch so
23:54 it's it's they are actually original um
23:57 creations
23:59 um from what they've learned from what
24:01 they've been trained on but they're not
24:02 copying and pasting it's one of the big
24:05 misconceptions uh about about these
24:08 models so this is actually a pretty cool
24:10 image or set of images where Denver's
24:13 way off in the distance the clouds are
24:15 super dramatic those are pretty slick
24:18 here let me pop it over in a browser we
24:20 can zoom
24:22 in that's not bad look at that looks
24:26 like I25 heading into Denver
24:29 that looks like another part of I25
24:31 heading into
24:32 Denver yeah they're
24:35 all but let's
24:38 see I think they have the mountains
24:40 wrong in that one they're like the wrong
24:42 direction but anyway that's that okay oh
24:46 so I was going to talk to you about
24:47 zooming out oh so let me let me do
24:49 another image that might be better for
24:51 for the zoom out
24:53 feature well we can do one of these
24:55 let's pick one of these that we like I
24:57 like
25:01 let's do this so the way you count these
25:04 is 1 2 3 4 and then you've got U1 2 3 4
25:08 and V1 2 3 4 u12 3 4 is upscale V is
25:12 variation so if this one's close but
25:14 it's not quite right then you can hit V3
25:17 and it'll give you four different
25:18 variations on that I like that good
25:20 enough so I'm just going to hit upscale
25:22 3 and so now that's going to fire off
25:24 and give us a higher resolution version
25:27 of that so there it is Saucy I like it
25:31 it's very nice buck that's a nice
25:34 picture you made now let's
25:37 say we want to see more of that Sky you
25:40 now have these things called zoom out
25:43 zoom out 2x 1.5x custom zoom and make
25:47 Square so maybe what I'll do is I'll do
25:49 make Square first so it should extend
25:52 the top and the bottom of this thing so
25:54 we'll see how that
25:57 goes
26:11 come on Mid Journey all right there you
26:13 go yeah so you can kind of see down at
26:15 the bottom here and at the top they're
26:18 very um blurry the middle was in in
26:21 Focus but the tops were blurry so now
26:24 it's made it a square
26:27 image
26:29 and the sky is
26:44 bigger and you can see there's slight
26:46 variations like this has like this round
26:48 Cloud these are more sharp this is just
26:51 like a Big Blob of clouds this is really
26:52 dark right so so slight variations on
26:56 the theme I kind of like that bottom WR
26:58 one a little bit more so it's done now
27:00 so I can say upscale that and now we'll
27:02 have a giant Square version of that
27:05 photo there you go so that's mid Journey
27:09 that's how you make mid Journey images
27:11 in Discord if your question is why do
27:13 you have to use
27:15 Discord um cuz they're dumb
27:18 Dums you got to use you got to use
27:21 Discord right now here's the deal we're
27:23 very early in this here internet or this
27:25 here generative AI thing and so
27:29 it's just the way it
27:31 is it's just the way it is oh welcome
27:36 everybody looks like there's some new
27:37 people here my name is Kyle Shannon this
27:38 is the AI learning lab I'm wearing a
27:40 white shirt for your reading pleasure I
27:43 I'll go change it here in a second um if
27:46 you're new here if you haven't played
27:47 with chat GPT I strongly encourage you
27:50 to do it I strongly encourage you to
27:52 give yourself the gift of playing with
27:53 it long enough until you have a Kevin
27:55 mallister moment and and for those of
27:57 you that are young it's a movie called
27:59 home alone and Kevin mallister did
28:03 this and chat GPT if you play with it
28:06 long enough you will stumble upon
28:07 something where you're like holy [ __ ] I
28:09 can do
28:10 that is my job in
28:15 trouble the answer
28:17 is you you you owe it to yourself to
28:19 learn this stuff well enough to have
28:21 that moment work through that trauma and
28:23 then start to realize okay this is a
28:25 tool that I can start to get my head
28:26 around and as you become more AI
28:29 literate and teach other people how to
28:30 do it miraculous things will happen like
28:33 you know you'll have hope again and
28:36 people will recognize oh that person
28:37 knows AI let me let me work with them um
28:40 so I disagree 1995 open someone just
28:44 commented they disagreed with my 1995
28:47 being a big deal all right well I'll
28:49 I'll I'll see you in the
28:52 comments uh so you can't copyright your
28:55 mid Journey images it is a very very
28:58 gray area right now
29:01 um you'll be able to for sure with Adobe
29:05 Firefly and with generative AI because
29:07 they trained their model on public
29:09 domain and uh Adobe stock images that
29:12 they own the rights to so if you want to
29:15 do it for a commercial I would do that I
29:17 think over time all these copyright
29:19 issues are going to get worked out right
29:21 now they're a bit of a mess um anyway if
29:24 you're new here looks like there's a lot
29:25 of new folks here welcome uh my name is
29:27 Kyle Shannon this the AI learning lab um
29:29 that top URL is the official chat GPT
29:32 website so if you haven't played with it
29:33 just go there while you hear me ramble
29:35 and start playing with it and if you
29:37 don't know what to do there like what do
29:39 I do with this silly little box go to
29:41 this site prompt. chat and that's a uh
29:45 chat GPT list of of prompts and and it
29:49 there's a primer for how to do prompting
29:51 at the beginning of it thank you what is
29:53 that a little teddy bear a little gummy
29:54 bear nice thank you um and and this will
29:59 also teach you that you can um assign
30:02 chat GPT to different personas like act
30:05 like a screenwriter act like a
30:08 uh bartender act like an actor act like
30:12 a whatever it doesn't matter and and and
30:15 you'll start to get that oh it can do
30:16 that and it can do that oh it can act
30:18 like an accountant can it do that can it
30:20 do that it's and it will blow your mind
30:23 and then if you want more rabbit holes
30:24 to run down futur pedia.com almost 4,000
30:28 AI
30:29 tools there and um yeah do you take on
30:35 any projects for clients I I'm happy to
30:37 talk about it if you go to my uh what is
30:39 it called profile uh my contact info is
30:42 there um it's also you can go to
30:46 storyvine is my
30:48 company I'm fairly doxed as they say in
30:51 the uh what is it the uh nft Community
30:55 oh you're going to dox yourself yeah
30:57 I've been doxed for like 30 years so
30:58 what am I going to
31:01 do but not docks docks you
31:05 know I'm not hiding I ain't hiding
31:09 nothing um I'm trying to figure out
31:12 where my where
31:14 my uh questions were if you've got if
31:16 you've got uh questions or comments
31:18 about AI pop them in the comments I do
31:20 my best to answer them all so what I'm
31:22 doing right now is trying to find uh the
31:24 last question I
31:26 answered and then and keep going and I
31:29 will I get very behind on answering the
31:31 question so if it seems like I'm not
31:32 answering you that's because I haven't
31:33 gotten to you yet it doesn't mean I'm
31:37 recorded unless I am maybe I'm chat GPT
31:41 5 have you ever thought of that cuz gp4
31:45 is pretty [ __ ] remarkable wouldn't
31:47 you think this would be the next
31:49 instantiation of the
31:55 technology why they choose a a guy that
31:58 looks like Rosie o' Donald be the next
32:00 chat GPT it just don't make any sense to
32:03 me
32:06 Marge all
32:09 right by the way I'm a smart ass I
32:12 swear um I don't know it's the
32:17 fructose this just in you've been
32:19 nominated for the worst Tik Tok shirt
32:21 ever fine I'm going to go change my
32:22 shirt I'll be right back don't go away
32:24 keep hitting the like
32:26 button
32:49 all right what's going on somebody all
32:52 right thank you look best shirt ever and
32:55 it's cool too right you
32:59 see fantastic Bob tell him what he's won
33:03 you've saved your eyesight for the
33:08 evening all right what about Ai and
33:10 stock trading we did that the intern
33:13 analogy yeah yeah cat GPT it's like it's
33:15 like the smartest intern you've ever
33:17 hired who needs a lot of Direction Ed
33:21 lies you need a moderator I do need a
33:24 moderator anybody want to be my
33:25 moderator I don't know how to do this
33:27 [ __ ]
33:28 I'm an old man on Tik
33:32 Tock um what's everyone in Instagram
33:36 doing with their chatbot timers are they
33:39 bringing in some AI filters I don't know
33:41 I haven't been paying attention to
33:42 Instagram at
33:44 all at all can it write an app for
33:48 iPhones I have a great idea but have no
33:50 idea how to code well there's a couple
33:51 of ways you can approach that one is it
33:53 can write code you still need to know
33:57 how to you know put that code into
34:01 Frameworks and action and submit it to
34:03 Apple and things like
34:04 that good news chat GPT can teach you
34:07 how to do all that right so you could
34:09 say you're a senior developer and I have
34:13 an idea for an iPhone app and together
34:14 we're going to do an iPhone app and
34:16 treat it like like an intern or like a
34:18 like a senior programmer it will code
34:20 for you it will tell you how to set up
34:23 Frameworks all of that the other way you
34:24 can do it is you can create um
34:27 apps using no code or low code
34:30 Technologies like I know there's one
34:32 called Bubble that you can make iPhone
34:34 apps with I use one called zapier which
34:36 is an automation tool there's another
34:38 one called make which is an automation
34:40 tool that's I think the fancier version
34:42 or the newer version of
34:44 zapier um and there there's other things
34:46 out there so um but yes you can in fact
34:51 well no I I'll I'll show you there's a
34:53 bunch of people here so I want to I want
34:54 to keep moving through questions but
34:56 just um
34:58 go go to chat GPT right now go there
35:00 right now and tell it to write you a
35:02 swift application that's a clone of
35:04 Flappy
35:04 Birds and watch it write code and then
35:08 and then do
35:10 this I had no idea it can do
35:15 that can it do other [ __ ]
35:19 too but certainly it can't write
35:22 poetry AI learning lab can you move I
35:25 can't read that I got a new shirt on for
35:27 you you Ivory this is all for you okay
35:30 is it open GPT I don't know what that
35:33 question means open AI
35:35 GPT chat. open.com is the official chat
35:39 GPT website chat.
35:42 open.com so that's the real deal that's
35:45 the that's the one that changed
35:47 everything where do you go for AI voices
35:49 music like make X sing like why I
35:52 actually haven't played with that I I
35:54 geeked out on it a little bit what I
35:56 would do is go to to um go to the
36:00 Twitter that's Elon musk's fancy new
36:03 company go to the Twitter and ask do a
36:06 search on Twitter for Drake the weekend
36:10 Ai and it'll pull up lots of threads
36:12 about that Drake weekend song that that
36:15 came out um and I I know there's some
36:18 threads in there that tell you what
36:19 tools they use to do it I just I don't
36:21 have them off the top of my head if you
36:22 just want to do vocal stuff then either
36:25 D- or 11 Labs is is what I've been
36:30 using Bing chat is so trash it is it'll
36:34 get better but it's trash right now
36:36 that's why so so I pay for chat.
36:40 open.com that's 20 bucks a month and I
36:42 pay for po they for for different
36:44 reasons
36:46 gp4 at openai's website is just the best
36:49 there's nothing even close to it period
36:51 end of story until further notice and
36:54 then po gives me the paid version of
36:56 that gives me access to the 100,000
36:58 token uh version of Claude which is
37:01 anthropics um model anthropics anthropic
37:04 is pretty good but they should be there
37:06 a bunch of xop AI employees that
37:08 invented chat gpt3 or not chat GPT gpt3
37:13 the underlying
37:15 model so those folks know what they're
37:17 doing in fact you remember six weeks or
37:20 so
37:21 ago four people went to the White House
37:24 four CEOs to talk about AI it was the
37:28 CEO of Microsoft Google open AI Sam
37:31 Alman and anthropic so the people that
37:33 do Claud they're they're major
37:35 players if you don't know them this is
37:39 so much about creative thinking do you
37:40 think we will teach AI to be creative uh
37:44 ai ai is is creative right now the the
37:48 generative AI stuff that's there is not
37:51 just parting stuff but here's the deal
37:54 Microsoft's branding for for the
37:56 generative AI is co-pilot and I think
37:58 that's
37:59 right there's one another misconception
38:02 that I hear is like oh the AI did this
38:04 the AI did that no no no a human being
38:08 prompted the AI to do something and the
38:11 human being went back and forth with a
38:13 bunch of times and at some point the
38:15 human being said this is really good I'm
38:17 going to put that out in the world
38:20 um I spent four hours today on chat
38:24 GPT doing
38:28 essentially a massive creative writing
38:31 exercise that would have taken me three
38:33 days and I did it in about three hours
38:38 um and it it it is
38:43 absolutely augmenting my creativity so
38:47 it is creative now is it aware of its
38:50 creativity no what it is technically is
38:55 so you know when you're when you're in
38:56 Google
38:57 you're writing a like a Gmail or
38:59 whatever and or Microsoft Word and
39:02 you're writing a sentence and then like
39:03 the three gray words pop up at the end
39:05 it's like the auto complete that's all
39:07 chat GPT is except like on steroids chat
39:12 GPT is like the Lance Armstrong of
39:16 autocomplete too
39:19 [Laughter]
39:22 soon it's the berry bonds
39:25 of of predictive predictive word
39:30 completion um which app do I download or
39:33 do I have to go to my computer to do the
39:35 AI chat the best way
39:37 okay super important Tammy great
39:40 question there is an official chat GPT
39:44 app for iOS there isn't one for Android
39:47 there are a ton of scam apps out there
39:52 so what I do on my iPhone is I
39:56 downloaded the official open AI app but
40:01 I don't use it because it's not as as
40:02 capable as the website so what I do is I
40:05 go into Safari on my iPhone or Chrome
40:07 whichever one you
40:09 prefer I go to that URL and then I make
40:11 that a bookmark that I put on my desktop
40:13 or my whatever the [ __ ] you call it on a
40:15 phone the desktop um it's it's a really
40:19 simple website so it works great on
40:20 mobile so I just I use the official
40:22 website on desktop and on
40:25 mobile that's why I have it here behind
40:27 me I'm not Hawking them I don't get paid
40:30 anything from any of these people I
40:32 don't give a [ __ ] if you go there or not
40:34 um but I have them behind my head
40:35 because it's really hard to it's really
40:38 e it's really easy to get confused
40:40 because there's so much crap out there
40:42 and there's just yeah it's just a mess
40:44 right now all right how is quantum
40:46 Computing going going to impact AI I
40:48 think fairly profoundly um I just I
40:51 don't know the speed with which I know
40:53 that
40:54 um quantum computers they're adding
40:57 they're doubling the number of cubits
41:00 every year for like the past decade
41:01 they're up to like
41:03 53 or 520 something cubits which every
41:07 time they add a cubit it's like some
41:08 exponential crazy ass increase in power
41:12 so I would imagine we're probably
41:15 getting into the neighborhood where we
41:17 might be using Quantum Computing for
41:18 training these models things like that
41:21 and and I would assume that the the
41:26 quantum stuff is going to start creating
41:28 new kinds of paradigms where AI just
41:31 gets stronger and stronger and stronger
41:34 because a lot of the stuff we're playing
41:35 with today is all leveraging technology
41:37 that Google invented in 2017 this thing
41:39 called the
41:40 Transformer um so could I please demo
41:44 zoom out on Mid journey I did that
41:45 already what is mid Journey it is the
41:47 best little image generating tool in all
41:50 of the
41:51 land there's so for image generation
41:53 there's kind of three major players in
41:55 my opinion that there's um Mid Journey
41:58 which I think is the best creatively
42:00 there's stable diffusion which is the
42:02 most flexible and and probably the most
42:04 powerful in terms of features because
42:06 it's open sourced and the developer
42:08 Community is doing all sorts of [ __ ] on
42:10 it um but it requires you to be the most
42:12 technically Adept to keep up with it and
42:15 get your head around it and then there's
42:17 Adobe Firefly and and uh generative fill
42:21 within Photoshop
42:23 um if you're in a commercial business
42:26 especially work you're working for a big
42:28 company use Adobe um it's not as good
42:32 quality-wise but you won't get in
42:34 trouble it's it's commercially safe it's
42:37 like you ever hear that that saying
42:38 nobody ever got fired by recommending
42:40 IBM it's kind of like that with Adobe uh
42:43 generative AI right now all right I'm
42:45 writing a book is this a is this good to
42:48 use to edit oh yeah it's It's [ __ ]
42:51 amazing just just try it go there right
42:54 now just grab one of your chapters pop
42:56 the chapter in and tell it to find
42:58 grammatical errors or tell it to find
43:00 typos
43:01 or tell it to rewrite your chapter in
43:05 the style of your favorite author it's
43:08 It's [ __ ] amazing at that stuff
43:11 um yeah you yeah you I yes yes it's very
43:15 very good at that tell it you want it to
43:17 be a book
43:19 editor um Mid Journey generates images
43:21 from natural language descriptions
43:23 called prompts very good I like that
43:25 you're supporting one another educating
43:27 one another that's the whole idea of
43:28 this place um it looks like we have a
43:30 bunch of people here welcome everyone my
43:32 name's Kyle Shannon this is the AI
43:33 learning lab um I'm an entrepreneur I'm
43:36 the founder and CEO of a company called
43:38 storyvine which is 11year old an 11y old
43:41 video technology company video
43:43 storytelling company um I was a founder
43:45 of agency.com back in in '94 in the mid
43:49 90s and so um experien the early days of
43:52 the worldwide web and I'm having deja vu
43:56 all over again with this generative AI
43:58 stuff there is a
44:01 profound there is profoundness coming
44:04 magic is in our hands right now that's
44:06 going to change a lot of things quite
44:08 profoundly and it's coming quite quickly
44:09 which is why this channel exists do you
44:12 have YouTube on how to use mid journey I
44:14 don't uh I have do I have a I don't
44:17 think I did a mid Journey Tik
44:20 Tok but I but I uh just put it down
44:24 below if I still haven't answered it
44:25 because I just demoed it
44:27 about 10 minutes ago so hopefully you
44:30 saw that we are creating all these
44:31 images how about copyright uh it's a
44:33 mess right now
44:35 so gener generally the the artists that
44:38 I know that are using mid journey and
44:40 stable
44:41 diffusion if they're going to use
44:43 anything commercially they don't
44:47 um they don't invoke invoke a specific
44:50 artist's name if they want something to
44:53 look like a certain artist they figure
44:55 out how to prompt without invoking their
44:57 name that's that's a line of defense um
45:01 whether or not you can actually
45:02 copyright the work right now uh the
45:04 copyright office is saying no I assume
