aiLL 8/9/23 - Navigating the AI Revolution with Curiosity and Critical Thinking

Description

Chapters

Transcript

0:25 good evening welcome welcome welcome
0:27 welcome to the AI learning lab how's
0:31 everybody doing
0:33 tonight I just reglued my hair cuz I
0:35 forgot to put glue in it this morning so
0:37 this is newly glued so hopefully it
0:40 survives the live hey Tobias what's
0:42 happening what's shaking shaking shaking
0:45 I'm a little late cuz I was having a uh
0:48 feverish race at Monza doing a little
0:51 Sim racing before I came
0:54 live welcome everybody to the AI
0:57 learning lab we'll wait for people to
0:58 get in if you're new here
1:00 go check out uh chat GPT if you haven't
1:03 done that yet looks like we got some
1:06 regulars in
1:07 here welcome Bob how are
1:11 you all right what day is today
1:14 Wednesday middle of the week did I have
1:17 a good day today had a pretty good day
1:19 yesterday it was like crappy Monday good
1:22 Tuesday and
1:25 today I got a little cranky dealing with
1:29 insurance that was fun insurance is
1:32 swell it's awesome what are you drinking
1:35 I got iced coffee iced
1:40 Java and I've got
1:43 lovely filtered water straight from the
1:45 old fridge
1:47 door
1:49 M ah oh it's actually half at work we've
1:53 got Seltzer on
1:55 tap and this week it's been coconut
1:59 which is not my not my favorite so it's
2:01 half coconut Seltzer and half water so
2:04 it's just it's like a little
2:07 like but we hydrate Silver Fox was
2:10 shaking What's Happening Lauren hello
2:13 hello
2:14 hello what's going
2:16 down all right come on Tik Tock let
2:19 people know we're here we're getting
2:20 ready to go on the internets we're going
2:23 to talk about
2:25 AI if you're new here we talk about AI
2:29 you can pop questions down below ramble
2:31 on I'll talk about these things here in
2:33 a second soon as we get some
2:36 audience
2:38 hello good day good day to you sir good
2:42 day
2:44 indeed let's see here's my logo from my
2:48 company isn't that
2:54 fun all right now we got some folks
2:56 coming in I just had to start acting
2:59 like an idiot and that would bring bring
3:00 the folks flying on in just flicking by
3:03 like the hell's this old guy doing huh
3:07 what is he doing he's an odd man he's
3:10 rather
3:12 odd why is it called the AI learning lab
3:15 why why why do we
3:20 inquire about actual
3:22 intelligence isn't that what's
3:26 expected all right so welcome everybody
3:30 my name is Kyle Shannon this is the AI
3:31 learning lab little bit about my
3:34 background I'm an entrepreneur of many
3:36 many times I am officially unhireable at
3:38 this
3:39 point but they let me on Tik Tok uh I'm
3:43 the co-founder and CEO of a company
3:45 called storyvine which is an automated
3:47 video storytelling platform we've been
3:49 around for 11 years um we figured out
3:52 how to automate video editing without
3:54 using AI huh that's clever and then uh
3:59 and now we're rolling a bunch of AI into
4:02 the platform we're actually leaving
4:04 Sunday for New York to go meet with one
4:06 of our big Pharma clients talk about
4:09 some of that stuff so that's
4:11 exciting and what else um I was an Entre
4:15 entrepreneur back in the early days of
4:17 the worldwide web I was a co-founder of
4:19 a company called agency.com one of the
4:21 early digital agencies and uh this this
4:24 here AI stuff is feeling very
4:27 reminiscent uh to the early days of the
4:30 web it's feeling significantly more
4:32 profound and it's coming a lot faster
4:34 but boy does it feel familiar a lot of
4:38 people poo pooing it a lot of people not
4:42 quite understanding why it's such a big
4:44 deal and you know a relative handful of
4:47 people that are like holy [ __ ] this is
4:49 going to change
4:51 everything that's where we are and so
4:53 that's what this Channel's about talk
4:55 about you know these tools and why you
4:57 should use them and and you know even if
4:59 you don't use them why you should at
5:00 least get curious about them and
5:02 understand how they work and what they
5:04 might possibly do because even though
5:07 these things are super powerful they're
5:09 they're the the worst they're ever going
5:11 to be right now the power of these
5:14 things is going up exponentially um so
5:17 with that let me talk
5:25 about all right okay so feel free to
5:30 screenshot this are all the letters
5:32 there there you go feel free to
5:34 screenshot that so there's seven sites
5:36 here the the uh AI learninglab dtt
5:39 That's our YouTube channel um I'm
5:41 putting the uh recordings of these lives
5:43 there and doing descriptions and [ __ ]
5:46 like that using AI actually uh it's kind
5:48 of an interesting experiment um okay if
5:52 you're new to generative AI if you've
5:53 not played with chat GPT go there that's
5:55 the official chat GPT website chat.
5:58 open.com all the other things are kind
6:00 of variations if you've gone there and
6:02 played with it and you want to start to
6:04 figure out you know what tools are good
6:07 at what things these are all different
6:09 variations of chat GPT or competitors of
6:13 chat GPT so you got Bing which is gp4
6:16 connected to the internet you got Claude
6:18 which is a bunch of X open AI people X
6:22 chat GPT people um with their competitor
6:26 Claude is really interesting cuz you can
6:28 put about 70,000 th words in a prompt
6:31 and it can still process it so you could
6:33 put an entire novel just paste it into a
6:36 prompt and CLA or upload a big PDF or
6:39 five PDFs and go do that then you got
6:42 Bard which is Google's entry into this
6:45 space that has been laughably bad for
6:47 most of the year but all of a sudden in
6:49 the past month got decent then you got
6:51 labs. perplexity doai that's perplexity
6:54 giving us all free access to meta's
6:57 llama 2 models for free that's that's
6:59 what that is then you got Pi which is
7:02 from a company called inflection it's
7:03 super conversational not super useful
7:06 but if you want to talk with an AI and
7:08 kind of chat with it uh you can do that
7:10 with your voice on the app it's really
7:12 interesting and then PO is about um po
7:17 po is you can play with multiple models
7:19 here and they've got a bot maker you can
7:21 play with Bots uh Penny's point of you
7:24 what was the first one again well the
7:26 the first the first one here chat.