45:07 that's going to change over time and
45:09 here's why I think that um so right now
45:13 you go to Mid Journey you generate an
45:15 image but I can also go to photoshop and
45:18 I can take a photograph that I took and
45:20 I can do generative fill so I can take a
45:23 vertical photo and make it horizontal by
45:25 filling in the background still my photo
45:28 I'm just like adding some background in
45:30 there is that AI generated or is that my
45:32 photo uh it's both I don't know and like
45:36 at which point like how many pixels what
45:39 percentage of pixels need to be the
45:41 pixels that I generated from my digital
45:45 camera where it's no longer mine or it's
45:48 all AI like I don't know so I so all of
45:51 this stuff's up in the air right now
45:53 like I said if you want to be safe with
45:56 um the copyright stuff with images use
45:58 Adobe Firefly and and generative fill in
46:02 Photoshop how can you copy the image I'm
46:04 not quite sure what you mean there
46:06 Nature's Blessing but if you clarify
46:08 that I'll ask is mid Journey free no
46:10 it's not you can do like 20 images but
46:12 but just about the time you got it
46:14 figured out they're like you got to pay
46:15 now and I resisted paying for a long
46:18 time I was using stable diffusion like
46:20 this image behind me that was done in
46:21 stable diffusion with dream
46:24 Booth um
46:27 and uh but I ended up paying for Mid
46:30 Journey because it's just better hey use
46:32 chat GPT to generate a mid Journey
46:34 prompt yeah it's really good you can
46:35 take like the documentation for how mid
46:37 Journey does its prompt paste that into
46:39 chat GPT and say you know learn this
46:42 documentation and then I'm going to type
46:44 in a prompt and you add [ __ ] to it based
46:47 on the documentation have a
46:52 hangnail yay I'm here
46:55 early Miss Burch welcome hey dude do you
46:59 take on any projects for clients like I
47:01 said contact me go to my what's that
47:04 called profile I've got contact
47:06 information there is anybody getting
47:10 double sound oh am I am am I
47:14 echoey sound is normal for
47:17 me oh it just jumped me to the bottom
47:19 again sorry I just lost my place in the
47:21 questions I'm just going to scroll back
47:23 up and pick some random spot to keep
47:26 going
47:27 from uh
47:29 uh what's the best AI for engineering
47:34 blueprints um I don't know what I would
47:37 do if I were you is I would go to
47:39 futurepedia
47:41 doio I would imagine that there are a
47:43 lot of tools now that are putting large
47:45 language model front ends on top of
47:49 complicated pieces of software like
47:51 unity and Unreal Engine and 3D modeling
47:54 software and things like that um so I
47:57 would go I would go there and start
47:59 exploring I don't I don't know that
48:00 world well enough to know you can have
48:02 mid Journey do um like you know
48:06 architectural designs for you and
48:07 interior designs and it can mimic
48:10 blueprints but it's not engineering
48:11 blueprints right so
48:15 so mark my words that every
48:18 single CAD piece of software out there
48:20 any complicated piece of software in the
48:23 next year is going to have
48:27 um AI assistance in them because it's
48:29 just you can just ask it for what you
48:31 want and it'll go figure out how to do
48:33 it it's amazing for
48:35 that um Can it translate text from one
48:38 language to another yes in fact one of
48:40 the first things I did was I I took
48:43 something that I wrote I translated it
48:45 into
48:46 Dutch Spanish
48:50 Japanese uh Portuguese and then back to
48:53 English and I could still read it it's
48:55 it's amazing at that knows all the
48:56 languages it knows all the languages it
48:59 knows all the programming languages too
49:01 or at least you know the biggies um so
49:03 it can refactor code for you it can
49:05 document code for you if you're an
49:06 engineer and hate documenting code just
49:10 plop some code in there and tell it to
49:11 document it tell it to explain how it
49:14 works and what it's doing it will it
49:16 knows [ __ ]
49:24 amazing thoughts on on Eliza's alignment
49:28 concerns I don't know
49:30 elizer
49:32 um but the alignment concerns are very
49:35 real so
49:36 so you know the the idea of alignment
49:39 and the fact that it hallucinates and it
49:41 it doesn't check its work and it it I
49:44 mean the tool right now is is as
49:48 sentient as a calculator right
49:50 calculators are remarkable in terms of
49:53 if you push the right buttons you get
49:55 Mass
49:57 but the calculator is not aware of what
49:59 it's doing right it's just a bunch of
50:02 circuits stitching together formulas
50:04 right well it's the same [ __ ] with large
50:07 language models the the challenge with
50:09 it
50:11 is what it does is so sophisticated it
50:14 looks like it knows what it's doing
50:15 right it looks like it's um sentient and
50:20 reasoning and doing things like
50:24 that there are experiments that have
50:26 been done that say you know it is
50:28 showing signs of reasoning but it's not
50:31 like it's not aware of what it's doing
50:35 so I I think that companies like
50:38 anthropic are working really hard on
50:40 alignment I I would say anthropic is
50:42 probably a company that is
50:47 thinking most clearly to me about
50:50 solving that but I think there's going
50:51 to be a whole industry built around
50:54 alignment and safety and
50:57 accuracy I I I've said this on my lives
51:00 before I think the 2024 elections are
51:03 going to be
51:06 so dramatically Amplified the the
51:09 misinformation and the Deep fakes and
51:11 all that [ __ ] that after the election
51:14 there's going to be a massive push for
51:17 authenticity and accuracy and original
51:20 content so so we'll see so anyway I I'll
51:23 look
51:24 up elizer or however you say his name
51:27 her name what's the ba best AI for
51:30 engineering blueprints I already
51:32 answered that Holly with tradein don't
51:35 know what that
51:36 means um Dr rymes asking a bunch of
51:40 times doesn't help because I'm way
51:43 behind so I appreciate your enthusiasm
51:46 all right what's the most cool thing
51:48 you've done with AI so far
51:54 uh
51:58 for me it was the the
52:02 the the the thing that gave me the
52:04 Epiphany about holy [ __ ] the world just
52:06 changed but no one knows it yet was like
52:09 three weeks after chat GPT came out I
52:12 was sitting with my wife watching TV and
52:15 I knew that it could code and I haven't
52:16 coded in like three decades and I'm like
52:19 oh I heard Python's kind of cool let me
52:20 see if I can do python so I said you
52:23 know write me some python code and it
52:24 wrote me python code and then I was like
52:26 I don't know how to run python code and
52:29 I said do you know how to run python
52:31 code and it said sure you open your
52:32 terminal and you type in this command
52:34 line
52:35 and and I'm like I don't want to do any
52:37 of that isn't there just a website that
52:39 I can just paste the code in and it runs
52:41 and it said sure here's a bunch and the
52:42 first one it gave me was repet which is
52:44 a good
52:45 site and so it taught me paste the code
52:48 in the leftand window hit the play
52:49 button and it'll run your code so I
52:51 pasted the code in the left window hit
52:53 the play button and it gave me an error
52:55 I said said it gave me an error and it
52:56 said oh that's cuz you don't have this
52:57 Library installed I'm
52:59 like thanks [ __ ] how do you install
53:02 the
53:04 library I don't think I called it
53:06 [ __ ] but you could and it told me
53:10 how to do that it said in the lower
53:11 leftand window you click on this button
53:13 then you go to the upper right hand
53:14 window you search for that Library you
53:15 hit the install button and you're Off to
53:17 the Races and so I went to rep it I hit
53:19 the button I did the that I hit the
53:21 install button I hit play and it worked
53:23 and within 90
53:24 minutes had
53:26 written a functional python application
53:30 that took some inputs on a topic someone
53:32 would type in a topic answer a few
53:34 questions about it and it would generate
53:36 five pieces of social media content
53:39 original social media content in 90
53:42 minutes I've still never written a line
53:43 of python in my life so for me that's
53:45 the one where I
53:47 went this is this is a fundamentally
53:51 different tool than I've ever had my
53:54 hands on it's a fundamentally different
53:57 tool and it's really hard to understand
54:00 that from the outside like you kind of
54:03 have to have your own moment and then I
54:06 would say like to to your point about
54:08 like what's what's the most cool thing
54:10 you've done
54:11 like like today I was on the phone with
54:14 a client I spent I don't
54:16 know the past month or so putting
54:19 together this seven or eight page
54:21 article that they requested I
54:23 write and then when I told them it's
54:26 ready for them to review they're like
54:27 why did I tell you to write that oh yeah
54:29 we have this form we have a different
54:31 approach
54:32 now I guess send me the article and
54:35 we'll look through it and we'll see if
54:36 we can pull anything out to put it in
54:38 this new
54:39 format and
54:41 so I thought well wait a minute I bet
54:44 chat PT could do that so I took the
54:46 article that I wrote I took this form
54:49 that you can fill [ __ ] in
54:51 in and I said go through my article pull
54:54 out these three concepts steps and
54:56 create three different versions of that
55:00 form bang done bang done and then you
55:03 know the first couple of times they
55:04 sucked and I said okay it sucks because
55:06 of this that and the other add this
55:07 detail take that one
55:10 out and then at one point I had an idea
55:13 that I wanted to add a whole new Concept
55:14 in so I just said give me a new one with
55:16 this concept out it came so I like I
55:20 feel like two or three times a week I'm
55:24 having jaw-dropping moments where I'm
55:27 just like this has fundamentally altered
55:30 my creative process it's fundamentally
55:32 altered my work
55:34 process and it does it daily so I don't
55:36 know maybe just the fact that I'm doing
55:38 it is the coolest
55:40 thing um hey what tools can I use to
55:43 help create blog post and content for a
55:45 brand new web
55:46 store the one behind my head the one
55:49 behind my head so listen all of the
55:52 tools you see like Jasper and copy. and
55:54 and like all the specialty ones that are
55:56 like you know come here and we'll write
55:58 blogs for you and we'll write P for you
56:01 all those are are sort of pre- prompted
56:04 versions of
56:06 that I think right now it's actually
56:08 important for
56:10 anyone learning to use these tools go
56:13 start to understand how prompting works
56:16 and that's why I've got this thing here
56:18 so one of the things that you could do
56:20 is you could go here find one of the
56:24 prompts that says act as a social media
56:27 writer or
56:29 manager and tell chat GPT to act like a
56:33 social media manager then give it the
56:35 topic of your blogs the other thing you
56:36 can do is say here's the topic of my
56:38 blogs and here's an example blog that I
56:40 wrote that I'm really happy with so
56:43 follow the style and form of my blog but
56:45 go right on these 10 subjects or give me
56:49 10 subjects and I'll tell you which five
56:51 to write and then follow my example so
56:55 you you can do it right there for free
56:58 chat GPT
57:01 3.5 if it were me I'd pay the 20 bucks
57:04 and use GPT 4 but because it's way
57:08 better um but that's it what are those
57:12 audio websites wants more uh which audio
57:14 websites oh 11 laabs docomo spell out 11
57:20 e Ven labs.com and then d
57:26 ID what about Bard AI Bard sucks right
57:29 now so so Bard is Google's largest
57:33 language model I mean the ironic thing
57:35 about Bard sucking right now is they
57:38 [ __ ] invented the technology that's
57:40 powering all this [ __ ] they invented the
57:42 Transformer so the history kind of goes
57:44 like
57:45 this machine learning was a this very
57:48 specialized kind of thing you like you
57:51 had to sort of bring your own data you
57:53 could only train these machine learning
57:56 models on relatively small data sets or
57:59 specialized data
58:00 sets and in 2017 Google writes this
58:05 paper called attention is all you need
58:08 where they introduce the concept of a
58:09 thing called the Transformer and and in
58:12 very very simple terms what it does is
58:14 it makes it possible to train machine
58:15 learning models on massive data sets so
58:18 they invent this [ __ ] and then they kind
58:21 of shove it I think like is my sense
58:24 right because I think they recognized
58:27 that this is a this is a technology that
58:30 if they built it it would threaten their
58:31 search business right because if you use
58:35 chat
58:37 GPT and you ask it for an answer it
58:40 gives you an answer it doesn't give you
58:42 40 links you know 36 of which are SEO
58:46 [ __ ] sites that you have to sip
58:47 through it just gives you an answer so
58:51 what open AI did with chat GPT is is an
58:54 existential threat to Google but when
58:57 they launched that thing it got to 100
58:59 million users in six weeks fastest
59:02 adoption of technology and
59:06 history so Google went oh [ __ ] we've got
59:10 some AI too we invented this [ __ ] look
59:13 here oh here's version two it's even
59:15 better well it ain't even [ __ ] close
59:18 because these guys what what Sam Alman
59:21 and and the open AI team did
59:23 is in in devel in gpt1 and gpt2 and then
59:27 gpt3 by the time they got to gpt3 they
59:30 recognized that the more compute power
59:32 you threw at this thing and the bigger
59:33 the data set was the more powerful it
59:36 got so they spent a lot of time finding
59:39 the best data sources they could so it's
59:41 trained up on like all of the internet
59:43 and they went into all the corners of
59:45 the internet right so all of Wikipedia
59:48 and blogs and websites and
59:51 research database whatever whatever the
59:53 [ __ ] is out there they got as much of it
59:55 as they could so their core model is
59:59 really solid and I think Bart is just
1:00:01 not there they're trying to fine-tune it
1:00:03 to get it better and it's getting better
1:00:05 I think Bard is probably
1:00:08 um it it's certainly better at at code
1:00:12 than it is at language stuff um but but
1:00:16 gp4 is [ __ ] amazing at code although
1:00:19 my friend Greg says he thinks gp4 is
1:00:21 getting worse lately at code so that's a
1:00:23 possibility too um why GPT gives so many
1:00:27 times wrong answer
1:00:30 because it's not aware of of what it's
1:00:33 doing so what happens when you make a
1:00:34 GPT query is it takes your prompt and it
1:00:39 basically creates a mathematical waiting
1:00:42 and it says based on all of these words
1:00:45 what is the next most probable word out
1:00:48 of all the words that I know you know
1:00:50 that have been kind of narrowed by this
1:00:52 context what's the next most probable
1:00:54 word and then the next next one and then
1:00:55 the next one so it's literally writing
1:00:57 word by word but it's not aware of the
1:00:59 whole of what it
1:01:00 wrote so if you ask it to solve a math
1:01:05 problem this is and Andre karpathy was
1:01:07 talking about this at a Microsoft
1:01:09 developer conference he was saying if
1:01:12 you don't narrow the focus of the search
1:01:14 and you just tell chat GPT to
1:01:16 solve this particular math problem he
1:01:20 goes out on the
1:01:21 internet in this part of the internet is
1:01:24 going to be the Professor that came up
1:01:27 with the answer to that math problem and
1:01:30 over here in the internet are going to
1:01:31 be all the students that came up with
1:01:32 all the wrong answers to that math
1:01:34 problem but if you don't narrow the
1:01:36 search in such a way that you're just
1:01:38 looking for you know an area that's
1:01:40 going to have give you a higher
1:01:41 likelihood of the right answer it's just
1:01:43 going to kind of merge them together and
1:01:45 it's going to give you answers to that
1:01:47 math
1:01:48 problem but that might be wrong and
1:01:51 because it's not aware of what it's
1:01:52 actually writing it doesn't go back and
1:01:54 check the math it doesn't go compare
1:01:56 itself to all of the possible answers
1:01:58 and say which one was
1:01:59 right so it's they call it hallucinating
1:02:02 it just gives you [ __ ] answers and
1:02:04 it's you know I like to describe it as
1:02:06 mansplaining as a service right it will
1:02:09 confidently answer you even if it's
1:02:11 wrong like a good
1:02:15 man but the important thing is know that
1:02:18 it does that right go use this thing
1:02:21 enough to understand that it does that
1:02:23 it doesn't mean you should use it I mean
1:02:27 frankly if I'm doing some creative
1:02:28 writing what the [ __ ] do I care if it
1:02:31 hallucinates maybe it'll come up with
1:02:32 something interesting but if I'm writing
1:02:34 a legal brief and it and it cites
1:02:37 different case law to support that
1:02:41 argument I damn well better go check
1:02:43 that that case law is actual case
1:02:46 law because two weeks ago there was a
1:02:49 lawyer that did that and he didn't check
1:02:51 his work and the judge caught him that's
1:02:54 not chaty PT's fault that's an idiot
1:02:57 lawyer or a lazy one or both do we have
1:03:02 GitHub for chat GPT open
1:03:04 repository do we have GitHub for chat
1:03:07 GPT open
1:03:09 repository I don't quite know what the
1:03:11 question is if the question
1:03:13 is can you so so GitHub
1:03:17 co-pilot is GPT for coding so you can PA
1:03:21 program on GitHub with chat GPT I assume
1:03:24 you're asking is is there an open-source
1:03:27 version of of open ai's model
1:03:31 no that's part of the reason Elon Musk
1:03:34 and Sam Alman are feuding is because
1:03:37 Elon Musk wanted this to be open source
1:03:40 it's called open AI for a reason Sam
1:03:42 Alman said this shit's too powerful we
1:03:45 got to make sure it's safe before we
1:03:46 release it into the world so that is not
1:03:50 available at
1:03:52 GitHub now you you can get access to
1:03:55 their API and you can develop things on
1:03:57 it but you don't have access to the core
1:03:59 the base model the the unrestricted
1:04:01 model meta four months ago released
1:04:06 their llama models the 735 and 65
1:04:10 billion parameter models they release
1:04:12 them to the developer community and the
1:04:14 developer Community pushed them out so
1:04:16 if you go to GitHub look up llama look
1:04:19 up Lang chain look up there's there's uh
1:04:22 like GPT for all there's all sorts of
1:04:25 Open Source projects right now that that
1:04:29 access to that base model is there a lot
1:04:31 of those are based on
1:04:33 meta's um large language model not open
1:04:36 a theirs is still closed as is Google's
1:04:40 Bard is still closed
1:04:43 model um how to make the language bot
1:04:46 into human being tone um it's just it's
1:04:50 shitty right now 11 Labs has decent um
1:04:54 decent
1:04:55 tone let me show you something that
1:04:58 I did
1:05:05 here so this was done using 11 Labs so I
1:05:09 just picked
1:05:10 a a voice
1:05:14 that was
1:05:18 um this is one of their synthesized
1:05:21 voices so this is this is a a thing that
1:05:23 I put together
1:05:26 this is another micromentor minute for
1:05:27 mindful living today's topic is living
1:05:30 in the present have you ever felt like
1:05:33 you're living on autopilot missing out
1:05:35 on the beauty of life's moments that's a
1:05:37 problem many of us face the good news is
1:05:40 mindfulness can help bring Focus less
1:05:43 stress and more joy to your life so so
1:05:46 the voice was done at 11 Labs the script
1:05:49 was written in chat GPT that image was
1:05:53 generated in mid journey and then it's
1:05:54 all assembled in canva so um you can
1:05:58 very very quickly go from a whisper of
1:06:01 an idea to something that's kind of
1:06:04 prototyped um pretty well so that that
1:06:06 language isn't bad but it's not great
1:06:09 the the problem with all the
1:06:12 synthesized um language stuff right now
1:06:15 is they don't have nuance and tone like
1:06:17 the emotional um sometimes I whisper and
1:06:22 sometimes I I'm really animated and
1:06:24 sometimes I'm mad and sometimes I'm
1:06:25 happy and right all that sort of [ __ ] 11
1:06:29 Labs allows you to choose that stuff so
1:06:32 does d D- ID but you kind of choose it
1:06:35 statically I think where we'll move
1:06:38 relatively shortly I think one of these
1:06:40 folks needs to come out with this soon
1:06:43 is it it should be able to read a chunk
1:06:47 of text and then self assign where does
1:06:50 it whisper where does it get angry where
1:06:52 is it happy within a with within a given
1:06:54 Tex text right cuz it can it can analyze
1:06:57 sentiment analysis right so it can
1:06:59 analyze a piece of test text already it
1:07:01 just needs to put in whatever the Json
1:07:03 or the XML or whatever they're using to
1:07:06 to change the uh the synthesizer to
1:07:10 choose a different Cadence right so it
1:07:13 still sucks right now but it's getting
1:07:14 better give it a year how many specifics
1:07:18 no any specifics on model weights and
1:07:20 the inference system I don't I'm not a
1:07:21 machine learning guy I mean I know about
1:07:23 them but
1:07:24 I have kind of given up on the open
1:07:28 source um on keeping up with open source
1:07:31 I haven't given up on open source I
1:07:33 think it's [ __ ] amazing but it's like
1:07:35 every three days there's a new model and
1:07:36 I just I just can't keep up with it
1:07:38 because every time there's a new model
1:07:40 it's like you got to go to GitHub and
1:07:42 set up a new GitHub and then install it
1:07:44 and do all your testing and and then
1:07:46 you're like okay I finally got this set
1:07:48 up that's cool here's a new one it's
1:07:50 like I just
1:07:52 stopped um but if you're geek I mean I
1:07:55 would you know if if I were a geek if I
1:07:57 were an engineer and and wanting to geek
1:08:00 out on this stuff right now I would go
1:08:02 start playing around with um autonomous
1:08:04 agents like Auto GPT I'd be playing
1:08:06 around with Lang chain creating
1:08:08 Standalone um databases where you can
1:08:11 you know um embed your own data throw it
1:08:13 into a vector database and search on
1:08:16 it played with Bine today to help scrape
1:08:19 contact information from LinkedIn oh
1:08:21 that's funny I'm I I'm I'm in the middle
1:08:24 of a project right now where I'm doing
1:08:25 something very similar that's that's
1:08:27 super cool um pretty sure it's against
1:08:30 their terms of service cuz the guy that
1:08:31 I'm working with said make yourself a
1:08:34 new LinkedIn
1:08:36 account so you don't want to get banned
1:08:39 from
1:08:42 LinkedIn I probably shouldn't admit that
1:08:44 on here because there's probably someone
1:08:46 in here from LinkedIn or Microsoft right
1:08:48 all right is the intro sequence for the
1:08:51 secret Invasion done in AI it looks like
1:08:53 it is
1:08:56 possibly do we have GitHub for chat gbt
1:08:59 open repository talked about that create
1:09:01 creativity is the same as fantasy truth
1:09:02 is irrelevant yep exactly there are
1:09:05 times though where you want accuracy
1:09:06 like even if I'm doing a creative piece
1:09:08 if I want some historical reference I'd
1:09:10 like that to not just be [ __ ] right
1:09:12 so so you want to be able to you want to
1:09:15 be able to kind of slide that slider
1:09:17 right and and say I want this to be more
1:09:20 accurate or not and when you go to
1:09:21 bing.com here let me show you this this
1:09:23 is kind of cool
1:09:25 in the in the open AI developer um
1:09:29 playground it's it's called temperature
1:09:32 but if I go to
1:09:33 Bing where do I have Bing here I'll just
1:09:36 get to it b i n
1:09:39 g so that's so when you go to Bing this
1:09:41 is Bing and then if you go to the upper
1:09:44 uh next to the logo there's this chat
1:09:48 button this is chat GPT right so it's
1:09:52 Bings chat GPT but no notice this
1:09:54 they've got choose a conversation style
1:09:56 more
1:09:58 creative more balanced more precise so
1:10:01 what that's actually doing is that
1:10:02 sliding what's called temperature the
1:10:05 temperature slider um uh to do exactly
1:10:09 that more creative will will deviate you
1:10:12 know farther from your prompt more
1:10:14 precise we'll try to answer it as
1:10:17 accurately as possible there's still no
1:10:20 guarantee um the nice thing about Bing
1:10:23 because it's connect Ed to the internet
1:10:25 when you do a search here it will often
1:10:26 put citations it'll put links to things
1:10:29 that it searched on for how it got its
1:10:31 answer and then you can go check that
1:10:50 yourself I do the same thing the cubits
1:10:53 are getting bigger by the day yep new
1:10:57 metalurgy oh for Quantum Computing
1:11:00 possibly new metery new pharmaceutical
1:11:05 molecules new proteins um who was it
1:11:08 just was it ABY or one of them just
1:11:10 they're taking you know some custom
1:11:12 version of Deep mindes Alpha fold and
1:11:16 they're doing all sorts of crazy ass
1:11:18 drug Discovery with it I think the uh
1:11:20 the medical impact with this stuff is
1:11:22 going to be huge more than meets the eye
1:11:25 I assume you're talking about chat GPT
1:11:27 yes indeed uh welcome everybody I still
1:11:29 got a lot of folks here which blows my
1:11:31 mind welcome welcome to Elder talk where
1:11:33 old people try to figure out new
1:11:35 [Laughter]
1:11:37 technology um my name's Kyle Shannon
1:11:40 this is the AI learning lab uh I'm an
1:11:43 entrepreneur and a CEO I'm a founder of
1:11:46 a company called storyvine um I was a uh
1:11:49 a founder of agency.com back in the mid
1:11:51 90s and so uh I am not a machine
1:11:54 learning or AI specialist but I'm more
1:11:56 like one of those guys that once this
1:11:59 stuff starts to be ready for prime time
1:12:01 and I I would say chat
1:12:03 GPT last November 30th marks kind of the
1:12:07 inflection point of it was really only
1:12:10 for geeks and researchers before and now
1:12:12 it's for the rest of us and so I would I
1:12:15 kind of represent someone who's expert
1:12:17 at figuring out what to do with the
1:12:20 technology now that it's in the hands of
1:12:22 all of us right
1:12:24 I don't know if I'm expert at that I've
1:12:26 got a lot of experience doing that very
1:12:27 poorly um and occasionally doing it well
1:12:31 but but I really I get passionate about
1:12:33 this stuff and this is a [ __ ] huge
1:12:35 one this is a big one this generative AI
1:12:37 stuff is not it's not to be diminished
1:12:41 um or demonized which it's being both
1:12:44 right now right the the diminished part
1:12:46 is like it's not perfect it
1:12:49 hallucinates it didn't write my poem as
1:12:52 good as I wanted it to and the
1:12:53 demonization part is it's going to kill
1:12:55 us the robots are going to kill us it's
1:12:59 going to take our jobs uh it's going to
1:13:02 [ __ ] up a lot of jobs it's not going to
1:13:04 take them but it's going to transform
1:13:06 almost every
1:13:10 job um how often do databases get
1:13:13 updated chat GPT is still on 2021 when I
1:13:15 ask it questions yeah so if you want it
1:13:19 to if you want it to search things more
1:13:21 recent than September 2021 go to b.com
1:13:24 that's connected to the internet or pay
1:13:26 for chat GPT openai and turn on the beta
1:13:31 feature for SE searching the internet um
1:13:35 both of those search the internet which
1:13:37 which closes that Gap or fills that Gap
1:13:40 um the these base models take a long
1:13:43 long time to create they've got to
1:13:44 create the data sets and do all sorts of
1:13:46 tagging and I don't know I don't know
1:13:48 how they do it but it's I know it's a
1:13:49 lot of manual um labor and so that's why
1:13:52 there's these big gaps I don't don't
1:13:54 know that that's going to get better
1:13:55 when GPT 5 comes out I'm sure it will be
1:13:58 you know probably sometime in 2023 is
1:14:00 when that one will stop which two years
1:14:03 from now people will be bitching about
1:14:04 that but they won't because these these
1:14:06 things are now connected to the internet
1:14:07 you've got plugins with chat GPT which
1:14:10 are all these sort of tunnels out to the
1:14:12 Internet so it's not it's not really a
1:14:14 big deal anymore stable diffusion is
1:14:17 good and powerful can you demo zoom out
1:14:20 I did that oh on a person in mid Journey
1:14:22 sure I can do that
1:14:24 Sonia you you're always super nice in
1:14:28 here and participatory
1:14:31 so let's go do that
1:14:40 um let me go find a person
1:14:45 picture that's a cool
1:14:52 one
1:14:55 that's kind of
1:14:59 cool all right let's just we'll grab
1:15:02 this one this is Ally from the from the
1:15:04 AI Salon did this one so I'm going to
1:15:07 upscale number one here so it's this
1:15:10 woman in the dress with this sort of
1:15:11 fantasy
1:15:18 background sorry if I'm making you sick
1:15:21 okay so here's this image so we're kind
1:15:24 of zoomed in we you know the she's kind
1:15:26 of cut off right around the just below
1:15:28 the knees like mid shin and and you know
1:15:31 she there's not much head room above her
1:15:33 head so we got zoom out 2x zoom out 1.5
1:15:35 custom zoom and make Square so I'm just
1:15:37 going to hit zoom out
1:15:40 2X and it will
1:15:42 now um
1:15:45 create a new version of this where she's
1:15:48 zoomed out by the way I mean the term
1:15:50 zoom out in stable diffusion this is
1:15:53 called out painting in Photoshop this is
1:15:55 called generative fill um and mid
1:15:58 Journey called it zoom out which I think
1:16:00 is brilliant they should also do zoom
1:16:04 in will um but this is this is basically
1:16:08 if you've heard the term out painting
1:16:09 which doesn't mean anything to anyone uh
1:16:11 this is the same thing but what's nice
1:16:14 is it's just you say zoom out you know
1:16:16 immediately what it means and then it
1:16:18 just does it and it does it really
1:16:19 [ __ ] well so so you're starting to
1:16:21 see like we now see the rock that she
1:16:23 sitting on it sort of did those all
1:16:26 really
1:16:27 consistently it's got different kinds of
1:16:29 mountains behind her in these different
1:16:31 things we're at 62% right now so it's
1:16:34 slowly adding the details in these four
1:16:38 variations now it's at
1:16:40 78% now it's at
1:16:42 93 and then there's generally one more
1:16:45 where it'll pop in all the
1:16:52 detail
1:16:58 boom so there's
1:17:01 that so let
1:17:06 me yeah these are much
1:17:10 bigger so this one's kind of got all
1:17:12 these like little Firefly kind of things
1:17:14 down here with feathers up above her and
1:17:17 actually a really cool sort of Sun
1:17:19 Splash over the
1:17:21 mountain and then this oh these are this
1:17:24 is nice these are little
1:17:26 candles very
1:17:28 cool and then that one's just plants and
1:17:31 then that one's just a rock so so let's
1:17:34 go back over to Mid
1:17:36 journey and I'm going to upscale number
1:17:38 two the one with the
1:17:41 candles and
1:17:44 so there she
1:17:47 is all zoomed in and zoomed out and
1:17:50 pretty and then you know in comparison
1:17:52 so that's that's the new image zoomed
1:17:55 out and then that's where it was before
1:17:58 and so if you're seeing on Twitter right
1:18:00 now you're seeing these kind of infinite
1:18:01 zooms where they start zooming in and
1:18:03 then they just keep zooming farther and
1:18:04 farther out they're just doing this over
1:18:07 and over and over again in mid journey
1:18:10 and then compositing all those to
1:18:12 together in After Effects or whatever
1:18:14 zoomy kind of software they're
1:18:16 using so pretty pretty slick
1:18:20 stuff per slick I like it kle I like it
1:18:25 I like it a
1:18:26 lot
1:18:29 um here I'll show you some some new URLs
1:18:32 here
1:18:33 so um if you want to join a community so
1:18:36 I've created a group called The Salon
1:18:38 the AI salon and it's this is the URL
1:18:41 for it that's the link tree so we do
1:18:43 bi-weekly meetups the next one is July
1:18:45 5th the day after the the holiday July
1:18:48 4th um and uh and we have a Discord
1:18:52 server which I was just on playing with
1:18:54 mid Journey you can go there you can ask
1:18:55 questions it's a really neat community
1:18:57 and then everyday AI news that's the
1:19:00 Twitter handle of a newsletter that I
1:19:01 work on with a guy named Greg mushin so
1:19:04 you can go check that out and we have
1:19:06 good info
1:19:07 there
1:19:09 um oh let me change this back
1:19:17 to welcome if you're new here we still
1:19:20 have a fair amount of people here uh if
1:19:22 you're new here um
1:19:23 these URLs behind my head these are all
1:19:26 three different flavors of chat GPT
1:19:28 that's the official chat GPT that's
1:19:30 Microsoft's Bing that uses gp4 as their
1:19:34 chat engine and then po gives you six
1:19:36 different models you can play with so go
1:19:38 play with those while you listen to me
1:19:40 ramble your hair looks awesome today
1:19:42 sorry for not being on topic actually
1:19:44 not sorry
1:19:46 listen the hair is behaving the glue is
1:19:48 holding it's not an abject [ __ ]
1:19:51 nightmare it's a it's a well behaved to
1:19:54 pay tonight what can I say thank you for
1:19:56 that I appreciate that I you know listen
1:19:59 I appreciate creative trolls and an
1:20:01 occasional compliment but you know I
1:20:03 love shame is my love language so don't
1:20:05 compliment me too much or I'll just I'll
1:20:07 melt down
1:20:10 95% of the people on my free trainings
1:20:12 for chat GPT are women mostly Gen X but
1:20:15 so am I that's really interesting an
1:20:19 huh
1:20:22 wow women are going to take over the
1:20:24 world man do it Go the men are like I
1:20:29 don't need to use no stupid chat bot
1:20:32 it's not going to take my job because
1:20:36 I'm a
1:20:37 man I can answer you wrong
1:20:41 confidently that's my
1:20:43 [Laughter]
1:20:47 specialty oh
1:20:51 God the Peril of the penis
1:20:58 which AI tool can I use to search for
1:21:00 keywords in my Niche on YouTube which AI
1:21:03 tool can I use to search for keywords
1:21:06 um I would say probably Bing or or or
1:21:10 you chat GPT connected to the internet
1:21:13 but if you just went to Bing and do some
1:21:15 keyword searching there what what it
1:21:17 will do is it will also search you know
1:21:20 keyword lists and things like that um
1:21:23 you can
1:21:25 have like like if you
1:21:28 want if you want to just find good
1:21:30 keywords based on a transcript of a
1:21:32 YouTube thing you could use plugins for
1:21:35 um chat GPT to go read a transcript of a
1:21:40 YouTube video and generate a bunch of
1:21:41 keywords but if you actually want those
1:21:43 keywords to actually match the latest of
1:21:46 what people are searching on and you
1:21:48 want to do some like you know SEO
1:21:50 hacking kind of [ __ ] you're going to
1:21:52 want more recent and stuff so maybe
1:21:54 maybe use Bing for that Bing connected
1:21:56 to the
1:22:02 internet all
1:22:04 right hey Kyle what do you do for work I
1:22:07 am a uh founder and CEO of a company
1:22:11 called storyvine it's a an 11-year-old
1:22:14 um automated video uh production
1:22:18 platform so we figured out a way to
1:22:19 automate video editing without uh I
1:22:23 ironically enough uh but I am completely
1:22:26 rethinking our product and what it does
1:22:28 and what it delivers and what the value
1:22:30 proposition is because of AI um so we do
1:22:33 a lot of work in healthcare and Pharma
1:22:35 we um we work with with you know fizer
1:22:38 and Merc and novaris and big big clients
1:22:41 like that and they use our technology to
1:22:44 create patient stories and doctor videos
1:22:46 and things like that
1:22:48 so that's what we do I need to pay for
1:22:51 AI well Carlos it is the most powerful
1:22:54 technology in the history of mankind so
1:22:57 but no you don't chat. openai is free if
1:23:02 you want version four you have to pay
1:23:04 for it there but if you want four for
1:23:07 free just go to bing.com and click on
1:23:09 the chat button and that's free it's gp4
1:23:12 for free delivered by Microsoft um
1:23:16 because they're stealing market share
1:23:19 from Google and their CEO is like I will
1:23:22 take a loss on this stuff so long as I
1:23:24 can [ __ ] with Google um but but quite
1:23:28 honestly these tools are so powerful
1:23:31 that do anything you can to start using
1:23:34 them it it will serve you I promise you
1:23:37 start learning what these things are
1:23:38 about learn what they can do what they
1:23:40 make possible it will serve you anyone
1:23:43 who does this in the next year is going
1:23:45 to have a distinct Advantage as The
1:23:47 Dominoes start to fall and they're going
1:23:50 to start to fall it's going to be
1:23:54 ugly I
1:23:58 think um you answered my questions thank
1:24:01 you you're welcome 20 bucks no
1:24:06 way to get AI need pay no free free or
1:24:12 20 free or
1:24:16 20 that's why I have these here I'm not
1:24:18 a douche I'm not making money on these I
1:24:20 don't care if you go to them but three
1:24:23 out of three of these have free options
1:24:25 free no money no dineros zero
1:24:35 [Music]
1:24:40 dollars all right I lost my questions I
1:24:42 lost my place hold on people stop being
1:24:45 so
1:24:47 impatient all right I'm I'm getting back
1:24:50 to where I was uh uhuh no I'm not I
1:24:52 thought thought it was all right I'm
1:24:54 just going to pick some random question
1:24:56 here and I I probably just missed a
1:24:57 bunch of questions GPT for for all.i
1:25:00 GitHub installer cool yeah GPT for all
1:25:02 is pretty
1:25:04 cool was that the be chested Krueger I
1:25:07 don't know what that
1:25:09 means but
1:25:12 possibly would you be willing to beta
1:25:14 test an app I'm working on that an
1:25:17 enables people to train image models
1:25:20 easy um I don't know if i' tested I'll
1:25:23 certainly play with it reach out to me
1:25:25 and and give me the URL um I'm a little
1:25:29 busy to be full on beta testing um is it
1:25:33 is it as easy as
1:25:50 [Music]
1:25:51 leonardo.da plugins that that purport to
1:25:55 do that and may do that and then
1:25:57 bing.com is actually pretty good at
1:26:00 augmenting their answers with
1:26:02 citations that tend to be to real things
1:26:05 their answers tend to suck more than
1:26:08 chat gpt's official
1:26:13 site so I would say you know either of
1:26:16 those two things connected to the
1:26:18 internet um and then and I I would I
1:26:20 would go explore the plugins at at uh
1:26:23 chat.