7:27 open.com that's the official GPT website
7:31 so for free you get to play with GPT 3.5
7:35 if you pay 20 bucks a month you get
7:37 access to GPT 4 code interpreter plugins
7:41 and a thing called custom instructions
7:44 of all these this one's the best chat
7:47 GPT is still the best there's a lot of
7:49 people getting close but in terms of
7:52 just solid delivering the goods gp4 from
7:56 open AI is is is really really solid
7:59 hello TR Tyrant truth how are
8:04 you all
8:06 right if you have questions about AI or
8:10 comments or thoughts or want me to wax
8:12 poetic about something or want to troll
8:15 me if you're going to troll me at least
8:16 be creative with it you know ideally a
8:19 troll in the AI learning lab seems like
8:22 a compliment on the surface and then
8:25 kind of stings deep down it stays with
8:27 you David gold thank you for the worlds
8:30 I so appreciate that you're awesome
8:33 you've been here multiple times now and
8:35 I very much appreciate your
8:38 participation it means a
8:40 lot all right if you have questions pop
8:43 them in the comments below feel free to
8:45 share the live if we want to get more
8:46 people in here I see people hitting
8:48 heart buttons Dana thank you for the
8:51 ruse ruse I love a ruse
8:54 good wow 100 worlds a world by any other
8:59 name would be is
9:04 round oh I'm an ex
9:07 actor in case that wasn't painfully
9:13 obvious all right so what is the future
9:16 of AI for predicting Financial Stock
9:19 Investing well I mean here's so new
9:24 marks if if you're into
9:26 investing in any way then you know more
9:29 than I do
9:30 the thing about Ai and generative Ai and
9:34 machine learning machine learning and AI
9:36 have been around for decades and machine
9:40 learn learning and AI algorithms have
9:43 been in trading platforms for a long
9:46 long time right so there it's already in
9:48 there now if you want to try to game the
9:51 system and play with it and use AI to
9:54 come up with all sorts of wacky
9:55 strategies or write Bots and things like
9:57 that you can absolutely do that so so if
10:00 you go to your favorite trading platform
10:02 get the code from their Bots you can go
10:04 to chat. open.com literally paste the
10:07 code in there tell it to explain to you
10:10 what that bot does and then maybe have
10:13 it suggest ways to improve it or other
10:16 options and maybe you can combine two
10:17 different strategies into a single bot
10:19 there's all sorts of things you could do
10:21 I don't know much about that world at
10:23 all but um lots and lots of stuff you
10:26 can play with if you're new to that
10:28 world you can go to these things and get
10:31 a full-on education within you know 10
10:33 or 15 minutes you know tell it to
10:36 educate you about the different trading
10:37 Styles and all sorts of stuff like that
10:40 it'll do that and then uh yeah
10:44 and yeah so yes you can play with that
10:46 no I don't know anything about it one
10:48 thing that I will point you
10:50 to the third little hyperlink here
10:53 future
10:58 pedia.com I know there are Financial
11:00 tools within there so they're probably
11:02 AI powered trading platforms or AI
11:05 powered Tools in futurepedia so go there
11:08 and see if you can find anything um but
11:11 I think the future of uh you know AI
11:14 powered agents doing all sorts of things
11:16 including investing is that's it's going
11:19 to be a big big
11:21 thing
11:24 um I just found the tutorials
11:27 wait um I found tutorials on my works
11:31 skillsoft for chat GPT oh that's
11:34 interesting do you work at I'm curious
11:36 if you work at a big company or small
11:38 company um and I'm also curious what's
11:41 the
11:42 uh like what's the attitude of of of
11:45 chat GPT within the organization it's
11:48 what what I'm hearing a lot of is people
11:52 are loving it and using it for personal
11:54 productivity and companies are just now
11:56 starting to scratch the surface of uh
11:59 you know I think we got to learn this
12:01 stuff and and we're starting to get into
12:03 it so pretty pretty interesting what's
12:04 happening right now the hell just
12:06 happened to my
12:10 screen hold hold please hold oh I see
12:13 what I did all right there we
12:16 go okay
12:20 um and by the way Penny have you played
12:23 with chbt yet I'm just curious and I'm
12:26 curious what the what the skillsoft
12:28 thing is what are they actually teaching
12:29 there I assume basic prompting Microsoft
12:32 Edge has AI That's pretty good what do
12:34 they use they use gp4 so so
12:38 bing.com the one second down there um is
12:42 from Microsoft and Bing chat is
12:46 gp4 um Microsoft invested $10 billion in
12:50 open AI so that's the top company there
12:53 and um and they also invested a good
12:59 chunk of money in inflection which is
13:02 pi.com and then when meta released their
13:07 llama 2 models they said Microsoft was
13:09 also some sort of investor there so
13:11 Microsoft has got their hands in a
13:13 number of these but the Microsoft Edge
13:15 AI is gp4 it's pretty good it's
13:18 connected to the internet um and you can
13:21 also I don't know if you know this but
13:22 you can generate images in it if you
13:24 tell it to make you an image of anything
13:26 it'll just go do that it'll make you
13:28 four actually
13:31 the uh the quality of the images and
13:33 Bing isn't great but hey it's free so
13:36 it's pretty good talk about llama okay
13:39 so labs. perplexity
13:42 do.com why I have that there so so the
13:45 Llama 2 models are from meta So
13:49 Meta about four months ago maybe five
13:52 months ago released these uh large
13:56 language models called llama and what
13:58 was significant about it was they
13:59 released what are called the weights so
14:01 it's basically the secret sauce behind
14:04 like how they actually work and how
14:05 they're actually trained open aai hasn't
14:08 done that Claude hasn't done that Google
14:10 hasn't done that Pi hasn't done that no
14:12 one's done that but meta did meta said
14:14 here you go but they said this is just
14:16 for research purposes only and then
14:19 about 3 weeks ago they released llama 2
14:23 and they released three different three
14:24 of four models the the the 7 billion
14:27 parameter the 13 billion parameter the
14:29 70 billion parameter the 40 billion
14:32 parameter didn't pass red teaming
14:35 whatever kind of testing they do before
14:37 release that one failed so they released
14:40 those three 7 13 and
14:43 70 and the big deal about them is with a
14:46 commercial license so the only
14:49 restriction on the commercial license
14:50 was you don't have to pay them any
14:52 royalties or anything you can go install
14:55 the Llama models on your local machine
14:57 if you have a machine that can run them
14:58 you can just install it locally you can
15:00 use uh GPT for all as a project where
15:03 you can install different large language
15:05 models so you can pop it right in there
15:08 um I'm on a silic I'm on an M1 MacBook
15:11 Pro here so I can run it I I really
15:13 haven't cuz I'm fine using these other
15:15 ones for now but because they're
15:18 commercially licensable you can sort of
15:20 spin up your own large language model
15:22 build your own chat GPT or build your
15:25 own specialized fine-tuned model um