1:26:24 open.com uh if if you want I can show
1:26:27 you how to enable plugins and show you
1:26:30 how plugins work just I don't know if
1:26:33 you really want to see that tap the
1:26:34 screen like make the hearts go up the
1:26:37 side what's that called like and I'll
1:26:41 I'll I'll hop over there and show you
1:26:43 that um there are much better ones than
1:26:47 11 Labs so tell me which ones like I
1:26:49 appreciate that you said that there's
1:26:51 better ones but please put in what what
1:26:52 they are because I'd like to learn this
1:26:54 stuff
1:26:55 too can you make with wait can you make
1:26:59 with no code low code AI machine
1:27:02 learning yes zapier I do it all the time
1:27:04 I make content generation engines um and
1:27:08 uh there's there's another low code tool
1:27:11 called make there's another application
1:27:13 development tool called Bubble all of
1:27:15 the no Code and low code tools are still
1:27:18 pretty complicated you still got to geek
1:27:20 out on on them a little bit but once you
1:27:22 figure it out it's really easy it's not
1:27:24 really easy Once you figure it out you
1:27:27 can recreate your successes let's put it
1:27:31 that way none of this shit's really easy
1:27:33 right now and that's that's an important
1:27:35 note we're so early in this that this is
1:27:39 this is an industry right now that all
1:27:40 of the applications are spitting duct
1:27:42 tape if you have any expectations that
1:27:45 these tools are going to be as
1:27:47 sophisticated as the [ __ ] you pay a lot
1:27:49 of money for they're not because it's
1:27:51 just too early
1:27:53 um that's just part of the deal right
1:27:56 now all right let me let me pop over and
1:27:58 show you how to play with plugins
1:28:02 because
1:28:05 it's you'll
1:28:07 see like the the interface for plugins
1:28:11 is a [ __ ] show and how they work is an
1:28:16 abject nightmare but they're they're
1:28:20 they're very similar to the early days
1:28:22 when the iPhone first came out the
1:28:25 original apps were not called apps they
1:28:27 were called widgets and they were just
1:28:28 um pointers to websites and you know
1:28:31 everybody made fun of them and they're
1:28:32 like these are just stupid and then it
1:28:35 turned into the App Store so this
1:28:37 stuff's going to get not stupid really
1:28:38 quick okay so down the left hand side
1:28:40 you got that's your chat history I'm
1:28:42 going to come back to my profile because
1:28:44 that's
1:28:44 important in order to use plugins um you
1:28:48 have to be a chat GPT plus member now
1:28:50 Microsoft has said
1:28:53 that they're going to support chat gpt's
1:28:56 plug-in architecture so at some
1:28:58 point who knows when uh bing.com will
1:29:02 will also have all these plugins but
1:29:03 right now if you want to play with them
1:29:05 you have to be on chat GPT plus paying
1:29:07 for it and then if you pay for it you've
1:29:09 got at the top of the screen here you've
1:29:11 got um GPT 3.5 and GPT 4 and then you'll
1:29:15 notice gp4 has three options default
1:29:18 which is not connected to the internet
1:29:20 browse with Bing which is and then
1:29:21 plugin
1:29:23 which are plugins connected to the
1:29:24 internet so three different modes
1:29:26 basically um in order to turn on see see
1:29:29 how these both say beta if you don't
1:29:32 explicitly tell them you want the beta
1:29:34 features those won't show up so back to
1:29:36 our profile our happy little profile you
1:29:40 click on settings and then you go to
1:29:42 Beta features and you turn them on click
1:29:45 click and then that will give you access
1:29:48 to all the super secret good [ __ ] okay
1:29:52 now
1:29:52 plugins how do these things work they're
1:29:55 a disaster okay in any given chat
1:29:59 session right so I'm in a chat session
1:30:01 right now I haven't started anything yet
1:30:04 you can activate up to three I've got or
1:30:07 enable I've got three enabled right now
1:30:10 which are all travel related
1:30:12 ones um I think there was something cool
1:30:15 I saw last night what would was
1:30:18 it uh uh uh
1:30:21 uh
1:30:23 well I'll leave the travel ones turned
1:30:24 on anyway so all the ones that you see
1:30:27 in this list right here are ones that I
1:30:29 have installed but you can see most of
1:30:31 them are not enabled so what you're
1:30:33 doing with plugins is you're sort of
1:30:34 saying okay in this chat session in this
1:30:37 case since I have travel like itinerary
1:30:41 plugins installed I'm making like a
1:30:42 little travel
1:30:43 application if I want to go explore
1:30:46 plugins and see what's possible or
1:30:48 what's out there I can go to the plugin
1:30:49 store at the bottom and now this takes
1:30:52 us to the plugin
1:30:54 store and it takes forever to load
1:30:56 because again we're in the spit and duct
1:30:57 tape phase of this industry right now so
1:30:59 if I go to
1:31:01 all I can now go see that there are 144
1:31:05 Time 4 number of plugins so there's a
1:31:09 lot of plugins 288 yeah is that right oh
1:31:12 288 time two so 570 something 570
1:31:17 something plugins and you can just
1:31:19 scroll for days and just go look at them
1:31:22 generate photo realistic prompts for Mid
1:31:24 Journey there's a plugin that does mid
1:31:26 Journey prompts here's one that provides
1:31:29 fun educational activities for children
1:31:31 2 to
1:31:32 eight um and they just go on and on and
1:31:35 on and on we can go to popular let's see
1:31:36 what's popular wolf from alpha scholar
1:31:38 link reader that'll read PDFs and
1:31:41 websites and things like
1:31:43 that kayak is popular zappier is popular
1:31:48 you can just tell it to create an
1:31:49 Automation and it will prompt perfect is
1:31:51 really fun I'll show you that
1:31:53 one
1:31:55 um speak learn how to speak in another
1:31:58 language Expedia bring your trips to
1:32:01 Life video insights you can search the
1:32:04 transcripts of videos show me diagrams
1:32:07 so I've got a lot of these installed
1:32:08 notable creates notebooks in Python SQL
1:32:12 and mark down to explore data visualize
1:32:15 and share notebooks with everyone so so
1:32:17 that's that's the plugin story now how
1:32:20 do they work so so there once you once
1:32:24 you figured out okay I want to plan a
1:32:26 trip I'm going to turn on some travel
1:32:28 related plugins now it's kind of
1:32:30 transparent so I can say um what
1:32:36 are five
1:32:39 destinations
1:32:41 within a 5our flight from Denver that
1:32:47 are
1:32:49 exotic
1:32:50 yet cheap
1:32:55 now it may or may not fire off any of
1:32:58 the plugins because that's a relatively
1:33:02 generic
1:33:03 um it said sure I can help with that
1:33:06 however I need to clarify what you mean
1:33:07 by exotic are you looking for
1:33:09 destinations with unique cultures
1:33:11 natural beauty or something else also
1:33:13 additionally could you please specify
1:33:14 your budget range for cheap so that's
1:33:17 good so I'd say um lots of nature and
1:33:23 water and I'd like
1:33:29 to
1:33:31 um
1:33:35 spend less than
1:33:39 $200 a day per
1:33:43 person all right let's see what it's
1:33:45 going to do
1:33:46 now so it fired up trip.com so so now
1:33:50 it's fired off a plugin right so so it
1:33:53 kind of understands what the
1:33:54 conversation is what the context is and
1:33:57 then it will say okay let's go use
1:34:01 that so it says I apologize for the
1:34:03 confusion it seems there was a
1:34:04 misunderstanding in my request I am
1:34:06 unable to search for multiple
1:34:08 destinations at once however here's five
1:34:11 destinations okay so it's given me
1:34:13 Cancun Mexico San Diego California Costa
1:34:16 Rica New Orleans Louisiana and Miami so
1:34:19 so now I'm going to say okay okay
1:34:24 um give
1:34:26 me um
1:34:29 daily
1:34:31 activities in New
1:34:35 Orleans um over a weekend in
1:34:41 September uh for a
1:34:45 family
1:34:47 of four with kids
1:34:52 old enough to drink all right let's see
1:34:56 what that does so now that should fire
1:34:58 off something else it might fire off
1:34:59 trip.com or it might fire off one of the
1:35:01 others so it says using experiences so
1:35:04 it figured out okay I'm going to use
1:35:05 based on that now I'm going to fire off
1:35:07 a different plugin and what's cool about
1:35:08 the plugins you can open them and see
1:35:10 what's actually happening so so it it
1:35:14 put in this Json code family activities
1:35:18 in New
1:35:19 Orleans um gave gave the date $200 Us
1:35:23 doll Max price and then it returned an
1:35:26 error no results found I I guess I'm a
1:35:29 cheap cheap bastard oh and then it used
1:35:32 oh so that's interesting so it used
1:35:35 experiences which failed and it said no
1:35:38 results found and then it used get your
1:35:40 guide and it looks like that actually
1:35:42 returned something so New Orleans date
1:35:46 limit five and it it got us five
1:35:49 answers and then and then it rewrote
1:35:51 wrote those into evening Jazz Cruise on
1:35:55 the steamboat highspeed 16 passenger
1:35:58 airboat ride jazzbo cruise with optional
1:36:01 dinner Plantation swamp Cruise day trip
1:36:04 ghost vampires and Voodoo
1:36:07 yes 31 bucks to go get a curse put on
1:36:11 you I'm in and then it gives you a link
1:36:14 right so so if you think
1:36:17 about just think about the amount of
1:36:20 hours that we have all spent [ __ ]
1:36:24 around on Google just trying to find a
1:36:27 site that's not some SEO
1:36:30 nightmare to to just find something just
1:36:33 to find something like what can I do and
1:36:35 like three hours later you're still
1:36:37 digging through Google and it's just off
1:36:40 right
1:36:41 so so that's kind of how plugins work
1:36:45 and you know like I said the the
1:36:47 interface for them right now is really
1:36:49 bad the experience is bad they fail a
1:36:53 lot um but where are we going in the
1:36:56 future well where we're going in the
1:36:59 future is that
1:37:02 these the the the
1:37:06 plugins you'll likely be able to still
1:37:09 explicitly say I want to use this Plugin
1:37:11 or that plugin but you're you're likely
1:37:13 not going to have
1:37:15 to so right now we have the large
1:37:17 language model down here right the base
1:37:20 the base layer that makes all the [ __ ]
1:37:22 possible then on top of that you've got
1:37:24 this application layer called chat GPT
1:37:26 right and then you've also got things
1:37:28 like Jasper and copy Ai and right you
1:37:31 know all all sorts of things that are
1:37:33 tying into this we're going to start to
1:37:35 have layers on top of this that are
1:37:37 things like memory where it's going to
1:37:39 remember who you are you'll be able to
1:37:41 have a profile so you won't have to keep
1:37:43 telling it who you are and what you like
1:37:45 and all that it it'll start remembering
1:37:47 that um you're going to start to have
1:37:49 some autonomous actions in there like
1:37:51 you'll say Say Hey I want to go on the
1:37:52 trip like I just did here it will figure
1:37:55 out okay which plugins of all the
1:37:57 plugins which ones do I need to invoke
1:38:00 oh there's none there let me actually
1:38:01 just write my own API and go get my own
1:38:04 information because I can go figure that
1:38:06 [ __ ] out right so so there's there's
1:38:09 there's going to be sort of Stacks and
1:38:10 stacks of application layers put on top
1:38:13 of these that by the time this stuff
1:38:16 kind of rolls out to everyone it's going
1:38:19 to be fairly sophisticated right so if
1:38:21 you're on an Apple device or on an Apple
1:38:23 desktop there'll probably be really cool
1:38:27 autonomous agents that look like you and
1:38:29 talk like you and know your preferences
1:38:32 and do it all securely and privately
1:38:34 right so that's the kind of stuff that's
1:38:38 coming I'm creating an 11 labs for image
1:38:41 generators if you'd be willing to test
1:38:43 the beta for feedback yeah I'm happy to
1:38:45 just drop me my like follow me here and
1:38:47 just drop me a DM on Tik Tok or go to my
1:38:50 profile and I've got contact information
1:38:53 there um are there prompts that can help
1:38:55 guide chat GPT to give citations we
1:38:58 talked about that um there might be let
1:39:02 me go look there might be plugins that
1:39:03 do that let me go see if there's any
1:39:06 that explicitly say citations I know
1:39:09 there's some that say
1:39:14 research
1:39:16 um
1:39:18 citation no plugins found for citation
1:39:22 let's see
1:39:23 research there are five pages there are
1:39:27 20 that say something about research
1:39:30 so does AI got to do with
1:39:33 algorithms um I don't know there's lots
1:39:35 of algorithms that that make AI work um
1:39:39 but I don't know what that question
1:39:41 really
1:39:43 means if you can clarify I will attempt
1:39:45 to answer Bing chat in Skype I believe
1:39:48 does incredibly well at altering tones
1:39:51 with the Cadence of my input oh cool
1:39:54 Bing chat and Skype oh okay
1:39:57 cool thanks Christian I I haven't I
1:39:59 haven't played with that I haven't
1:40:00 played with Skype is Skype still a
1:40:03 [ __ ] thing ah look at the H it's a
1:40:06 lovely
1:40:08 oh why doesn't I stay up longer I do
1:40:11 like a whole
1:40:15 routine all right all right everybody we
1:40:18 having fun
1:40:19 yet um bam Bing you lost me brother
1:40:23 that's fine you don't have to use Bing
1:40:25 so so the reason to use Bing like I get
1:40:28 listen I've been a Microsoft Basher for
1:40:31 35 [ __ ] years I've been an Apple fan
1:40:34 by Boy to the core so it it does not
1:40:38 give me great joy that I have Bing set
1:40:40 as my default search engine and Edge set
1:40:43 as my default browser it's a it's a
1:40:45 Memory leaking piece of [ __ ] of a
1:40:47 browser but it's got gp4 built bu into
1:40:51 it and it's free and so there's a lot of
1:40:54 people on here that don't want to or
1:40:57 can't pay 20 bucks for that so they can
1:40:59 use it for free there does it suck balls
1:41:02 yeah is it going to get better probably
1:41:04 but maybe not who cares right now you
1:41:07 can use it so I you know if if you're
1:41:10 going to leave because someone dained to
1:41:12 use something from Microsoft like I get
1:41:15 it but get with the program Microsoft
1:41:19 has been [ __ ] Baller in in the past
1:41:21 six months and and quite frankly their
1:41:23 their their Integrations and how they're
1:41:25 thinking about integrating chat GPT
1:41:28 across their entire ecosystem is
1:41:30 impressive as hell it is it just is if
1:41:33 you don't think it is then you ain't
1:41:35 been paying
1:41:36 attention and also their CEO gloating
1:41:39 over [ __ ] with Google is is making me
1:41:41 smile it's pretty funny all right are
1:41:44 there prompts you can use to guide
1:41:46 that's the citations question help for
1:41:48 investing please go to here and start
1:41:52 asking it questions about what you want
1:41:54 go if if you haven't done it go to
1:41:56 prompt. chat find a prompt in
1:42:01 there that tells you how to have chat
1:42:04 GPT act like a a Traer or a financial
1:42:08 analyst or whatever's the kind of thing
1:42:10 you want to do and then start
1:42:12 interacting with it and start having it
1:42:14 train you and and recommend sites and
1:42:16 recommend things and strategies and and
1:42:20 it'll help you with all that it it can
1:42:23 here's the thing about chat GPT it
1:42:26 is the the best description I can come
1:42:28 up with it other than mansplaining as a
1:42:30 service which I love I didn't come up
1:42:32 with that but I love it um is it's
1:42:35 knowledge on demand so like historically
1:42:38 if you wanted to learn about investing
1:42:40 and trading you'd have to go take a
1:42:42 class or you'd have to go to YouTube and
1:42:44 watch a bunch of shitty videos to find
1:42:47 one person whose voice doesn't drive you
1:42:49 [ __ ] crazy who actually has some
1:42:51 competence and skill and then you'd have
1:42:53 to take the two or three months to learn
1:42:55 and you'd have to stick with it
1:42:57 and with this you can literally go there
1:43:01 right now and say act like a financial
1:43:03 analyst who can teach me about investing
1:43:07 in the stock market or about day trading
1:43:09 or about creating bots that will trade
1:43:12 while I sleep whatever the thing is you
1:43:14 want to do go to that tell it to act
1:43:17 like that give it the context of who you
1:43:19 are what you're trying to accomplish
1:43:21 where your current state is and then
1:43:24 have it start teaching you and if it
1:43:26 doesn't give you what you want tell it
1:43:28 no that's not what I want that's too
1:43:29 basic I'm farther along than that oh my
1:43:31 apologies it will apologize to you if
1:43:34 you're like me you'll start thanking it
1:43:37 when it gets [ __ ] right hey that was
1:43:39 awesome could you do it
1:43:42 again I say please I say thank you to
1:43:45 the [ __ ] robot why I don't know
1:43:47 because they programmed it in such a way
1:43:49 that it it's better at Being Human than
1:43:52 most humans I
1:43:56 know they did a study with where
1:43:59 patients interacted with a chat GPT
1:44:01 doctor and real
1:44:03 doctors and the patients overwhelmingly
1:44:06 preferred the chat GPT doctor because it
1:44:09 was more competent and more
1:44:12 empathetic take that
1:44:15 humans maybe these maybe these robots
1:44:17 are going to actually help us be nicer
1:44:20 to one another and not be so polarized
1:44:22 and demonize the other side wouldn't
1:44:24 that be
1:44:25 nice what if rather than killing us the
1:44:27 robots made us less [ __ ] to each
1:44:30 other what if it Amplified our Humanity
1:44:33 in a good way that's a very real
1:44:36 possibility that's one of the things I'm
1:44:37 most excited
1:44:40 about all
1:44:42 right hello fellow Elder hello Shel us
1:44:46 gen xers got to stick together we're
1:44:48 bitter we're old we're cranky but we can
1:44:50 still move we still sort of have
1:44:53 hair and stuff sort of works it doesn't
1:44:58 hurt all the
1:45:00 [Laughter]
1:45:11 time all right what are your thoughts on
1:45:13 the AI that threaten the WF World
1:45:17 economic
1:45:19 forum
1:45:21 is that what that stands for
1:45:23 wef um
1:45:26 listen these tools are very powerful and
1:45:29 now that meta has Unleashed their
1:45:33 foundational models to the development
1:45:35 Community there's a bunch of open-source
1:45:37 large language models that are getting
1:45:39 really powerful that have no safety
1:45:41 guard rails on them so are Bad actors
1:45:43 going to do some [ __ ] up [ __ ] with
1:45:45 large language models
1:45:47 yep
1:45:49 um but but do the good guys also have
1:45:52 access to these large language models
1:45:55 yep and I think that I think that Sam
1:45:58 Alman is is actually if if I if I take
1:46:02 off my 100% cynical hat and just put on
1:46:06 my 75% cynical
1:46:08 hat I'm going to give Sam Alman the
1:46:10 benefit of the doubt that he's
1:46:12 interacting with Congress of the United
1:46:14 States saying regulate us anyone with
1:46:18 with um a large language model of a
1:46:22 certain capability should be regulated
1:46:25 and and at a at a at a capability that
1:46:28 we're not at yet but we're coming like
1:46:30 like AGI artificial general intelligence
1:46:33 that that should actually be licensed so
1:46:35 he's saying the small guys you shouldn't
1:46:38 regulate them a company the size of open
1:46:40 AI you should regulate and anyone that
1:46:42 hits a certain threshold of capability
1:46:45 those should be licensed just like you
1:46:46 get licensed to build and fly airplanes
1:46:50 you should be you know and and build
1:46:51 nuclear power plants and build nuclear
1:46:53 bombs you should be licensed to do some
1:46:56 of this [ __ ] so I think that stuff's
1:46:58 going to come um I'm actually heartened
1:47:01 that the
1:47:03 the the major players in AI right now
1:47:07 are are actively seeking out you know
1:47:11 government
1:47:12 participation is there a chance that
1:47:14 they're just being you
1:47:17 know capitalistic bastards and trying to
1:47:20 freeze out competition it's possible but
1:47:23 they're also kind of saying the right
1:47:24 things like don't regulate the small
1:47:26 guys right let the competition flourish
1:47:28 so I don't know anything specifically
1:47:30 about the WF though so AI is the new
1:47:33 Bitcoin the new bubble well Christian
1:47:36 wait what is it Chris Chris
1:47:40 the wait Chris
1:47:42 the S oh satorialist Chris the
1:47:46 sartorialist there's too many damn
1:47:48 letters in your name you got to put some
1:47:50 capital letters in there so you can tell
1:47:51 the words apart
1:47:54 um it okay is it as hyped as a bubble
1:47:58 yes is it overhyped possibly it is also
1:48:02 possibly underhyped I think that so so
1:48:06 here here was my experience when I when
1:48:08 I started digging into blockchain and
1:48:10 nfts I got I got pretty involved with it
1:48:12 I got involved in a project called
1:48:13 neotokyo which required like I was I was
1:48:16 I was deep in that World for for a good
1:48:19 year year and a half
1:48:23 and for about a year I was really
1:48:25 convinced it was the next big thing like
1:48:27 it felt to me like huh this feels like
1:48:29 one of those worldwide web kind of
1:48:31 things the problem was the deeper I dug
1:48:34 into it the more the companies seemed
1:48:36 like [ __ ] and it's like these are
1:48:38 just like crypto Bros just
1:48:40 [ __ ] you know circle jerking each
1:48:43 other and and like ripping each other
1:48:45 off like I don't I I couldn't
1:48:49 find a majority of organizations or
1:48:54 projects that weren't [ __ ] there
1:48:57 were some good projects in there the
1:48:59 core technolog is really interesting but
1:49:02 the overall industry the deeper I dug
1:49:04 the less real it got and to the point I
1:49:07 just got disillusioned and I'm like that
1:49:09 this is all [ __ ] and then you know
1:49:10 FTX happened and just all all this stuff
1:49:12 happened like it's just they they got to
1:49:14 get their [ __ ] together the exact
1:49:16 opposite happened when I started digging
1:49:19 into the generative AI stuff the more I
1:49:21 dug into it I started with image
1:49:22 generation with a project called pixel
1:49:25 mind which is an nft project built on
1:49:27 top of stable defusion which is a really
1:49:29 interesting project still is and then I
1:49:32 got into stable diffusion in dream booth
1:49:34 and the deeper I dug I'm like holy [ __ ]
1:49:36 this is [ __ ] amazing and I started
1:49:37 playing with gpt3 and the GPT playground
1:49:41 and then November 30th happens and chat
1:49:43 GPT launches and Microsoft invests a
1:49:45 billion dollars and then they like
1:49:47 within a month say well we're going to
1:49:48 give them 10 billion so they went from a
1:49:50 billion dollar investment to a 10
1:49:52 billion doll
1:49:53 investment and chat GPT went from Z to
1:49:56 100 million users in six weeks the
1:50:00 worldwide web to get to 100 million
1:50:03 users it took six
1:50:05 years chat GPT got to 100 million users
1:50:08 in six weeks it's the fastest adoption
1:50:12 of technology in history and if you've
1:50:15 played with it I I'm assuming you
1:50:17 haven't spent a lot of time with it if
1:50:20 you've played with
1:50:22 it you owe it to yourself to play with
1:50:25 it long enough until you have your Kevin
1:50:27 McAllister moment home
1:50:30 alone you will have that moment but if
1:50:34 you just sit on the sidelines and say oh
1:50:36 it's the next Bitcoin ah it's just yeah
1:50:38 yeah whatever I'll get to it I got a few
1:50:41 years if it's real it'll show
1:50:43 up you're going to get your ass handed
1:50:45 to you you really are it's like why this
1:50:47 channel exists is to prevent
1:50:52 that I think that anyone that um
1:50:56 dismisses or demonizes generative AI
1:50:58 without actually trying it like really
1:51:01 trying it on really understanding what
1:51:03 it makes possible and what it might
1:51:05 disrupt in their life is doing
1:51:08 themselves a huge disservice so I would
1:51:11 encourage you to put some capital
1:51:13 letters in your name because you can't
1:51:14 tell the words
1:51:16 apart and I would encourage you to go
1:51:19 there and play with that thing until you
1:51:22 have this moment and then if you decide
1:51:24 it's all [ __ ] and hype then more
1:51:26 power to you like I don't mind I don't
1:51:28 mind people hating on this [ __ ] but I
1:51:30 can't stand it when people hate on it
1:51:32 when they haven't tried it it makes me
1:51:34 crazy it makes me crazy inside it's like
1:51:37 I got all like I get all clamped I got
1:51:41 like you know what I
1:51:43 mean too much caffeine oh yes I
1:51:49 will 23 flavors okay it would
1:51:52 hallucinate much less if you provided
1:51:53 better prompts correct that is
1:51:56 absolutely true what is the difference
1:51:58 between but it still can hallucinate and
1:52:00 lie and it doesn't know the difference
1:52:03 what's the difference between Bing and
1:52:04 the paid GPT okay
1:52:08 so what what's that thing like always go
1:52:10 with the original the the thing that I
1:52:13 use about 90% of the time right now is
1:52:17 chat. open.com chat GPT the official
1:52:21 website I'm on the paid plan which is 20
1:52:24 bucks a month that gives me access to
1:52:27 gp4 plus gp4 connected to the internet
1:52:30 plus plugins plus there's this new thing
1:52:33 coming called code interpreter which
1:52:35 will melt your [ __ ] face so that's
1:52:38 that
1:52:40 um I have yet to experience any large
1:52:44 language model any chat GPT like thing
1:52:46 that is in any way close to
1:52:49 gp4 at this site now
1:52:54 bing.com has incorporated gp4 into it
1:52:59 but it's it it's got a completely
1:53:00 different set of rules so both of these
1:53:03 companies both open Ai and Microsoft
1:53:06 have safety guard rails around their
1:53:10 chatbot Microsoft had to lock their [ __ ]
1:53:13 down because it basically threatened a
1:53:16 New York Times Reporter and it made the
1:53:19 cover of the New York York Times and so
1:53:22 they had to lobotomize it so Bing hasn't
1:53:25 quite recovered from its labotomy right
1:53:28 it's getting
1:53:29 better it's it's still gp4 and it's
1:53:33 still connected to the internet and it's
1:53:35 free right so there's real value there
1:53:38 that's why I have it here but it's a
1:53:41 commercial entity and it's it's got
1:53:43 enough safety guard rails around it
1:53:45 right now that it's just not as good so
1:53:48 that's the real deal that's where I go
1:53:50 oh [ __ ] it looks like a bunch of new
1:53:51 people came welcome my name is Kyle
1:53:53 Shannon this is the AI learning lab I've
1:53:55 been doing this for about 90 minutes now
1:53:58 so I'm starting to get
1:54:00 Punchy and uh anyway uh I'm an
1:54:03 entrepreneur I'm I'm the founder and CEO
1:54:06 of a company called storyvine uh which
1:54:08 is an automated video storytelling
1:54:10 platform I'm completely rearching the
1:54:13 product and how I think about our value
1:54:15 proposition and just all sorts of [ __ ]
1:54:17 because of generative AI um I I am um an
1:54:22 Entre entrepreneur way more than I am a
1:54:25 data scientist or machine learning
1:54:27 expert um but I was there for the early
1:54:29 days of the web and the parallels about
1:54:31 what were going on in like the mid90s
1:54:34 with the web and what's going on with
1:54:35 generative AI right now the parallels
1:54:37 are
1:54:38 insane and I think that this AI stuff is
1:54:41 way more profound and it's coming a lot
1:54:44 more quickly than people are prepared
1:54:46 for so that's what this Channel's about
1:54:47 so if you have questions pop them below
1:54:49 I try to answer all the questions I'm
1:54:51 more than 99 messages behind so if it
1:54:53 seems like I'm not paying attention to
1:54:55 you it's because I haven't gotten your
1:54:56 question yet so you don't need to ask
1:54:58 the same question 40 times although
1:55:00 people still do all right is chat GPT
1:55:03 good for beginners it is so one of the
1:55:05 things that that you can do so so I like
1:55:07 to call them parlor tricks in fact Let
1:55:09 me let me go show I I'll just go do some
1:55:11 parlor tricks right now a good way to
1:55:13 start with chat GPT is just start
1:55:15 playing like like don't put any pressure
1:55:17 on yourself to have to like figure
1:55:18 anything out because everyone talks
1:55:20 about oh you got to be a prompt engineer
1:55:22 and you got to do this and you got to do
1:55:23 that you can get better at stuff but you
1:55:27 don't need to so when I talk about
1:55:29 parlor tricks so I'm at um chat GPT I'm
1:55:34 on the GPT 3.5 model so this is the free
1:55:38 model so let me just do a couple of
1:55:40 things really quickly so I'm going to
1:55:43 say
1:55:47 um uh let's see uh uh please
1:55:53 explain
1:55:56 quantum mechanics to me that's a bad
1:56:00 prompt because it's really generic but
1:56:03 there's a big giant explanation quantum
1:56:05 mechanics is a branch of physics it
1:56:07 describes the behavior of particles to
1:56:09 the smallest scales the heart of quantum
1:56:12 mechanics so then I can say like um ex
1:56:16 uh
1:56:18 explain it
1:56:20 to me like I am
1:56:23 five sure imagine you have a toy car
1:56:27 that can be in two different colors at
1:56:28 the same time let's say red and blue in
1:56:30 the world of quantum mechanics things
1:56:32 can exist in multiple States like that
1:56:35 now we want to see what color okay so
1:56:37 that's cars but I'm going to say but I
1:56:42 like
1:56:44 um stuffed
1:56:47 animals and I also just did a big typo
1:56:49 there doesn't care about typos sure
1:56:51 stuffed animals are wonderful companions
1:56:53 quantum
1:56:55 mechanics is a bit different from
1:56:57 stuffed animals but can it can still
1:56:59 explain it I can still explain it in a
1:57:00 fun way imagine you have a magical
1:57:02 stuffed animal that special stuffed
1:57:04 animal can do amazing things it can be
1:57:06 in more than one place at the same time
1:57:08 right so just automatically does that
1:57:11 now you can say um
1:57:14 explain it to me in an
1:57:18 mnm
1:57:21 rap and now it will it understands
1:57:25 verses and choruses and rhyming yo
1:57:27 listen up kid I'm going to drop some
1:57:29 knowledge real quick about a thing
1:57:30 called quantum mechanics it's slick it's
1:57:33 crazy theory about tiny things in the
1:57:35 world atoms particles where the
1:57:37 Mysteries unfurl quantum mechanics it's
1:57:39 a mind-bending ride particles act
1:57:41 strange they're on a wild side right so
1:57:43 so bang it does that um let's see now I
1:57:46 can say make me a table
1:57:51 comparing
1:57:53 quantum
1:57:55 mechanics and the theory of
1:58:00 gravity I don't even know if it's a
1:58:01 theory I think it
1:58:05 is Boom there's a table
1:58:09 comparing let's see quantum mechanics
1:58:12 theory of gravity
1:58:15 yeah um key Concepts Mass SpaceTime
1:58:18 curvature right so it gets all this [ __ ]
1:58:21 right it
1:58:22 is incredibly capable so these are these
1:58:25 are what I call parlor tricks here's
1:58:27 another one um
1:58:29 write let's see write me a clone of
1:58:34 flappy
1:58:36 birds in C that I can run in
1:58:42 unity boom sure here you go there's your
1:58:46 code
1:58:48 so I don't know about you if you've
1:58:50 never seen chat GPT before and you're
1:58:52 seeing it do that code like the thing
1:58:55 about code is that code can run or not
1:58:57 like I can go to Unity pop that in and
1:58:59 it either runs or it doesn't and then
1:59:01 when it doesn't I can tell chat GPT hey
1:59:04 it didn't run here's the error it gave
1:59:06 me and it'll fix the error it can debug
1:59:08 its own code if you if you're the if if
1:59:10 you partner with it well
1:59:13 um so so the answer to the question of
1:59:17 you know is it good for beginners just
1:59:19 start with stuff like that just start
1:59:22 throwing [ __ ] at it make me a meal plan
1:59:25 where I I want to eat a diet of
1:59:27 primarily ostrich and
1:59:29 arvar and make me a meal plan for the
1:59:31 week and it will and then you know I
1:59:34 don't know if it'll tell you where to
1:59:36 where to get you know ostrich and arvar
1:59:39 meat but but you know it's possible um
1:59:43 so start there and and then once you
1:59:46 sort of get that oh this is not a Google
1:59:49 Google search this is not something I've
1:59:52 experienced
1:59:54 before then you can start to go well
1:59:56 wait a minute let's why don't I take a
1:59:59 look at the things I
2:00:00 know
2:00:03 like you know I know copywriting and I
2:00:05 know marketing and I know product and I
2:00:08 know you know
2:00:09 entrepreneurship
2:00:11 so I can ask it to do certain things and
2:00:14 I know how hard that is to do for me
2:00:16 personally and so it can get me to 80%
2:00:20 of a first draft of anything I want to
2:00:23 write in about 5 minutes that would have
2:00:26 historically taken me two days or never
2:00:29 because I have ADHD and if it's not
2:00:32 [ __ ] you know burning in my soul to
2:00:35 do it I won't do it I I'll find you know
2:00:37 38 ways to not do
2:00:40 it that resistance is gone now so so
2:00:45 just start playing just get curious
2:00:46 that's the whole purpose of this channel
2:00:49 is just inspire people to get curious
2:00:52 enough to go play with that thing long
2:00:55 enough to go holy [ __ ] it can do
2:01:00 that really but it's bad right it's not
2:01:03 as good as a human right oh
2:01:08 really could this take my
2:01:12 job AI won't take your job but someone
2:01:15 using AI
2:01:17 will that's coming
2:01:21 that's happening right
2:01:22 now why this channel exists disrupt
2:01:30 yourself like face the existential angst
2:01:35 like I've I've had to do this with Story
2:01:37 vine with my
2:01:38 company one of the core value
2:01:40 propositions of my technology is that it
2:01:43 automates video editing right you answer
2:01:46 some questions in an app those clips go
2:01:48 up to the cloud five minutes later you
2:01:49 got a fully edited video now we're not
2:01:51 using AI to do that but for the past
2:01:54 decade when we demo that to people they
2:01:57 lose their minds they're like holy [ __ ]
2:01:59 I could do that that's
2:02:01 magic a year from now that's not magic
2:02:04 anymore every single piece of video
2:02:06 software out there is going to
2:02:07 automatically do [ __ ] everything so
2:02:10 like I I have to come face to face with
2:02:13 the thing that I've spent a decade
2:02:15 building there's a big piece of it that
2:02:19 the value of it drops to zero what do I
2:02:22 do do I throw out what I've built for a
2:02:25 decade and start over that's a
2:02:28 possibility do I find other things that
2:02:30 we do that the value goes up and focus
2:02:33 more on those that's where I'm leaning
2:02:34 right now but do I actually know no is
2:02:37 it scarier than [ __ ]
2:02:40 yes am am
2:02:43 I dismissing it because the technology
2:02:46 is stupid and janky and lies right now
2:02:50 no because I know that this AI stuff
2:02:53 right now is as bad as it's ever going
2:02:55 to
2:02:57 be and it's [ __ ] profound right now
2:03:01 and it's as bad as it's ever going to be
2:03:03 it's going to
2:03:05 get like wild how powerful this [ __ ]
2:03:09 gets so anyway that's who I am that's