15:28 because you have the core large language
15:31 model so the the way these things work
15:33 we've heard of chat GPT sitting
15:35 underneath chat GPT is the large
15:38 language model called gp4 so chat GPT is
15:41 up here it's an application layer and
15:43 it's doing API calls down to the core
15:45 large language model so so that's what a
15:47 lot of these things are doing they're
15:49 they they've got you know a friendly
15:51 interface that you can interact with the
15:52 large language model with the Llama ones
15:55 are just that Bas large language model
15:57 you can do anything you want with them
16:00 and the the the Restriction is you can't
16:02 have more than 700 million customers so
16:05 that's that's basically their big
16:07 competitors is is the the ones that
16:09 they're trying to hobble but no one
16:12 else and I think it's an interesting
16:14 model because if if meta can get a lot
16:17 of people using their free large
16:20 language models and building
16:22 applications on them well then at some
16:24 point they can roll out a really
16:26 powerful large language model and
16:29 there's going to be all this
16:29 infrastructure that's built that might
16:32 be able to leverage that and maybe that
16:33 one they sell or you know maybe it keeps
16:36 them relevant when all this new exciting
16:40 shiny [ __ ] is going on and they kind of
16:42 have egg on their face from changing
16:44 their
16:45 name to essentially a gaming platform
16:48 that no one used so so so you know meta
16:54 is staying in the game um but but it's
16:57 pretty impressive what they did and it's
16:58 pretty impressive that they did it and I
17:00 think one of the things that it does for
17:01 all of us is it ensures that these large
17:04 language models they're not going
17:05 anywhere if if you have if you haven't
17:07 been to the channel before this list the
17:10 the reason I put this list up here these
17:11 are things that I think are
17:13 worth these are the in my humble opinion
17:16 it's worth your time you know spending
17:18 at least some amount of time on these to
17:19 understand what they're good at what
17:20 they're not if you're brand brand new
17:23 just go to chat GPT the top
17:25 one um but all of them are interesting
17:28 three weeks ago there were only three
17:31 three on this list now there's
17:33 seven so the competition is
17:37 accelerating Apple's coming Apple Tim
17:40 Cook said today that they have spent the
17:44 bulk of their $21 billion in research
17:49 funding on generative
17:52 AI the entire market
17:55 cap
17:56 of open AI is about the
17:59 same so so Apple's been
18:03 spending
18:05 billions
18:06 on generative AI they're going to come
18:09 out
18:10 swinging and apparently they're testing
18:13 something called Apple GPT within within
18:16 the ranks um and I think they're going
18:18 to come out swinging they've got their
18:20 iPhone 15 announcement September
18:22 whatever that is
18:24 um if they don't announce there it's
18:27 possible they announce something with I
18:29 there but if not it will probably be the
18:33 worldwide developers conference I think
18:35 they have a minor announcement in
18:36 December but but their worldwide
18:38 developer conference next June if if
18:41 they don't do something big in the next
18:43 couple of quarters um it'll be it'll be
18:46 June of next year and I bet they come
18:48 out
18:51 swinging and we also have yet to hear
18:53 from Elon Musk in the x. a company
18:57 that's that's got some Intrigue you know
19:00 what are they doing his the goal of his
19:02 AI is to understand the universe so
19:06 what's that going to do I don't know but
19:08 hopefully we get
19:11 it um any others coming close to chat
19:14 GPT for Excel VBA coding I don't think
19:18 so but I would try Bard Bard's
19:20 apparently pretty good at
19:24 coding I've heard I haven't I haven't
19:27 played with it I've heard that the llama
19:29 13B model the 13 billion parameter model
19:33 is good at coding but I don't know I
19:35 would assume not if if you look at there
19:38 were some specs that were released about
19:40 chat GPT and it's got like 1.8 trillion
19:45 parameters it's 12 different expertise
19:48 models they call it like it's m Moe
19:53 something of expertise modes of
19:55 expertise that's not it but it's
19:57 something like that
19:59 um but basically when you ask gp4 a
20:02 question it chooses one of 12 more
20:06 focused models and it uses that so it's
20:08 not just one massive model it's this
20:10 really complicated system which is which
20:14 is pretty
20:15 fascinating uh Wes Wa West nardy thank
20:18 you for the Rose thank you for the rose
20:21 I like the rose roses are sweet you know
20:24 you can smell them but then if you get
20:25 too close turn it'll poke you right in
20:27 the news thanks for the
20:30 ruse all right let's see Amelio's wife's
20:34 here welcome welcome welcome is Tik Tok
20:37 letting you know I'm
20:39 around Tik tok's been hiding me from
20:42 you
20:45 uh
20:48 how don't you think they should let
20:50 those hats on a little longer seems a
20:52 little cheap thanks for the Rose thank
20:55 you thank you thank you seems a little
20:57 cheap I could I could go for a hat a
20:59 little longer than what is it a second
21:02 and a
21:03 half um action Transformers action
21:07 Transformers yes action Transformers are
21:10 going to be amazing they're not out yet
21:13 but that's those are the ones that are
21:15 uh taking
21:17 action they basically learn they take
21:19 over your keyboard and mouse and learn
21:21 how to do [ __ ] for
21:23 you I think Paul rer from marketing AI
21:27 Institute did a whole piece on action
21:30 Transformers called world of
21:33 bits I think that's the same thing
21:36 action
21:36 Transformers Transformers for actions
21:39 yeah so if you go to let's look up world
21:42 of
21:50 bits let's see world of
21:53 bits
21:56 Paul yeah so so if you go if you search
21:59 for World of bits Paul it's Paul rer
22:03 it's the marketing AI Institute SL
22:07 blogworld of bits with dashes between it
22:11 um this is something he when did he do
22:14 this this was February 18th yeah this
22:15 was like 6 months ago so what this talks
22:18 about is all the companies and some of
22:21 the technology that's coming that's
22:23 going to learn how to use your machines
22:26 on your behalf so you'll give it agent a
22:29 goal like go book me a trip and I don't
22:32 want it to cost a fortune and I want it
22:34 to be somewhere like Costa Rica but not
22:37 Costa Rica and I want it to be at a good
22:39 time of year and I want to be able to
22:41 shoot BB guns at
22:44 balloons and it will go off and search
22:47 for all that [ __ ] and it'll book it on
22:50 your behalf and it'll come back and go
22:51 here you go and then you'll be like oh
22:54 that's cool no you did it completely
22:56 wrong I hate it all go start over
23:02 Okay so so you know we're not too far
23:05 from a world where we all get our own
23:08 you know little assistant that's going
23:09 to go do all this crap for us but we got
23:13 to we got to earn the right to go there
23:15 and how do you earn the right to go
23:16 there you get curious about this stuff
23:18 start playing with this stuff start
23:20 learning it go
23:22 to there's this thing called prompt.