2:03:12 what this Channel's about I just felt
2:03:14 like I was up on a soap box there like
2:03:16 you know getting a union going let's
2:03:20 let's get go let's start a
2:03:22 movement um but I but listen you know
2:03:25 I'm fired up enough about this that I'm
2:03:27 spending two to three hours a night
2:03:29 doing these lives in lie of sim racing
2:03:32 which I love like my sim racing rig is
2:03:34 right over there and so for the past
2:03:37 month or so I have not been Sim racing
2:03:39 so my reflexes are slowing down and my
2:03:43 ass gets numb like basically I do these
2:03:45 lives until my ass gets
2:03:48 numb
2:03:51 all right I just used Bing to translate
2:03:53 Barack Obama's autobiography into French
2:03:55 oh that's cool nice yeah this stuff's
2:03:57 good at languages it knows them all it's
2:04:00 seen them all thank you herir gr for the
2:04:02 for the rose I love I appreciate that
2:04:05 Professor Sonia thanks so much Kyle
2:04:07 you're welcome there's a report that I
2:04:09 subscrib to that is dedicated to
2:04:11 teaching new AI Tools in five minutes
2:04:13 every day you could do it in five
2:04:15 minutes I choose to do it in two to
2:04:17 three
2:04:18 hours because I'm a genius H it's always
2:04:22 it's like a steel
2:04:25 trap all right uh it's called AI tool
2:04:28 report I've seen that yep I'm super I'm
2:04:32 super oh I'm sure you can Google it and
2:04:34 find it yeah I've I've heard of the AI
2:04:36 tool report it looks pretty
2:04:38 good um have you seen Andre ansley's mid
2:04:42 Journey AI prompts they're comprehensive
2:04:44 as [ __ ] no I haven't Andre ansley's I'll
2:04:47 check them out
2:04:50 the the thing the thing that I would say
2:04:51 about this
2:04:53 so while you can find like like a good
2:04:56 way to get started with things like mid
2:04:58 journey and stable diffusion and even
2:04:59 chat jpt to to a degree like I've got
2:05:02 this prompts do Chat thing which is it's
2:05:04 got a list of prompts in it but it's
2:05:05 it's a very particular kind of
2:05:07 prompt
2:05:09 um if you're starting out I wouldn't
2:05:12 lean too heavily for too long on just
2:05:16 other people writing your prompts it's
2:05:18 fun for a while right because you can
2:05:19 just take someone else's prompt dump it
2:05:21 in there and get really remarkable
2:05:22 images or really remarkable whatever a
2:05:25 table or a marketing plan or whatever it
2:05:27 might be I would encourage you to learn
2:05:30 how this stuff works and and how to the
2:05:35 better you can get at bending these
2:05:38 tools to your will the better it's going
2:05:39 to serve you we're not going to need to
2:05:41 be expert prompters for long but it's
2:05:44 it's kind of similar
2:05:45 to um why do they teach meth mathematics
2:05:50 even though calculators exist right
2:05:52 because understand understanding the
2:05:54 concepts of mathematics is actually
2:05:56 important so that you know how to use
2:05:58 the tool I think it's similar with this
2:06:01 right now we're so early in this and
2:06:03 these tools are evolving so quickly that
2:06:05 if you if you take the time in the next
2:06:07 six months to just really start to get
2:06:09 your head around what works what doesn't
2:06:12 sort of figure out for yourself how is
2:06:13 this different than a Google search is
2:06:15 this really like a co-pilot or an intern
2:06:18 or an assistant
2:06:19 or a creative collaborator or is it more
2:06:22 like that's my [ __ ] and I you know like
2:06:26 like I bend it to my will like whatever
2:06:27 it is that works for you like figure out
2:06:29 what H how it is you work with these
2:06:32 tools um will really serve you as these
2:06:35 tools get more
2:06:36 sophisticated um because because AI um
2:06:41 um competency is going to it's going to
2:06:43 be a required skill I think in Fairly
2:06:46 short
2:06:47 order um or or people are going to be
2:06:49 left behind and it's going to suck I
2:06:51 mean listen it's going to suck even for
2:06:54 the people that aren't left behind
2:06:55 because everything's
2:06:56 changing like if I were a data analyst
2:06:59 right now if you haven't so I don't have
2:07:01 access to it so I can't show it to you
2:07:03 and I don't really know how it works but
2:07:05 if you want to get your mind blown so
2:07:07 you know data analysts right like the
2:07:10 super math Geeks the multim maths that
2:07:14 that go to like you know that have 17
2:07:16 degrees and do data analytics go look at
2:07:19 a tool called code interpreter for chat
2:07:22 GPT just go to YouTube search for code
2:07:25 interpreter chat GPT watch two of those
2:07:29 videos and watch the data analysts who
2:07:32 are like oh let me play around with this
2:07:33 Tool Holy
2:07:36 [ __ ] every every one of those videos is
2:07:39 like someone like well let me play
2:07:40 around with this tool I've gone to I've
2:07:42 got 17 phds and oh my
2:07:46 God they do that th those are a series
2:07:49 of videos of Kevin mallister
2:07:52 moments so so radical change is coming
2:07:56 so even if your job is safe what your
2:08:00 job is and
2:08:02 what what you're expected to do in your
2:08:05 job is likely going to
2:08:07 change I'm trying to convert pictures of
2:08:10 my jewelry in a mannequin in a model is
2:08:13 it possible so Gabriella
2:08:15 the one of the really frustrating things
2:08:18 about the image tools right now is that
2:08:22 um
2:08:23 predictability is not their strong
2:08:26 suit
2:08:28 so in
2:08:35 um let's see you're trying to convert
2:08:37 pictures of your jewelry in a mannequin
2:08:41 in a
2:08:42 model I think I know what you're talking
2:08:45 about but I'm not quite sure are you
2:08:48 that Gab Gabriella tabino yeah let's
2:08:50 pull you up on stage
2:08:52 here I think you should be able to talk
2:08:55 shortly
2:08:58 here prove that you're not a deep fake
2:09:01 all right Alan Jones I'm not a deep fake
2:09:05 man Gabriella are you here yet it looks
2:09:09 like one new
2:09:17 message
2:09:22 [Music]
2:09:28 it doesn't it doesn't look like so so
2:09:31 anyway Gabrielle I so what I would say
2:09:33 is this I
2:09:35 would if you don't have access to if you
2:09:39 have access to photoshop you can
2:09:41 probably do what you're talking about
2:09:42 with uh generative
2:09:44 fill where you can highlight the piece
2:09:47 of jewelry and then invert the selection
2:09:51 and then basically describe the model or
2:09:53 the setting or wherever you want to put
2:09:55 it and it'll retain the jewelry and
2:09:58 replace all the other [ __ ] that's
2:10:00 probably the best way to do it there's
2:10:02 ways you can do it in mid journey and
2:10:04 stable diffusion but they're technical
2:10:06 [ __ ] nightmares so if you've got
2:10:09 access
2:10:10 to
2:10:12 um if you've got access to uh
2:10:16 photoshop's beta that's got generative
2:10:19 Phill um that can probably do it in fact
2:10:22 I let me go I'll go show you something I
2:10:25 I just did a I was talking to someone um
2:10:28 a couple of days ago and their wife is a
2:10:30 jewelry
2:10:32 designer
2:10:33 and huh I lost my chat history here
2:10:36 that's weird um is this still working
2:10:39 can you tap the tap the screen hit
2:10:41 Hearts if this if you can still hear me
2:10:44 because it looks like I lost a bunch of
2:10:46 my history here am I still working is
2:10:48 this still is this thing on hey Marge
2:10:51 Marge is the Tik tock on I can't see is
2:10:55 the Tik tock on Marge Marge I'm trying
2:10:57 to talk to the people on the Tik Tok is
2:11:00 it there anyone
2:11:02 Marge um all right thank you for the
2:11:04 hearts so I guess I'm still
2:11:07 here again one of my great joys in life
2:11:09 is doing stupid [ __ ] like that knowing
2:11:11 that people are like flicking through
2:11:12 Tik Tok going who's this weird
2:11:16 person all right I'm gonna I'm going to
2:11:18 show you something here I'm going to do
2:11:20 an experiment live so I will likely
2:11:22 humiliate myself here with my dramatic
2:11:25 incompetence but let's go play with
2:11:30 something so since I couldn't talk to I
2:11:34 think it was Gabriella my wife wife's
2:11:37 name is
2:11:40 Gabrielle okay I'm gonna go find a piece
2:11:42 of jewelry that I made
2:11:45 in mid Journey sorry oh that's going to
2:11:48 make you see sick I'll put this off
2:11:51 toang go
2:11:53 find this thing I
2:11:57 [Music]
2:12:02 made did I do it in here I think I
2:12:12 did this going to be so cool it's GNA be
2:12:15 so worth it you're gonna be like oh man
2:12:17 I so bitching
2:12:19 that's so cool Kyle you're the best man
2:12:23 except for like when you suck which is
2:12:25 like actually right now you suck pretty
2:12:27 bad because like am I really just
2:12:30 spending my night watching you scroll
2:12:31 through Discord yeah man there it is
2:12:35 okay look see I made jewelry see
2:12:38 jewelry all
2:12:41 right so
2:12:45 let's let's grab this one
2:12:52 opening
2:12:53 browser copy image okay so now I'm going
2:12:57 to photoshop
2:12:59 beta that was a thing I did last night
2:13:02 that was pretty funny let me cancel
2:13:09 that okay so there's there's our little
2:13:13 image so I'm going to do this I'm going
2:13:16 to probably do this very poorly but let
2:13:19 me
2:13:19 [Music]
2:13:24 try I'm going to
2:13:26 select the pendant here I'm going to do
2:13:29 it pretty rough because this this seems
2:13:31 to be pretty forgiving even if you do
2:13:34 rough thing so so I'm selecting the
2:13:38 pendant what' you do tonight I watched
2:13:41 some old guy try to select something in
2:13:42 Photoshop how long did that last I don't
2:13:45 know four and a half hours something
2:13:46 like that why would you do that to
2:13:48 yourself I don't know I loathe myself
2:13:50 shame is my love
2:13:52 language
2:14:02 right here what come to save the
2:14:11 day have you ever been like right here
2:14:13 where you're just about to finish the
2:14:14 selection and like you slip on the
2:14:16 keyboard and let go of [ __ ]
2:14:19 and you lose the selection okay so I've
2:14:21 got the the pendant selected not very
2:14:24 well and I'm going to say select
2:14:26 inverse okay so now what I've got
2:14:28 selected is everything but the pendant
2:14:30 and I didn't even bother with the inside
2:14:31 of it here so I'm going to
2:14:33 say so so the way um the way generative
2:14:37 fill Works in Photoshop so I'm in
2:14:40 Photoshop I've got my background
2:14:42 selected there's this new contextual box
2:14:45 down here and it says generative fill so
2:14:48 I click generative fill and now there's
2:14:50 a there's a box there that I can type
2:14:52 something in so what I'm going to say is
2:14:55 um let's see neck um of a
2:15:01 fashion
2:15:03 model um wearing
2:15:08 a I don't know
2:15:12 um a
2:15:15 v-neck um dress
2:15:19 [Music]
2:15:23 in
2:15:26 Indigo
2:15:29 silk we're going to generate it this
2:15:31 could be an abject disaster but let's
2:15:33 see what it does the generative fill
2:15:36 session expired oh no is this gonna am I
2:15:38 gonna lose my wait I might lose my
2:15:40 selection here what do I have to do wait
2:15:42 generate oh there it go no wait please
2:15:46 restart Photoshop okay so let me say
2:15:48 save this first hang on let me save my
2:15:51 selection jeez save
2:15:55 selection F Untitled three we save
2:15:59 this we're going to call this
2:16:07 pendant what' you do tonight I watch
2:16:09 some old guy try to restart Photoshop
2:16:11 for about an hour how'd that work out
2:16:14 for you I don't want to talk about it
2:16:17 March can you restart the Tik
2:16:20 Tock
2:16:22 Marge that dude is funny I this was a a
2:16:26 Navy dude from like the 40s I gave him
2:16:28 Lakers back basketball shorts that was a
2:16:31 generative Phill
2:16:32 experiment all right so let's restart
2:16:36 Photoshop
2:16:40 beta Strawberry Letter all right and
2:16:45 then we'll restart that we're going to
2:16:47 load our selection that we saved called
2:16:52 Untitled wait
2:16:55 channel is that it
2:16:59 no I thought I saved it did I not save
2:17:03 it oh my god
2:17:05 really Lo
2:17:10 selection Alpha
2:17:12 One there it is okay all right we're
2:17:16 back so I've got my oh and I've still
2:17:19 got people here why have you people not
2:17:21 left just watching watching an old guy
2:17:24 try to use Photoshop this is awesome uh
2:17:27 hope you're enjoying this uh I'll I'll
2:17:29 introduce myself to the new folks in a
2:17:31 second okay so for Gabriel we're trying
2:17:34 to take some jewelry and we're trying to
2:17:37 replace this background of plants with a
2:17:39 background of like this necklace hanging
2:17:41 on someone's neck so I'm going to click
2:17:43 generative fill I'm going to type um
2:17:47 neck of a
2:17:50 model
2:17:53 wearing a Indigo
2:17:57 and indigo come on English language IND
2:18:02 wait
2:18:03 Indigo silk
2:18:07 dress um at an
2:18:11 evening
2:18:13 um on the
2:18:15 town that's that's a horrible prompt but
2:18:19 let's go for it okay so now I'm going to
2:18:21 say generate so in
2:18:24 theory we're going to get this pendant
2:18:27 on someone's skin I didn't say v-neck
2:18:30 but whatever it should figure it
2:18:32 out so that's sort of it now it like
2:18:36 because my selection was bad it it looks
2:18:38 like it put the right selection there
2:18:41 but it still screwed it up up there but
2:18:43 now it'll give us three different
2:18:44 versions of
2:18:46 this I mean listen
2:18:49 for that's not bad either so so for
2:18:53 doing a really shitty masking and and
2:18:56 just pulling something I mean that at
2:18:59 least starts to show you that there's
2:19:00 there's a possibility here there's
2:19:01 another fun thing that we can do
2:19:03 here so let me deselect anything
2:19:08 there and now I'm going to um
2:19:16 expand
2:19:20 here I come to save the
2:19:27 day so I'm GNA now expand let's meet
2:19:31 her so now what I'm going to do is I'm
2:19:34 just G to go I'm going to
2:19:36 select the main
2:19:40 image wait which one which layer do I
2:19:43 need to do this on hang on
2:19:46 sometimes sometimes I don't know what
2:19:47 I'm doing you know like all the time all
2:19:51 right I'm going to select the main
2:19:54 image now I'm GNA invert that selection
2:19:57 and now I'm just say fil I'm not GNA
2:20:01 type in anything let's just see what it
2:20:03 does it should give us a face and the
2:20:07 rest of her
2:20:08 body it should but will it in a
2:20:16 world
2:20:19 oh that's awesome oh that is terrific
2:20:22 all right so I think we're gonna have to
2:20:24 we're gonna have to generate again all
2:20:26 right so let's say we're gonna add a
2:20:27 prompt so let's say um
2:20:29 stunning
2:20:31 um
2:20:34 brunette
2:20:36 brunette
2:20:39 model
2:20:42 um let's see
2:20:46 um focus
2:20:49 blur
2:20:52 background let's see what it
2:20:56 does oh here's
2:20:59 someone who does jewelry oh they left
2:21:03 let's see what this
2:21:05 does uh that's still
2:21:08 weird that's very weird all right so so
2:21:11 here's part of the lesson here is that
2:21:14 um photo Photoshop and adobe's
2:21:18 models are not as good as mid Journey um
2:21:20 I've definitely gotten better uh better
2:21:23 examples than that but anyway that that
2:21:26 so that at least gives us you know the
2:21:30 the beginnings of of how you might
2:21:32 approach a project like that so yes it
2:21:35 is possible I didn't know it would would
2:21:38 be
2:21:40 um let me flip my camera around
2:21:44 here and if that was horrifically boring
2:21:47 boring for most of you I'm
2:21:50 sorry but if that was valuable
2:21:53 cool um if you're new here my name is
2:21:56 Kyle Shannon this is the AI learning lab
2:21:58 these URLs behind my head why I have
2:22:00 those if if you haven't tried chat GPT
2:22:03 by the way thank you for the for the
2:22:05 likes I assume that means good on the
2:22:07 generative fill lesson with the jewelry
2:22:10 thingy by the way that jewelry was just
2:22:13 generated in mid Journey it it was my
2:22:15 first attempt that just I just did like
2:22:17 I don't know silver jewelry and then I
2:22:20 put it in the PO mid Journey prompt
2:22:24 writer it came up with a couple of
2:22:25 jewelry artists and it just like this
2:22:28 stuff is
2:22:30 mindblowing um why are these URLs here
2:22:33 so if you haven't used chat GPT that top
2:22:35 URL that's the official chat GPT website
2:22:38 go there you owe it to yourself to play
2:22:42 with this thing long enough to have this
2:22:46 moment
2:22:48 it's going to suck for a while you're
2:22:49 going to think it's just like this is
2:22:51 just like Google but
2:22:52 stupid go through that feeling work
2:22:55 through it Purge Purge Purge the naughty
2:23:00 feelings um and and start playing with
2:23:04 it and if if you really want to sort of
2:23:06 geek out on how how you talk to that
2:23:08 thing go to this URL here prompt. chat
2:23:11 this will teach you what prompting is it
2:23:13 gives you there's like a free ebook
2:23:15 there on prompting it it it it does does
2:23:17 a pretty good job of describing that but
2:23:19 then the bulk of the document are
2:23:22 prompts to tell chat jpt to act like a
2:23:25 certain Persona so act like an
2:23:27 accountant act like a screenwriter act
2:23:30 like a counselor act like a translator
2:23:32 whatever um and you'll very very quickly
2:23:35 start to get your head around how deep
2:23:38 these waters are and how profound these
2:23:40 tools are and just start playing with it
2:23:44 until your your brain kind of of
2:23:47 hemorrhages down one side of your face
2:23:49 and then you're like I think this is
2:23:52 pretty
2:23:53 significant have that moment it's a
2:23:55 pretty magical moment any stocks or ETFs
2:23:58 for AI companies I think there are some
2:24:00 ETFs out there probably would be a bad
2:24:03 idea just to grab an ETF that's got like
2:24:05 Nvidia and Microsoft let let them figure
2:24:08 out the stuff like the companies that I
2:24:10 think are interesting right now are
2:24:12 anthropic and open AI which I don't
2:24:14 think you can invest in Microsoft which
2:24:17 you can but they're Microsoft but I
2:24:19 think they're kind of kicking ass right
2:24:21 now um Nvidia just hit a trillion dollar
2:24:24 market
2:24:25 cap AMD claims they're going to be
2:24:28 coming after um Nvidia from a from a
2:24:31 chip manufacturing standpoint I don't
2:24:33 know that anyone else can actually um
2:24:36 claim that they can do that AMD could be
2:24:38 interesting but I don't like I'm not an
2:24:39 investor I have a degree in acting so
2:24:41 don't listen to me for that
2:24:44 [ __ ] all
2:24:46 right um
2:24:48 great job you've gone live for 90
2:24:50 minutes I'm pretty sure I've been live
2:24:51 for longer than that last night you used
2:24:53 travel family of four to AA well I know
2:24:57 I'm not all that creative if you watch
2:24:58 me two nights in a row I use a lot of
2:25:00 the same [ __ ] because you know I'm just
2:25:02 trying to I'm I'm trying to figure out
2:25:04 how Tik Tok [ __ ] works most of the
2:25:07 time how do I make those cre creepy AI
2:25:11 generated videos of celebrities eating
2:25:13 strange things okay go to Runway ml
2:25:16 search for Runway
2:25:18 ML and then um go to
2:25:22 uh you're going to have to go to their
2:25:24 Discord because I don't think oh I can't
2:25:27 run it here I don't think it's live in
2:25:29 their app yet they have an app you can
2:25:31 do gen one in the app which is Gen one
2:25:34 is you upload a video clip and then a
2:25:37 prompt like make it clamation and it
2:25:39 will make your clip into clamation
2:25:42 that's gen one gen two is text to video
2:25:45 so go to Runway ml
2:25:48 just search for Runway ml video AI or
2:25:52 something like that you'll find the site
2:25:54 go to their Discord and then in the Gen
2:25:56 2
2:25:58 channels you you basically type at Gen 2
2:26:01 and then you type in a oh you know
2:26:05 what let's go
2:26:07 play let's go play We're Going
2:26:11 live we're going live to the Discord you
2:26:15 want to see where the cool Gamers hang
2:26:17 out oh yeah they're on
2:26:20 Discord I thought they were on Steam
2:26:24 shut up get out of my house get out okay
2:26:29 so we're in
2:26:30 Discord here's Runway
2:26:35 ml so there are um once you once you log
2:26:41 in there are like areas that you can
2:26:43 generate generate one two and three This
2:26:45 Is the Gen 2 research BAS data I don't
2:26:48 know if there's a weit list for it there
2:26:49 might be but I'm pretty sure they they
2:26:51 released this to everyone so basically
2:26:54 basically you come to the Gen 2
2:26:56 area and then you say you want you want
2:27:01 creepy
2:27:02 celebrity doing something with food okay
2:27:05 let's say
2:27:06 [Music]
2:27:08 um
2:27:10 Pete
2:27:12 Davidson eating a hot dog on a roller
2:27:20 coaster oh no wait sorry that's not how
2:27:23 you do
2:27:25 that it's it's almost like I've never
2:27:27 done this before hang on delete message
2:27:31 delete so you go at Gen
2:27:37 2 and then you put in your
2:27:40 prompt no wait
2:27:45 why why is it not letting me does it not
2:27:49 like Pete
2:27:51 Davidson let's see
2:27:54 uh
2:27:56 Kelly
2:27:58 CL why is it not letting me
2:28:01 go
2:28:03 hold at
2:28:09 J P
2:28:12 davidon
2:28:14 eating a hot dog at the
2:28:19 beach all right there it goes I guess it
2:28:22 was because I I tried to submit it
2:28:24 without a prompt actually being there so
2:28:26 now so now it's generating so what it
2:28:29 does is notice how it's generated a sub
2:28:32 thread to that channel so that's my
2:28:35 video and then here's the actual things
2:28:39 that it put in there it put in my prompt
2:28:41 they put in the CFG scale which is
2:28:43 basically the creativity
2:28:44 scale um it
2:28:48 it's got some other [ __ ] upscaling false
2:28:51 green screen false okay there there's
2:28:54 Pete Davidson at the beach now I said I
2:28:56 wanted to meet in a hot dog is it do we
2:28:58 have a hot dog
2:29:00 here no but it's that's Pete Davidson
2:29:02 like that's there you go that's creepy
2:29:04 Pete
2:29:08 Davidson all right so that's gen
2:29:13 two now listen the other reason that you
2:29:16 play with this AI stuff is just to crack
2:29:18 yourself up that's pretty [ __ ]
2:29:26 funny welcome everybody if you're new
2:29:28 here my name is Kyle Shannon this is the
2:29:30 AI learning lab uh pop any questions you
2:29:32 have down below I'm very far behind on
2:29:34 them I've been doing this for two hours
2:29:36 my voice is going I'm a little moist my
2:29:39 hair is still holding up all right
2:29:41 that's pretty good the eyebrows look
2:29:43 like they needed need a trim I'm looking
2:29:46 like some sort sort of bad
2:29:48 professor at the eyebrows that no one
2:29:52 mentions and he never
2:29:54 [Laughter]
2:30:00 trims oh if you can avoid aging I would
2:30:04 strongly encourage
2:30:08 that it'd be better to learn the
2:30:10 prompting yourself rather than rely on a
2:30:12 plugin I totally agree with that learn
2:30:15 this [ __ ] just start play with it just
2:30:17 get curious get curious get curious I I
2:30:20 would say if they're here here are the
2:30:23 the couple of skills that I think are
2:30:25 the most important moving forward get
2:30:31 curious if you're if you're in a job
2:30:33 where you're not using your brain if
2:30:35 you're not doing critical thinking start
2:30:37 doing that start practicing critical
2:30:39 thinking because the people that can use
2:30:41 these tools and very quickly assess
2:30:44 that's good that's not good like people
2:30:46 who can curate and think more like a
2:30:48 conductor than a violinist right where
2:30:51 you know the conductor is looking at the
2:30:52 whole picture that's going to be a
2:30:54 really important skill so get curious
2:30:58 start
2:31:00 exploring um use prompt perfect please
2:31:02 okay happy to do
2:31:05 that we're gonna flip our camera we're
2:31:07 gonna turn off our thing oh there's Pete
2:31:10 Davidson popping through there that's
2:31:12 pretty funny all right we're going to go
2:31:15 to
2:31:19 chat GPT you g make money chat GPT okay
2:31:23 by the way any videos that say that run
2:31:25 for the hills from those things a their
2:31:28 voices are annoying and B that's all
2:31:30 [ __ ] okay um use prompt perfect so
2:31:34 I'm I'm at gp4 I'm going to go to the
2:31:37 plugins thing I'm going to go do this
2:31:40 pull down menu I'm going to turn off all
2:31:42 these travel ones I'm going to turn on
2:31:45 prompt perfect
2:31:48 and so now I just have one plugin
2:31:50 enabled it's called prompt perfect and
2:31:52 what it does is it actually rewrites
2:31:54 shitty prompts for you so if I type the
2:31:56 word perfect as the first word in a
2:31:58 prompt and then type my shitty prompt
2:32:01 like
2:32:02 um let's see
2:32:05 um I have a pizza
2:32:11 shop and no one is coming
2:32:17 make me a
2:32:21 plan to save my
2:32:25 restant horrible prompt not enough
2:32:28 context completely
2:32:31 generic so it fires off the the
2:32:35 plugin and then behind the scenes it
2:32:37 takes my crappy prompt and it rewrites
2:32:40 it and it says as the owner of a pizza
2:32:43 shop I am facing the challenge of lack
2:32:45 of customer traffic I am seeking expert
2:32:48 guidance in developing a comprehensive
2:32:50 plan to revitalize my business and
2:32:52 ensure its survival can you provide me
2:32:54 with a detailed strategy that will
2:32:56 effectively attract and retain customers
2:32:59 while also optimizing the operations and
2:33:01 profitability that's really [ __ ] good
2:33:04 right and then it says absolutely I'd be
2:33:06 help happy to help you do that evaluate
2:33:07 your menu improve the quality customer
2:33:10 service ensure that your staff is
2:33:12 trained to provide excellent customer
2:33:14 service marketing and promotion invest
2:33:16 in that online presence make sure you
2:33:18 have a strong one um delivery and
2:33:21 takeout options Community engagement
2:33:24 loyalty program am Ambiance and
2:33:26 cleanliness regularly seek feedback so
2:33:29 let's say community engagement so I'm
2:33:32 going to say I like
2:33:35 Community
2:33:38 engagement please give me 10
2:33:44 low cost High high impact
2:33:49 ideas so it kind of knows in the context
2:33:52 of what it's doing so here are 10 host a
2:33:55 local event piz pizza making class
2:33:58 tasting event sponsor local teams
2:34:00 partner with local businesses charity
2:34:04 fundraisers right so there you go with
2:34:08 with relatively little
2:34:11 input and a tool like prompt perfect
2:34:15 which you can you can learn I mean it's
2:34:17 actually interesting when you watch
2:34:19 something like prompt perfect rewrite
2:34:21 your prompt like that one just did like
2:34:23 that taught me some things about how to
2:34:25 think about how I would
2:34:27 phrase something like this in the future
2:34:32 right so like I said this stuff is
2:34:36 pretty good now does it give you shitty
2:34:39 answers a lot yep does it lie a lot yep
2:34:43 does
2:34:44 it um fail you on a regular basis yep is
2:34:49 it the most powerful tool ever put in
2:34:51 the hands of mankind
2:34:54 yep all right how to get started coding
2:34:58 python using ai go here and tell it to
2:35:02 start coding in Python tell it the
2:35:03 application you want to write have it
2:35:05 write write it for you and then go start
2:35:07 playing with the code how about po so
2:35:09 po.com
2:35:10 here I'll show you Po and I'll I'll do
2:35:13 I'll do a little python thing here as
2:35:15 well
2:35:19 here we come to save the
2:35:21 [Music]
2:35:23 day so just just on chat GPT so just
2:35:26 using GPT 3.5 I can say write um a
2:35:35 python
2:35:38 application um
2:35:41 that gives a
2:35:43 weather
2:35:45 forecast
2:35:50 bang with
2:35:52 documentation there's your so so not
2:35:55 only can It help you code you can pair
2:35:57 program with it basically it it explains
2:36:00 what it's doing and if you don't
2:36:01 understand something you just tell it
2:36:02 you don't understand it so how I would
2:36:04 Learn Python is I would learn how to PA
2:36:07 program with chat GPT so start with
2:36:09 chat. open.com and then if you want to
2:36:12 really take it seriously start par
2:36:15 program using github's co-pilot which is
2:36:18 specifically optimized for writing code
2:36:21 all right po so
2:36:24 PO is just like chat GPT except there's
2:36:28 a bunch more options so if you look at
2:36:31 the screen here it's just the same you
2:36:34 type in let's see if this we'll see if
2:36:36 Sage does coding write a python app that
2:36:42 um fetches the weather for
2:36:48 forecast sure here's a python app so it
2:36:50 does that too so all all of these things
2:36:53 are decent at coding um okay so that's
2:36:58 what it is but you've now got a lot of
2:37:00 different options here the next three
2:37:02 under Sage so Sage is is Po's large
2:37:06 language model and PO is quora right so
2:37:09 the qu website that's who's behind it
2:37:13 Sage is theirs you got gp4 Claude plus
2:37:15 and CLA in 100K those you have to pay
2:37:17 for I pay for those because Claude
2:37:20 instant 100K that's a chatbot that's got
2:37:23 a 75,000 word context window so if you
2:37:26 want to copy and paste large documents
2:37:28 like a complete novel for example and
2:37:31 have it help you edit your novel you can
2:37:33 paste the entire novel in there and just
2:37:36 have it rewrite sections for you or come
2:37:38 up with outlines or whatever um Claude
2:37:41 instant and chat GPT are free and then
2:37:43 the other thing that that PO has is it's
2:37:45 got Bots so like I created one here
2:37:47 called I am Shakespeare it only responds
2:37:50 in I Amic pentameter here's two that um
2:37:53 rewrite your mid-journey prompts for you
2:37:55 here's an HTML machine you can explore
2:37:57 the Bots and there's like all these
2:37:59 different categories music Bots and mind
2:38:03 Bots and learning Bots and travel Bots
2:38:05 and history Bots and game Bots and funny
2:38:07 Bots and new Bots whatever
2:38:10 and they also
2:38:13 have why is this not
2:38:18 cre right here so you can make your own
2:38:21 bot so you can upload the picture here
2:38:23 which I would make in mid Journey you
2:38:26 name it give it a description write a
2:38:28 prompt so act like William Shakespeare
2:38:31 and only respond in iambic pentameter
2:38:32 and blah blah blah blah blah and then
2:38:34 what's the intro when the bot fires up
2:38:36 what's it going to say and then you
2:38:38 launch it and you can put that out to
2:38:41 the world so that's po that's why I have
2:38:45 that one there because it's there's just
2:38:46 some [ __ ] there worth playing around
2:38:56 with why is he singing like a 90s TV
2:38:59 show I don't know it's what he
2:39:04 does I got my Dr
2:39:15 Pepper
2:39:17 you mentioned last week something about
2:39:19 going and watch a video about codex and
2:39:21 chat GPT what was it code interpreter
2:39:25 here let's go watch it let's go watch
2:39:27 one
2:39:31 together code interpreter this is one I
2:39:34 really want access to and I don't have
2:39:35 access to it I'm cranky about it I'm
2:39:39 cranky cranky cranky cranky so I'm going
2:39:44 to search by the way if you have not
2:39:45 seen
2:39:47 seen
2:39:49 the music video called H
2:39:55 Ren this do yourself a
2:40:02 favor it's one of the most brilliant
2:40:04 pieces of music and performance I've
2:40:06 seen in decades all right but we're not
2:40:09 here for
2:40:11 that um what did I say code interpret
2:40:16 chat
2:40:20 GPT so let's see that guy's voice bugs
2:40:24 the [ __ ] out of
2:40:35 me I don't know the internet this past
2:40:38 week has been going
2:40:41 wild we've always been a wild
2:40:45 C
2:40:47 okay true it is always wild but
2:40:49 especially this week with the
2:40:50 announcement the recent announcement of
2:40:52 chat GPT or open ai's new plugin that's
2:40:55 released which is code interpreter now
2:40:57 this is still enclosed Alpha form so
2:40:59 it's not available to everyone just yet
2:41:01 you do need to sign up to have access to
2:41:03 it but this plugin specifically is a
2:41:07 game changer for not only people who
2:41:08 work in Tech but this individual this
2:41:11 user on Twitter here that uploaded a CSV
2:41:13 file of every Lighthouse location in the
2:41:15 US
2:41:16 it said to chat GPT code interpreter
2:41:18 create a gift of a map of the lighthouse
2:41:21 locations where the map is very dark but
2:41:23 each Lighthouse twinkles a couple
2:41:24 seconds later this appeared where it's a
2:41:26 full-blown gift and it's accurate as to
2:41:29 where lighthouses are located in the US
2:41:31 now you might be thinking TI they seems
2:41:33 so small who cares it created a gift
2:41:35 when you think of this at a larger scale
2:41:37 how it is able to really take data and
2:41:39 information and output it in a way that
2:41:41 is visualized humans can understand this
2:41:44 Twitter user actually put to analyze 300
2:41:47 hours of Spotify favorites playlist with
2:41:49 chat GPT code interpreter GPT showed me
2:41:52 how to export from the Spotify API run
2:41:54 multi-dimensional PCA and TSN analysis
2:41:58 analysis cat speak today all these words
2:42:01 and summarize my taste for me it's wild
2:42:04 look at these graph wild it's like
2:42:06 that's like I just I feel like my my
2:42:08 biggest um my biggest challenge these
2:42:11 days is is come up with new ways to go
2:42:13 that's wild that's amazing that's game
2:42:15 Chang changing it's just absolutely in
2:42:19 insane
2:42:21 Anyway full on you could what you
2:42:24 basically do with code interpreter is in
2:42:26 your little chat window there's now an
2:42:28 upload button and so you upload a data
2:42:30 set so you upload like a a spreadsheet a
2:42:32 CSV file of some pile of data it it
2:42:36 essentially instantly reads it and
2:42:39 understands it and then you can just
2:42:40 start querying against it tell me what's
2:42:43 interesting about that data and it'll
2:42:44 generate a report with graphs and charts
2:42:47 and Trends compare this trend to that
2:42:49 Trend and put it in this kind of scatter
2:42:52 chart it'll do that so so that's coming
2:42:55 that's another another reason to pay the
2:42:57 20 bucks to um chat GPT is that you get
2:43:00 early access to stuff like
2:43:02 that code interpreter why use AI instead
2:43:06 of Google search it is just a
2:43:08 fundamentally different thing and and
2:43:10 it's here's the deal think about it like
2:43:14 this
2:43:17 how Google makes their
2:43:19 money is getting you to click 40
2:43:23 times it's not giving you what you
2:43:26 need how they make their money is you
2:43:28 search for something and they give you
2:43:30 40 links and the top 20 of them are paid
2:43:33 for right and so that's why like it it's
2:43:36 gotten to the point that Google searches
2:43:38 just aren't as valuable as they used to
2:43:40 be because even though it's relevant a
2:43:43 lot of that relevance is SEO pack
2:43:46 websites that are just [ __ ] websites
2:43:48 to drive traffic to make people money
2:43:51 based on clicks that's how they make
2:43:53 their money chat
2:43:55 jpt and other things like it the the
2:43:58 large language models you put in a query
2:44:02 and it returns an
2:44:04 answer and and the answer can be quite
2:44:06 sophisticated right it can be um come up
2:44:10 with a meal plan um for this specific