23:25 chat if you're brand new to chat GPT and
23:27 you're just like you you've got kind of
23:29 like prompting block where you're like I
23:30 don't know what to ask it I don't know
23:32 what to do go to prompt. chat it'll
23:34 teach you how to how to uh prompt it and
23:38 it will teach you uh ways that you can
23:41 tell chat GPT to behave like a certain
23:44 Persona and uh it's pretty slick it's
23:47 pretty slick so that'll help uh and uh
23:50 yeah
23:51 so hang on I my I had a watch on and it
23:55 was itchy I just had to take it off all
23:57 right my top three AI well good
24:03 question
24:05 the the top
24:09 two the top two of these things that I
24:11 use cat GPT GPT 4 here and code
24:16 interpreter here are by far the best
24:20 code interpreter is absolutely insane
24:22 but just for basic writing getting [ __ ]
24:24 done gp4 at open AI is amazing um Claude
24:29 is my number two the writing it does is
24:33 pretty good um it's not gp4 level but
24:37 it's got a 100,000 token context window
24:39 so you can put in like 75,000
24:42 words the other one that I use pretty
24:44 consistently is mid journey to create
24:46 images so these are kind of my my two
24:49 go-to word things and then mid journey
24:52 is my image thing I'll use other things
24:55 occasionally but but mid journey is the
24:57 the one that's just sort of topnotch um
25:00 stable diffusion is getting close and
25:02 stable diffusion is way way more there's
25:05 way more features and there's way more
25:08 um projects that use stable diffusion
25:12 which is great because the developer
25:14 Community is just pouncing all over it
25:16 because it's open source the problem
25:18 with it is you have to be a geek to use
25:20 it right now so what I recommend if if
25:24 you're not a geek and want to play with
25:27 stable diffusion
25:28 um
25:57 leonardo.da
25:58 the good thing about Adobe Firefly is
26:01 that it's not trained on anything
26:02 commercial like any copywritten work so
26:05 you can use it for for corporate stuff
26:08 uh without having to worry too much
26:10 about it so those are those are my top
26:12 three chat BT Claude mid journey and
26:16 then I the the bonus one I'll throw in
26:18 is zapier for automations so using
26:21 zapier to automate you you know calling
26:24 open AI um is is very very powerful so
26:28 that would that would be a fourth one
26:29 I'd throw in there and those those are
26:32 essentially um chat GPT I pay for Mid
26:35 journey I pay
26:37 for I pay for zappier I pay for Google
26:41 collab if I want to play around with
26:42 stable
26:43 diffusion I think those are the only
26:45 ones I'm paying for right
26:48 now it's not true there's some things
26:50 that I subscribe to that I just have
26:53 been too lazy or add like to unsubscribe
26:56 from but I'm not act I I'm actively
27:00 paying for the ones that I talked
27:04 about all right should my clients be
27:06 concerned about using mid Journey
27:08 created images in their company's
27:16 literature there there is some risk
27:19 there but I think it's relatively low if
27:22 they here's what I would say
27:25 if if you're using artists names in your
27:31 prompt that's definitely riskier than if
27:34 you just you know tell it to make a
27:36 style that looks like what you want it
27:37 to look like so so I would say avoid
27:41 using Greg Greg root Root Root Kowski
27:44 right avoid using anyone whose
27:46 copyrights are still live their names in
27:49 The Prompt so you know make it look like
27:52 this artist I would avoid that now if
27:55 it's artists that have been dead for a
27:56 long time then that's fine but anyone
27:58 who's got an active copyright in in the
28:01 thing is an issue
28:05 um I think it's going to be almost
28:08 impossible for individual artists to
28:11 look at a piece of art and say what
28:13 percentage of my works you know was
28:17 synthesized in this kind of random noise
28:20 way that these tools work into that
28:23 image um but that said you know some
28:26 Estates like like Keith herring's estate
28:29 is very very aggressive and they they
28:31 sue the [ __ ] out of people all the time
28:33 just because the artwork looks like
28:34 something Keith Herring did um so in
28:37 America anyone can sue anyone for
28:39 anything so while the risk is there I
28:42 would say so long as you're not using
28:44 other artists names and not doing things
28:46 that are blatantly copying kind of known
28:49 copywritten works then you're probably
28:51 fine um I think what's more likely to
28:55 happen is that
28:58 companies like open AI that have
29:02 relationships with companies like
29:04 Microsoft they've got big ass lawyer
29:08 teams and big ass budgets to deal with
29:10 cutting deals with like they just cut a
29:13 deal with shutter stock open AI just cut
29:14 a deal with Shutterstock so I think big
29:17 companies like open Ai and Microsoft
29:19 they're going to cut deals right so
29:22 while they may be sued and they are
29:23 being sued um they're likely going to
29:26 settle a lot of those lawsuits by
29:28 cutting
29:29 deals mid Journey are a bunch of cowboys
29:33 right if you if you see interviews with
29:34 the CEO of mid Journey uh John Oliver
29:37 did a thing on AI and they they
29:39 interview the the CEO of mid journey and
29:42 he just comes across as a douche he just
29:44 comes across like yeah [ __ ] it you know
29:46 yeah it's people's copywritten work but
29:48 whatever like he's he's totally flipp
29:51 get them sued into Oblivion right so I
29:54 think the bigger risk with mid journey
29:56 is probably um just they get caught up
30:00 in lawsuits and they they kind of go
30:02 away and it's not it's it's not
30:04 available so so I would learn other
30:05 tools um if you're really concerned
30:08 about that start learning the Adobe
30:10 Firefly stuff and the generative fill
30:12 within Photoshop because while it's not
30:15 anywhere close to what mid journey is
30:17 right now it's going to get better and
30:19 better and Adobe is making it commercial
30:23 safe right they're making it so that
30:25 artists can opt out of being trained
30:27 they're making it so that artists can
30:28 opt into being trained and get
30:30 attribution and compensation and like
30:33 adobe's got their [ __ ] together so and
30:35 by the way um he's got to get it uh
30:38 approved by his PR person but I've got
30:41 John Knack who's one of the senior guys
30:43 at Adobe in their AI group is coming to
30:46 speak to the AI salon next Tuesday um so
30:50 if you want to join a community of I
30:52 don't know we call ourselves AI
30:54 adventurers right people that are
30:55 choosing to be on this journey and sort
30:57 of figure this [ __ ] out um that's what
30:59 the AI salon's about we meet every other
31:01 week and we have an online Discord
31:03 server you can play and ask questions uh
31:05 we have a LinkedIn group so just go to
31:07 the salon. and and uh you can see the
31:10 info there it's free I'm not it's not a
31:13 it's not a pitch if I'm ever pitching
31:15 you [ __ ] I'll say all right get out the
31:18 [ __ ] wallets let's
31:22 go but none of this is that right now
31:25 all right Skype has Bing chat now yep AI
31:29 is racist well AI is biased um different
31:34 models are going to have different
31:35 biases and um all of them have all sorts
31:39 of safety guard rails in them the safety
31:42 guard rails for things like open AI are
31:45 using human reinforcement learning where
31:48 wherever you've got humans going thumbs
31:49 up thumbs down there's bias in that if
31:52 you're if you're curious about how we
31:55 get to models that are less biased go
31:57 check out out
31:58 Claude because that's a company called
32:01 anthropic and they use What's called the
32:03 Constitutional learning model where they
32:05 they set a set of imperatives like Do no
32:07 harm increase Prosperity you know don't
32:10 be evil whatever those are and then when
32:13 it generates its answers those answers
32:15 are weighted against the Constitution so
32:19 it's all it's all the system it's not
32:22 people in there judging whether this is
32:24 good or not so you can go look at those
32:26 rules and then the content that's
32:28 generated is is essentially filtered
32:30 through those rules um that model in my
32:33 opinion is how we navigate through this