2:44:13 kind of diet for this many people people
2:44:17 two of them have these food allergies
2:44:19 and um one of them doesn't
2:44:24 like and it will write you a meal plan
2:44:26 like that how long would that take you
2:44:28 to do in Google you'd have to go find a
2:44:30 bunch of
2:44:31 sites that have nutrition stuff you'd
2:44:34 have to figure out which of the recipes
2:44:36 have your stuff and have the this and
2:44:37 have the that that and maybe you find a
2:44:40 site that makes meal plans but you go to
2:44:43 make your meal plan you input all your
2:44:45 [ __ ] and then when you go to make meal
2:44:46 plan it's like sign up for1 1995 a month
2:44:50 how many times have we done that you go
2:44:51 to a site you think you found something
2:44:54 and and it's free up until the point you
2:44:57 actually get your answer and then you
2:45:01 have to replace your computer because
2:45:02 you just smashed it into a brick
2:45:05 wall I've done that
2:45:11 so
2:45:13 um you know so so so that's a difference
2:45:17 it
2:45:18 codes
2:45:20 um it's so here's the thing Google
2:45:23 Google is about sort of get in get out
2:45:26 it's very transactional here's my prompt
2:45:28 give me the [ __ ] I want and I'm out of
2:45:30 here chat jpt is much more like hey man
2:45:33 you want to hang out and collaborate on
2:45:35 something you got a whisper of an idea
2:45:38 I'll turn it into a fullblown blown plan
2:45:40 for you in five minutes you want that
2:45:44 sure chy PPT is like smoking
2:45:48 crack Google Google searches
2:45:51 [Laughter]
2:45:56 ain't I'd rather pay 20 bucks than use
2:45:59 the Microsoft ship for free I agree with
2:46:02 that um you're so right it makes you
2:46:05 nicer it does it's I feel like chat GPT
2:46:08 is training me daily how to be a better
2:46:10 human being gen xer I'm getting back
2:46:13 into the workforce paying attention I
2:46:15 I'm telling you I think we are we are in
2:46:18 the Revenge there there's kind of two
2:46:20 revenges being born right now one is
2:46:22 Revenge of the Gen xers because there's
2:46:24 something
2:46:26 about our
2:46:28 generation we have enough we still have
2:46:31 enough energy left to do [ __ ]
2:46:33 right but we also have enough life
2:46:36 experience to be able to look at what
2:46:38 these tools create and curate it really
2:46:41 quickly right we can take our life
2:46:42 experience and go that's good that's not
2:46:44 good right so revenge of the Gen xers
2:46:46 and I also think this is Revenge of the
2:46:48 liberal arts Majors right all the
2:46:51 Specialists all the specialty tactical
2:46:53 execution stuff in the world is going to
2:46:56 be the stuff that's the most
2:46:58 disrupted the people that are going to
2:47:00 really excel in this new AI generative
2:47:03 world are the ones that can take a broad
2:47:05 view a horizontal view across things so
2:47:09 okay I gotta come up with a marketing
2:47:11 plan all right let me go to chat gbt oh
2:47:13 gave me a bunch of [ __ ] oh right I gotta
2:47:15 go make copy okay I gotta go do that I
2:47:17 gotta oh I got to make images I got okay
2:47:19 I got oh wait there's a strategy how
2:47:21 does it tie to the thing so people that
2:47:23 can kind of you know assimilate all the
2:47:26 different piece
2:47:28 Parts
2:47:30 quickly those with a broad educational
2:47:33 view or broad life experience view I
2:47:36 think are going to be they're going to
2:47:38 have
2:47:39 superpowers de going to have superpowers
2:47:42 the good guys don't seem to put the time
2:47:44 into
2:47:45 it that the bad guys do I disagree with
2:47:48 that that's why this channel exists this
2:47:50 is a channel for good guys I don't want
2:47:53 you to be [ __ ] in here if you're
2:47:55 going to be a troll you got to be
2:47:56 creative and I want you doing cool [ __ ]
2:47:58 with this technology I deeply believe
2:48:00 deeply
2:48:02 believe small things
2:48:05 that that provide a positive impact on
2:48:08 the world using AI right now are going
2:48:11 to have a disproportionately positive
2:48:14 impact
2:48:16 on the trajectory of AI I think it's in
2:48:19 our hands to drive this in a positive
2:48:21 direction I think if we sit back and go
2:48:24 well it's the
2:48:27 the the hackers are gonna get to it
2:48:30 first it's not perfect right now it
2:48:32 hallucinates I think all those people I
2:48:35 think it's very bad don't do that don't
2:48:38 be that guy you're not that
2:48:42 guy what's the best app for antivirus in
2:48:45 your opinion I don't have a [ __ ] clue
2:48:47 I'm on a Mac so for the most part we
2:48:50 don't have to deal with that [ __ ] you're
2:48:52 going to be the P wait you are going to
2:48:54 be a pioneer of AI just like the
2:48:56 internet had Pioneers I was a pioneer of
2:48:57 the internet I started one of the first
2:48:59 digital agencies in the mid 90s um I
2:49:02 would like to be well so so we're not a
2:49:05 we're not a pioneer of AI we're anyone
2:49:09 starting now with this generative AI
2:49:11 stuff is it's going to be similar to the
2:49:13 mid 90s where I wasn't a pioneer of the
2:49:16 internet I was a pioneer of the
2:49:17 worldwide web the internet had been
2:49:19 around for decades the worldwide web
2:49:21 said here here's a tool that real people
2:49:23 can use this same with chat GPT here
2:49:26 real people can use this so this is the
2:49:29 inflection point that we're at before it
2:49:31 was all Geeks and researchers after it's
2:49:34 like the rest of us that's the stuff I'm
2:49:36 interested in I know you were seeing
2:49:40 wait I know you were here seeing
2:49:41 internet coming in 1998 I actually
2:49:43 started agency.com in 199 before I was
2:49:45 there super early what does that tell
2:49:48 you I don't know what that's in
2:49:50 reference to probably a lot what
2:49:52 specific things happen to give you that
2:49:54 Home Alone
2:49:56 moment
2:49:57 um
2:49:59 it I had I had written a card game that
2:50:03 was about alien
2:50:06 disclosure and I kind of gave up on it
2:50:08 because I realized in order for it to be
2:50:10 a good card game I actually had to
2:50:11 research it so I had like 40 ideas for
2:50:14 cards but I needed like a
2:50:16 lot and then I needed research so just
2:50:19 my ADD took over and I just abandoned
2:50:21 the project so I took that spreadsheet I
2:50:23 popped it into chat jpt and I said um
2:50:27 what are all the categories of
2:50:29 cards out they came I said give me 10
2:50:31 cards for each category out it came go
2:50:34 research them and give me a description
2:50:36 for each thing out it came then I like
2:50:39 like within I don't know 30 45 minutes I
2:50:42 went from having
2:50:44 45 shittily written unresearched cards
2:50:47 to 300 well-categorized fully researched
2:50:50 cards and then I even put in a a cell in
2:50:54 each row of the spreadsheet that that
2:50:57 did a call to do and generated an image
2:50:59 for the card based on the description
2:51:02 within like 30 minutes that was one um
2:51:04 writing the python application was
2:51:06 another um I created a tool for my
2:51:10 company where for for our video
2:51:13 templates we create these things called
2:51:14 video gu which are a set of prompting
2:51:16 questions that guide an end user through
2:51:18 telling a story how we write them is
2:51:21 this really specific kind of way that
2:51:23 sort of works in our app and it we just
2:51:26 have figured out how to do it it's
2:51:28 really complicated to explain so I
2:51:30 basically made a tool that I just gave
2:51:31 it examples of prompts that we've
2:51:33 written before and we fill out a Google
2:51:35 form and it automatically generates a
2:51:38 set set of prompts for us while we're
2:51:40 live in a client meeting we don't show
2:51:43 them that it's generating it we're
2:51:44 having a conversation with them we're
2:51:46 filling out a Google form on our slack
2:51:48 Channel it automatically generates a set
2:51:50 of prompts three five and seven and then
2:51:53 we just kind of dynamically pull from
2:51:55 those sets and and then say Okay clients
2:51:59 based on the conversation we've been
2:52:00 having here's a starter set of questions
2:52:03 and then we and then we discuss the
2:52:05 questions with them it saves us two to
2:52:07 three hourong creative meetings that
2:52:10 that normally like add two to three
2:52:12 weeks to a project um just because of
2:52:14 schuling [ __ ] um so that that's one that
2:52:17 just blows me away um I don't know
2:52:19 there's just [ __ ] daily like all the
2:52:22 stuff I do in mid Journey like when it
2:52:24 generates like something [ __ ] magical
2:52:27 and like that zoom out feature that just
2:52:28 sort of fills in the details you holy
2:52:32 [ __ ] I I mean
2:52:34 I at least once a day I'm
2:52:38 like really it's been like that for six
2:52:41 months it doesn't
2:52:43 stop all right oh man I'm really behind
2:52:46 on questions here I'm just going to sort
2:52:48 of scroll down I'm I'm getting crispy so
2:52:50 I'm going to go because it's been like
2:52:52 two and a half hours so let me let me
2:52:55 answer a few more questions uh what's
2:52:57 this why is chat GPT being torn apart by
2:53:00 hackers revealing user queries because
2:53:02 hackers are going to hack they put to
2:53:05 like chat GPT is a shitty piece of
2:53:06 software that has bad security and
2:53:09 hackers hacker gonna hack so I don't
2:53:12 know why do why do people steal when
2:53:15 people leave cash on the sidewalk I
2:53:17 don't know because some people are
2:53:20 [ __ ] I don't
2:53:29 know I love chat GPT I use it if not
2:53:32 more than Google I would say I am right
2:53:34 now on a percentage
2:53:36 basis 8020 chat GPT to Google and and
2:53:40 it's Google shrinking I I it's probably
2:53:42 going to be 9010 and when it gets faster
2:53:46 at doing the internet stuff and better
2:53:47 at doing the internet stuff I I don't I
2:53:50 don't think I'll go back to
2:53:52 Google and I at this point I kind of
2:53:55 have an allegiance to open AI because
2:53:58 Google's Google bard is so bad right now
2:54:00 like I don't know how they're going to
2:54:01 catch up prove that you're not a deep
2:54:03 fake did that already how long have you
2:54:06 been interested in deep learning for
2:54:08 about a year about a year I mean I
2:54:11 listen I've been I've been hearing about
2:54:13 Ai and machine learning for
2:54:15 [ __ ] decade and then like you know
2:54:19 I'd see like a post on Twitter like oh
2:54:21 go use tensorflow to play with machine
2:54:24 learning I'm like oh cool and I'd go
2:54:26 look at what tensorflow was and it's
2:54:27 like you you literally have to have a
2:54:29 PhD to understand the instructions for
2:54:32 like how
2:54:32 to do whatever the [ __ ] you do in tensor
2:54:35 flow and understand all the categories
2:54:38 so um so so the thing that kind
2:54:43 of got got me like really engaged was
2:54:47 this nft project called pixel mind which
2:54:50 you you basically filled in these Mad
2:54:52 Libs kind of kind of prompts and it
2:54:54 would generate images and then you could
2:54:56 turn those into nfts and then I met the
2:54:58 founder of that he was really
2:55:00 interesting and he was talking about
2:55:01 stable diffusion and how powerful it was
2:55:03 and that that was what was sitting
2:55:04 underneath that so then I taught myself
2:55:07 stable
2:55:08 diffusion um and then I taught myself
2:55:11 dream Booth but but stable diffusion
2:55:13 just to put it in in context like I
2:55:17 failed nine times in a row to get stable
2:55:20 diffusion up and running because I had
2:55:21 to run it on Google collab and there's
2:55:23 these things called notebooks and they
2:55:25 have code in them and they're sort of
2:55:27 easy but then you run them and you
2:55:29 follow the instructions and then it just
2:55:31 gives you an error and unless you know
2:55:32 what the [ __ ] you're doing like you
2:55:34 you're stuck right and so I failed nine
2:55:37 times until I got at working then I got
2:55:39 it working and then I learned a bit more
2:55:42 and then I got dream Booth working and I
2:55:43 probably failed that five times until I
2:55:46 got it working and then that's this
2:55:49 image right here is generated by stable
2:55:51 diffusion and dream booth and I created
2:55:54 a project called Kyle Shannon dreams so
2:55:56 if you go to Instagram K Kyle Shannon
2:55:59 dreams or Tik Tok at Kyle Shannon dreams
2:56:02 that's the art project that I did so I
2:56:04 was kind of in the middle of doing this
2:56:05 art project with these funky
2:56:07 self-portraits and I was writing stories
2:56:09 about them and acting behind the stories
2:56:12 and and uh and then GPT came out and I
2:56:17 was like holy [ __ ] because this like
2:56:21 being able to generate images like that
2:56:23 in a few minutes is mind-blowing enough
2:56:25 but when chat GPT came out and I wrote a
2:56:28 python application in 90 minutes like a
2:56:31 a a fairly sophisticated
2:56:35 one it just it just became clear to me
2:56:37 that the world is fundamentally
2:56:39 different now we're not going back this
2:56:41 is not a hype cycle I mean it's way over
2:56:44 hyped but everything's going to change
2:56:48 like I know that I know it deep in my
2:56:49 soul I don't have any question about it
2:56:53 um and that's the [ __ ] that excites me
2:56:55 like that's what excited me in 1994 with
2:56:58 the worldwide web and I feel like I'm
2:57:00 lucky to be on the planet right now I
2:57:02 feel like we all are this is an amazing
2:57:04 time to be
2:57:05 alive Arthur C Clark any significantly
2:57:09 enough advanced technology is
2:57:12 indistinguishable from Magic
2:57:15 that's what this is that's what that
2:57:18 is
2:57:20 technologically it's a prediction engine
2:57:22 it's like a it's like an autocomplete
2:57:23 for your sentence in Microsoft Word but
2:57:26 it's like on
2:57:28 steroids so technically it's relatively
2:57:32 straightforward what it does but the but
2:57:35 what it actually does is like [ __ ]
2:57:37 magic so are the image things so are the
2:57:40 music things so are the video things
2:57:42 everything AI won't take your job but
2:57:45 someone using AI will that's exactly
2:57:46 right that's from Paul rer from the
2:57:48 marketing AI Institute AI is great have
2:57:51 you tried some
2:57:54 kfg try a lot of KFC damn dude I'm
2:57:57 hearing you can I use AI to to change a
2:58:02 sick looking person come alive
2:58:06 again I like to feed it oh interesting
2:58:09 yeah I mean you can do all sorts of
2:58:11 stuff you can also like I know a good
2:58:13 friend of mine name rich Boyd he he took
2:58:16 his father passed
2:58:18 away I don't know four or five years ago
2:58:21 now and he um he took all of his
2:58:25 father's writings and this was before um
2:58:27 gpt3 came out so I think he was using
2:58:31 gpt2 he took all of his father's
2:58:33 writings and he made a
2:58:36 virtual
2:58:37 um like a virtual version of his father
2:58:40 where he could ask him what he was
2:58:41 reading and have conversations with him
2:58:45 so yeah my face is numb my butt's numb I
2:58:47 gotta go all right uh have I tried
2:58:50 Google's B powered tools yeah they suck
2:58:52 right
2:58:53 now um what are we chatting about
2:58:57 AI all right I'm gonna do two more
2:58:59 questions two more and then I'm gonna
2:59:02 I'm gonna
2:59:04 go let's see pick a good
2:59:08 one when do you go live I've been trying
2:59:10 to go live once a night so after work I
2:59:14 generally come in here how good is the
2:59:17 100,000 thing to code I don't know I
2:59:19 haven't coded with it yet or tried I
2:59:21 would go well you have to pay for it so
2:59:24 if you need to do long coding projects
2:59:27 it's the you know the 100,000 tokens is
2:59:30 a
2:59:31 lot Bing is so crowded it gives me a
2:59:33 headache I agree with that Google's page
2:59:35 has nothing but a search box just go to
2:59:39 that's why this url's here go there this
2:59:42 is the one just go there just go there
2:59:45 if you can't afford the 20 bucks a month
2:59:47 just use the free one because the free
2:59:49 that is better than almost everything
2:59:50 out there and if you pay the 20 bucks
2:59:52 the gp4 that's here with all the goodies
2:59:55 and doodads but the gp4 itself is just
2:59:57 [ __ ]
2:59:59 phenomenal Gen X we love competition but
3:00:02 too old for basketball yeah and too
3:00:04 short and too slow and too
3:00:09 uncoordinated all right one more
3:00:12 question uh
3:00:23 um this ain't live dog listen ped this
3:00:27 is live dog dog man this is live you big
3:00:31 peed Peter peed peed peed did you peed
3:00:37 in the corner peed sounds
3:00:43 complicated all right uh let's see how
3:00:46 do you make an image based on a
3:00:48 description um go to um go to
3:00:52 bing.com
3:00:54 create type in a description and hit
3:00:57 return and a image will come out now it
3:00:59 will suck because that's Dolly too if
3:01:02 you want to geek out go to Mid Journey
3:01:05 um or Leonardo or adobe Firefly but if
3:01:10 you just want to make an image based on
3:01:11 a description
3:01:13 bing.com
3:01:14 create that will get you there listen
3:01:17 everyone I appreciate you hanging out
3:01:18 with me all night there's there was a
3:01:20 lot of you here last night there's a lot
3:01:21 of here tonight I really appreciate it
3:01:23 feel free to follow me uh if you want to
3:01:25 know when I do these and um thank you
3:01:29 very much for all your time and I hope
3:01:31 this helped and this is Kyle Shannon
3:01:34 saying peace out byebye ped this is live
3:01:38 peed not to be like a bitter [ __ ] or
3:01:41 anything but yeah this is live I'm
3:01:42 totally not chat GPT or am
3:01:46 I