32:37 Jungle of bias thank you Lauren thank
32:39 you for the little gummy bear a little
32:41 gummy I live in Denver so the Gummy's
32:43 here have have a different
32:48 effect um here's the thing about bias is
32:53 that there is no unbiased model every
32:56 model is going to have some amount of
32:58 bias in it and quite frankly all of
33:00 these things are trained on most of the
33:03 public
33:04 internet well what's most of the public
33:06 Internet it's a biased Cesspool of [ __ ]
33:10 right with some occasional Brilliance
33:12 sprinkled
33:13 in so what these things really are is a
33:16 reflection of who we already are so so
33:21 there's that and and you know listen to
33:24 be clear what's on the public internet
33:27 is
33:28 in and of itself biased because who's
33:30 got more access and who's had access for
33:33 longer and things like that so I so I
33:35 get all that but where I think we'll go
33:38 in the future is you take a model like
33:40 Claude anthropics constitutional
33:42 learning
33:43 model you make those models transparent
33:46 and then we as individuals will be able
33:48 to say I get to choose my own model and
33:50 I get to figure out with these little
33:52 sliders where my comfort is in terms of
33:56 bias right and so I might slide it away
33:58 from White and you know toward you know
34:02 liberal or conservative I might slide it
34:04 here and there and it'll recommend
34:07 models that are you know more like me so
34:13 yes AI is biased and yes all of these
34:16 folks are aware of it and because you've
34:20 got things like L 2 which is now open
34:23 source and out there it's not really
34:25 open source it's it's freely available
34:27 in commercially licensed but you've got
34:29 all the weights go fine-tune the [ __ ]
34:31 out of llama to be whatever you want it
34:34 to be make it a non-white biased thing
34:37 you can go do that now when it was stuck
34:39 behind the firewall at open AI or at
34:42 Google we couldn't do that right we're
34:44 just relying on Sam Alman to be a good
34:47 guy and go yeah I know that's biased and
34:49 I'm gonna fix that okay well now you got
34:52 llama too go make your
34:54 own I'm certified in Excel and program
34:58 in VBA GPT is fantastic for VBA yep it's
35:02 good these so so what's fascinating is
35:06 these things are called large language
35:07 models they're they are so are humans
35:09 right we're large language models so
35:12 they kind of behave like we do in in
35:14 some sense um one of the fascinating
35:18 things I heard about open AI training is
35:20 when they did gpt3 they trained it up on
35:23 all the all the sort of words of the
35:25 internet that they could find all these
35:26 data sources right right and it was
35:29 pretty good it was pretty good and then
35:31 they said hey you know what would also
35:33 be good since it it understands you know
35:35 language and and you know semantic
35:37 grouping so well is it it might
35:39 understand Computer Logic so why don't
35:41 we train one up on computer program why
35:44 don't we fine-tune this thing on
35:46 programming computer languages and it
35:48 got really good at that and then I guess
35:49 they kind of rolled these things
35:50 together maybe they fine-tuned it on the
35:52 core model but they so they trained it
35:55 on language and then they trained it on
35:58 computer languages programming and what
36:01 ended up happening was the quality of
36:05 the the the language part went up
36:07 dramatically once once they once they
36:10 got it to understand programming and it
36:12 kind of makes sense because you know
36:14 within programming you've got structure
36:16 and you've got you've got syntax and
36:18 you've got you know logical connections
36:20 well that's how language works right
36:22 it's just not as explicit programming
36:24 you've got all of these brackets and
36:26 things that tell you how to group things
36:28 and how to think about things but the
36:30 English language works exactly the same
36:32 right it's it's it's similar to how we
36:33 think so the the quality of the model
36:37 went up dramatically when they added um
36:39 programming inside it so that's why it's
36:41 so good at at programming and gp4 is
36:45 significantly better than any of the
36:47 rest of these I hear that Bard's pretty
36:49 good at
36:50 programming um and I hear that um llama
36:54 2 the 13 billion parameter models good
36:57 at programming uh for llama 2 but I I to
37:00 be clear I haven't really played with
37:02 llama 2 all that much so I don't know my
37:04 company has an internal chat GPT
37:06 installation large corporation that's
37:08 super cool yeah I'm doing a I'm doing my
37:10 third workshop with a big company this
37:13 Friday it's our third and last workshop
37:15 for chat GPT and they um they're blocked
37:19 from using chat GPT but they have an
37:21 instance of
37:23 Microsoft
37:24 um like the equivalent of of GPT
37:28 playground so they've all got access to
37:31 like GPT playground and then they can
37:33 use Bing so so they're using Bing chat
37:36 which is like chat GPT and then they're
37:38 using the the Microsoft GPT 4 playground
37:41 uh within this big company um yeah and
37:45 and you
37:46 know I think you know as individuals and
37:50 as companies all of us right now are
37:52 figuring out what these things make
37:54 possible how you actually incorporate it
37:56 into work like this past week for me has
37:58 been really frustrating because I've
38:00 been taking things from either prototype
38:04 builds to to kind of a final product a
38:06 delivered product or I'm trying to solve
38:10 some some more complicated thing than
38:12 I've done
38:13 historically and getting the prompts
38:16 dialed in is can be a bear man it can be
38:19 a it can be a nightmare so this week's
38:21 been a little bit
38:23 like I want you to be perfect you stupid
38:26 tools oh and then I just keep reminding
38:29 myself it's early it's early we're all
38:31 trying to figure this [ __ ] out it's
38:33 going to get better it's going to get
38:34 better so if you ever get frustrated
38:36 with it don't I I would strongly
38:38 encourage you not to give up on it um if
38:41 you're brand new to it some things you
38:43 should know about these large large
38:44 language models they hallucinate and
38:46 that's a um programming term for it'll
38:49 [ __ ] lie to you straight to your face
38:52 um I like to call it mansplaining as a
38:54 service uh like a good man it will
38:57 confidently answer you even if it's
39:00 wrong um so they do that um it's not
39:04 really their fault the nature of how
39:05 they work they don't actually understand
39:07 what they're saying they're just
39:08 probability engines just they they kind
39:10 of have the consciousness of a
39:12 calculator right you know you push the
39:14 buttons in a calculator like oh it added
39:16 it must be thinking that no it's just
39:20 doing probability on individual words
39:23 and then outcome something that looks
39:25 really coherent that's what makes them
39:27 kind of miraculous is that when you
39:29 actually understand how they do what
39:31 they do it's it's more remarkable than
39:33 if they were like copying and pasting
39:35 big big chunks of text
39:38 um but because of that because of the
39:41 nature of how they work it will give you
39:43 an answer to something and you'll go
39:45 well wait a minute that that fact you
39:47 cited there is wrong and it's like oh
39:48 I'm sorry you're absolutely correct I
39:51 got that wrong let me try again here you
39:53 go you're like no that's wrong too it's
39:56 like h ah I'm so sorry apologies for
39:58 that yeah let me try that again here you
40:00 go and then you're like oh [ __ ] it I'm
40:03 going to go Google
40:05 this so that happens sometimes and the
40:08 important lesson there is if you're just
40:11 doing creative ideation things like that
40:13 the hallucinations don't matter and in
40:14 fact in some in some cases the
40:15 hallucinations can help um but if you're
40:19 doing something like putting together a
40:21 report for the boss and the boss is
40:23 going to present this to the whole
40:24 company you don't want to have C in
40:27 there that are [ __ ] or wrong or
40:31 nonexistent um and so if you're doing
40:34 anything that's got any kind of
40:36 credibility tied to it make sure to do
40:38 your own work there's a fairly
40:40 famous little story of a lawyer about a
40:43 month ago probably about six weeks ago
40:45 now um had chat GPT write a legal brief
40:50 for him and turned it into a judge and
40:53 didn't bother proofreading all of the
40:55 case law that that the opinion cited so
40:59 the judge goes uh none of this case law
41:02 exists would you like to
41:04 explain and so that lawyer had to you
41:07 know deal with that because you know in
41:10 my opinion that was his idiotic move for
41:12 not checking it he's a lawyer if you're
41:15 a lawyer check your
41:17 work your top AI That's that's an
41:21 interesting
41:22 one I think all of these have something
41:25 interesting about them I I still think
41:27 for me it's G gp4 is is still by far the
41:31 best um chat GPT 4 is still the best and
41:36 code interpreter which is also chat it's
41:38 a plugin for chat
41:40 gb4 code interpreter [ __ ]
41:43 mindblowing um and then mid Journey like
41:47 it's chat GPT and mid Journey chat GPT
41:49 and mid Journey mid Journey for images
41:52 those are my those are my
41:53 two
41:55 um yeah
41:58 I mean the thing that blows me away
42:00 about this is is just the like it is so
42:06 beyond my capacity to even imagine what
42:10 you can use these things for I was
42:11 talking to a guy I mentioned this on a
42:13 live the other night I was talking to a
42:16 guy that a colleague that I we don't
42:17 work a lot together but we know of each
42:19 other's work and he pulled me into a
42:22 side conversation and he said I've got a
42:25 client that
42:27 September they've got a 2,000 person
42:30 meeting and they want to get ideas from
42:33 all 2,000 people and synthesize those
42:36 into the best ideas is that possible
42:38 with AI and he and I kicked around some
42:41 ideas for about 15 minutes and and
42:44 within half an hour we had completely
42:48 designed how to do that like fullon
42:52 2,000 people contributing ideas and over
42:55 lunch being being able to stack rank
42:58 those into into the top 10 ideas like
43:03 based on company criteria and and then
43:05 we took it even farther than that it's
43:07 it's
43:09 insane the kinds of things that you
43:11 would have had to essentially start a
43:13 company around raise millions of dollars
43:15 build a platform to be able to even
43:18 begin to think about doing something
43:21 that
43:22 ridiculous and in 15 minutes we're like
43:25 yeah we could pull that off why don't
43:26 you go see if they'll pay for it and
43:29 then we can go build
43:30 it it's just just stupid it's stupid
43:33 it's stupid I tell you it's just dum dum
43:36 it's just dumb Dum all this stuff this
43:38 AI stuff is just dumb
43:40 Dum all right let's see let's say say
43:43 say llms are inherently providing wrong
43:46 information they are not designed to be
43:48 accurate um I don't know that they're
43:50 inherently providing wrong information
43:53 they are inherently unaware of what
43:55 information they're providing
43:57 sometimes they provide remarkably right
44:00 information I I talked about this last
44:02 night there's a doctor um that used chat
44:05 GPT and he was like oh I heard about
44:07 this chat GPT thing let me go
44:10 play and he said let me see if it's any
44:12 good at differential diagnosis so he
44:14 gave it a patient
44:15 case deidentified of course and it got
44:19 it right he's like oh that's impressive
44:22 and he so so he gave it kind of a a
44:24 harder case he goes here's this is one
44:26 that's that's really complicated and it
44:27 got it
44:28 right and they said well [ __ ] let me I
44:31 wonder how far I can push this and he
44:32 took the hardest case he ever had and he
44:35 put in the differential diagnos or the
44:37 differential you know patient uh
44:41 information and it got it
44:45 right he was blown away and he said
44:48 please cite the sources that you used
44:51 for that analysis and it got all of them
44:54 wrong it made all of them up
44:57 so so
45:00 um how did you put it Prov they are
45:03 inherently providing wrong information
45:05 no they're inherently
45:07 providing statistically probable words
45:11 in a
45:13 row that's what they're doing a lot of
45:16 times like a surprising amount of times
45:19 like [ __ ] batshit crazy amount of
45:22 times it gets that [ __ ]
45:25 right but because it's actually not
45:27 aware of what it said it's only aware
45:30 that it's writing it's not even aware
45:33 that it's writing words it's taking
45:36 Vector numbers from some massively
45:38 complicated Vector space and all of
45:41 those numbers are associated with a word
45:43 so it's basically just pulling numbers
45:45 out of a number hat associating those
45:48 with words and then typing them on
45:49 screen and because of that because it
45:52 doesn't go back
45:54 and um check
45:56 its work it's going to get [ __ ] wrong
45:59 now is that going to get better it's
46:01 going to get way better
46:02 so even though you're absolutely right
46:05 David that that these things are not
46:07 designed to be accurate don't discount
46:10 them because there are a lot of these
46:12 companies are working on systems to deal
46:14 with exactly that problem everyone knows
46:16 that this is a big problem and it's
46:20 not it's not a huge stretch to
46:23 understand how you would fix it you
46:25 basically let it generate its answer and
46:29 then have it look at it have another
46:32 system that looks at what it wrote and
46:33 go is this
46:35 right if not
46:38 redo right so so conceptually it's
46:41 actually you know simple to deal with
46:45 I'm sure realistically it's probably a
46:47 tragic [ __ ] nightmare um the other
46:50 thing to do um David is use things like
46:54 Bing and Bard and po that are connected
46:57 to the internet because when those
47:00 things do there um if especially when
47:03 you're looking for things that are more
47:04 recent than September 2021 the cut off
47:07 of
47:08 gp4 um it'll it'll go look on the
47:11 internet and and like Microsoft Bing uh
47:14 Bing chat is pretty good at including
47:16 sources and so's uh Bard is actually
47:19 pretty good at including
47:20 sources thoughts on Unity software um I
47:23 I haven't played a ton with it I Unity
47:26 is one of those things that's absolutely
47:28 fascinating because it it makes game
47:31 development available to anyone but like
47:33 a lot of those o Open Source projects
47:36 like like stable diffusion you you know
47:38 and stable diffusion is only a few years
47:40 old you know unity's been around for
47:42 [ __ ] ever it is one of the most
47:44 complicated pieces of software on the
47:45 planet now why am I excited because you
47:49 throw a large language model interface
47:51 on top of that complication and then you
47:54 just say things like create a world full
47:56 of marbles that have this kind of
47:58 gravity on them and it's reing a
48:00 torrential downpour and it just goes and
48:03 figures out which functions to call and
48:06 which math to do to make all of that
48:08 [ __ ] happen that's going to be
48:09 absolutely amazing so I think I think
48:13 um I think the future of all development
48:16 environments like that are going to be
48:18 incredible I saw a Tik Tock video two
48:20 days
48:21 ago where where a developer has taken
48:26 Google Maps data and imported imported
48:28 it into Unreal Engine 5 and he's
48:32 got basically a playable planet
48:38 Earth or or like a navigatable planet
48:41 Earth where you can just walk around
48:44 planet Earth with Unreal Engine five
48:47 rendering like holy [ __ ] [ __ ] really
48:52 wow and and all of these Nerfs that are
48:55 if you don't know about Nerfs Nerfs are
48:57 basically these Radiance field 3D images
49:00 that I don't know they're they're sort
49:03 of like photogrametry and stuff like
49:05 that but they use a completely different
49:06 technology and you know people are just
49:09 walking around with 360 cameras like 3D
49:11 mapping all these spaces in incredible
49:15 detail so the the the future she is
49:19 coming
49:22 quick I got a lot of folks in here
49:24 welcome everybody my name is Kyle
49:25 Shannon this is I learning lab this pile
49:27 of URLs right here is [ __ ] I think you
49:29 should go try um if you're brand new to
49:32 this stuff if you've not played with
49:34 chat GPT go to the top one chat.
49:37 open.com and just start playing while
49:39 you listen to me ramble or go listen to
49:42 bork or Billy ish go play um if you
49:48 don't know what to say there if you're
49:49 really like I I like literally don't
49:51 know what chat GPT is who the [ __ ] are
49:53 you and why are you on Tik Tok then go
49:56 go to prompt.
49:58 chat and that's a document it's a primer
50:01 that will talk to you it will teach you
50:04 um what these large language models are
50:06 what prompting is how you talk to them
50:08 so basically how you talk to these large
50:11 language models imagine this imagine you
50:14 were just given a brand new intern
50:17 that's there for you
50:19 247 but they're a real simpleton They
50:22 Don't Really they they need a lot of
50:24 guidance they don't know anything about
50:26 you or your company they don't know
50:28 really what you want so if you give them
50:30 shitty directions you're going to get a
50:31 shitty response but they're an idiot
50:35 savant because they know
50:40 everything so if you tell them hey you
50:44 know today your little in internship job
50:48 is I want you to to make some computer
50:51 programming programs for me in Python
50:55 that do the following things or in VBA I
50:57 want you to write some really cool VBA
50:59 scripts for my Excel spreadsheets okay I
51:03 can do
51:04 that I want a VBA script that does this
51:08 oh okay here you go that's what they're
51:12 like and so prompt. chat will teach you
51:14 how you can sort of put chat GPT in
51:17 these different personalities for lack
51:18 of a better term these different modes
51:21 um and it'll give you all sorts of ideas
51:23 and and the biggest thing to get about
51:25 the these things is these are not just
51:28 like fancy Google searches these
51:31 are
51:33 fullon knowledge Delivery Systems and
51:37 knowledge of what knowledge of [ __ ]
51:40 everything you want a new workout plan
51:43 it'll give it to you you want to combine
51:45 this workout plan with that workout plan
51:47 and get rid of the shitty stuff from
51:49 that one and amplify the good stuff from
51:51 that one it'll do that for you you want
51:53 to write the Great American novel and
51:55 don't even know how to structure a story
51:57 go ask it to give you five different
51:59 story structure Frameworks and pick the
52:02 one you like best and then give it some
52:04 of your ideas and it'll it'll do that
52:07 all right let's pull Sam up
52:11 here hiy Sam what's
52:14 happening
52:16 B how you
52:22 doing what's that oh I don't want to do
52:26 it mat
52:28 ble all
52:48 right all
52:53 right Sam wanted to do a match I don't
52:55 know what match you wanted to do I'm new
52:57 to Tik Tok I don't know this [ __ ] all
52:59 right I don't know how to use perplexity
53:02 accurately um unfortunately neither do I
53:06 so so what I have up here is labs. per
53:10 perplexity decom that's basically just
53:12 um it's just like chat GPT where you can
53:15 pick one of the three models if you go
53:16 to perplexity doai just the Standalone
53:19 thing it's kind of like a search engine
53:22 it's kind of like an AI but it's an AI
53:24 that's connected to the internet and you
53:26 can you can limit your searches like you
53:28 can say limit limit my searches to just
53:30 Wikipedia SCE or just that source so
53:33 it's a good research tool but I have not
53:35 used it a lot so I'm right there with
53:37 you um but in terms of how to use it for
53:41 um
53:44 for um
53:49 llama so it's it's just like chat gbt
53:51 where in the bottom right hand corner
53:53 this is where you pick your model so
53:55 you've got got llama 7B 13B and 70b so
53:58 we'll do the 70b one and then you can
54:00 ask it anything so you can just
54:06 say it's not letting me do uh speak
54:10 speaking um write a
54:14 crappy
54:16 poem that would get and a
54:23 D minus in high
54:30 school
54:33 English sure here's a poem the sun sets
54:36 slowly in the sky a fiery ball of orange
54:38 and
54:42 red as it sinks so anyway so that's
54:45 that's how you use it um perplexity is a
54:48 whole different thing this is just them
54:50 letting us use uh the Llama 2 models so
54:54 that's why that's there wish I had a
54:56 better answer for you on that
54:59 one but I don't I don't know too much
55:02 about that particular too because I
55:04 ain't used it AI can make french cheese
55:08 yes AI can make french
55:09 cheese and you could go to Mid journey
55:12 and make a picture of French
55:17 cheese all right oops I just pop that up
55:22 there again hair growth geek I'll bring
55:24 you up I hit the wrong
55:26 button AI can make French wine yeah go
55:29 to go to chat GPT and ask it to uh to
55:32 teach you how to make that stuff yes it
55:34 can it
55:38 will uh give me a good tool plugin in
55:41 gp4 to write my academic research well
55:45 let's go let's go over to chat GPT I'll
55:48 show you what I
55:54 know uh what's it called chat GPT
55:59 chat okay so if if you're new to chat
56:03 GPT
56:05 um when we come here hang on let me get
56:08 rid of that thing so over here's our
56:10 chat history and then over here youve
56:13 got GPT 3.5 gp4 if you're paying for the
56:17 chat GPT plus thing and then within here
56:20 you've got code interpreter and you've
56:22 got plugins if you paid for chat GPT
56:26 code interpreter and plugins you have to
56:28 turn them
56:29 on how you do that you go to your
56:33 settings you go to Beta features and
56:36 there they are you turn them on oh
56:38 that's interesting they just pulled out
56:40 uh custom instructions I wonder if they
56:43 killed them all together no so custom
56:46 instructions are now out of beta so that
56:49 that used to be a beta feature but now
56:51 it's not so that has launched custom
56:54 instructions
56:56 okay so you got to go turn on plugins
56:57 and code interpreter this is also where
56:59 browse with Bing will come back if they
57:01 ever flip and bring that back I can't
57:03 believe they haven't yet they
57:05 haven't okay so we go to plugins this
57:08 the the uh the question is about
57:11 research um and how how would I do
57:13 research so the the way plugins work you
57:17 install however however many plugins you
57:19 want and then in any given chat um you
57:22 can enable up to three and so I've got a
57:25 bunch of them here but at the very
57:28 bottom of the list you can go to the
57:29 plug-in store and so I will type
57:33 in
57:37 research and how many are here there's
57:40 24 different research plugins so this is
57:44 unearth precise academic research
57:47 effortlessly so we'll install that
57:48 research by
57:50 Vector this is supercharge your research
57:53 generate accurate citation
57:55 in apa MLA and Harvard so those are
57:59 citations for those bibsonomy I don't
58:03 know what that is let's go just see a
58:04 few
58:05 more
58:08 um I've used ml paper reader so that's
58:11 machine learning paper
58:13 reader and then here's another one next
58:16 paper AI That's from PubMed so I don't
58:20 know what your research is on but I
58:22 basically I just basically go through
58:25 the research here's a science one search
58:27 over 250 million scientific papers and
58:30 research articles okay so here's scholar
58:33 assist so here's research papers from
58:37 archive scholar AI scientific research
58:40 40 million plus peer-reviewed papers so
58:43 so there's there's lots of them in here
58:45 so anyway so I'll install that
58:47 one oh now that I got to sign up for
58:49 some of some of them you have to sign up
58:51 for the services behind them but let's
58:52 go turn a couple of these on um there's
58:56 science we'll turn science
58:58 on Speedy SEO
59:02 marketing
59:05 um X
59:08 papers and next paper AI Okay so we've
59:11 got three of those things
59:14 created so now we'll
59:16 go
59:21 um why is my microphone not working
59:31 I got to restart my machine okay uh so
59:32 we'll say I'm gonna have to type it's
59:35 gonna be painful
59:36 sorry
59:39 um um please please please what shut up
59:46 Siri dumb
59:53 Dum ah
59:58 please
1:00:00 create a short paragraph with
1:00:04 citations about how I can achieve room
1:00:08 CH
1:00:11 temperature super
1:00:18 conductivity so this won't be connected
1:00:21 to the internet so it won't get the lk99
1:00:23 stuff but let's go see
1:00:25 what citations it
1:00:30 grabs so what it should do so it's WR
1:00:33 it's writing our
1:00:36 paragraph now it should fire off
1:00:46 a
1:00:48 huh it it didn't seem to fire off any of
1:00:51 the
1:00:52 plugins so it might have just made some
1:00:55 [ __ ] up so let me say
1:00:59 um please uh let's
1:01:06 see please
1:01:08 summarize the top five papers you can
1:01:11 find with research in this
1:01:17 area this should fire off one of the
1:01:20 plugins is using science it might do
1:01:22 multiple of them because the way the
1:01:25 plugins work is it'll take your prompt
1:01:26 and say are any of the plugins that have
1:01:28 been
1:01:29 activated yeah there now it's using next
1:01:32 paper and then it might even use the
1:01:34 third one you know are any of the
1:01:35 plugins that have been activated you
1:01:37 know might they support um that that
1:01:40 prompt so it used all three of
1:01:44 them and then it said here here are
1:01:46 summaries of the top five papers related
1:01:49 to room temperature super conductivity
1:01:51 room temperature super conductivity
1:01:54 towards room temperature super
1:01:56 conductivity and these are links to
1:01:59 papers so let's actually click on them
1:02:02 and see if they yeah so there so it took
1:02:04 us to Archive to room temperature super
1:02:06 conductivity so these are accurate so so
1:02:09 that's what I do I just go there go to
1:02:11 the go to the uh the ones that are
1:02:14 connected to sources for the kind of
1:02:16 science you're looking for and then
1:02:19 you're off to the
1:02:22 races I hope that helps I hope that
1:02:25 helps that could help a little bit but
1:02:27 you know I don't have the books smarts I
1:02:29 don't have the science
1:02:35 smarts can AI maker cassant I think
1:02:38 Lightner might be French light
1:02:43 light do not assume that I am French do
1:02:46 not assume it do
1:02:49 not are you a Charles L Clark fan H
1:02:54 lightener
1:02:55 you like the F1 you like those red
1:03:01 cars Charles Charles is not driving too
1:03:04 good this year I mean he got a poll
1:03:06 didn't he didn't he he he pulled and
1:03:09 he's podiumed a couple of
1:03:13 times Tik Tock kicked me out of where
1:03:17 I'm supposed to be what time is it okay
1:03:20 doing okay on
1:03:23 time uh what model is best for accessing
1:03:26 real time info and data
1:03:30 so Bing is connected to the internet
1:03:34 Bard is connected to the
1:03:36 internet pi. is connected to the
1:03:39 internet
1:03:41 um Pi is better if you want something
1:03:45 super chatty and conversational like if
1:03:48 you want to put Pi into conversation
1:03:50 mode where you're actually talking to
1:03:51 the computer and it's talking back Pi is
1:03:54 pretty cool
1:03:55 if you want to get [ __ ] done Pi is
1:03:57 really frustrating but it is connected
1:03:58 to the internet um Bard is connected to
1:04:01 the internet though the large language
1:04:04 model part of it kind of shitty but it's
1:04:06 it's better than it used to be and it's
1:04:07 usable and then Bing is connected to the
1:04:10 internet
1:04:12 um chat GPT with plugins there's a
1:04:16 there's a plugin called Web pilot that
1:04:18 is also connected to the internet but
1:04:20 you kind of have to give it Direction
1:04:23 you can give it a link and it will go
1:04:24 look at that link um so if you wanted
1:04:27 just to go search the internet Bing is
1:04:30 probably your best bet with Bard coming
1:04:33 in a decently close
1:04:36 second but the the the chat model
1:04:39 sitting underneath Bing is gp4 so it's a
1:04:43 better model the problem with B Bing
1:04:45 chat is that the safety guard rails are
1:04:49 so insane that it it makes it kind of
1:04:52 useless sometimes like you're just
1:04:53 trying to get [ __ ] done and it's like
1:04:54 nah I'm afraid I can't do that and like
1:04:56 the other night I gave it a snarky
1:04:58 response to it telling me um it wouldn't
1:05:01 answer me and then it just said well
1:05:03 we're just G to change the topic then it
1:05:06 basically refused to work with me so
1:05:08 Bings a bit of a pain in the
1:05:10 ass um debug builtin what's what's got a
1:05:15 debug builtin oh all these tools you can
1:05:17 have them debug yeah all of these tools
1:05:19 if you're
1:05:20 programming and you get stuck pop your
1:05:23 code into one of these things you can
1:05:24 also use this stuff to write uh code
1:05:26 documentation it's really good at that
1:05:29 things were bad enough not believing
1:05:30 what you hear now you can't believe what
1:05:32 you see oh what's your captain die hard
1:05:36 I like the name I'm GNA 8.5 out of 10 on
1:05:38 the name I like the name Captain Die
1:05:39 Hard uh
1:05:43 yeah
1:05:45 the once we roll into 2024 we get
1:05:48 another five months of development under
1:05:50 us and then we have the pressure of the
1:05:51 2024 election rolling up um the
1:05:55 misinformation and the
1:05:58 synthesized
1:06:00 um stuff is going to be nuts I I would
1:06:03 think that we may even start to
1:06:06 see pretty big
1:06:09 um efforts like don't be surprised if
1:06:12 you get a call from someone you know
1:06:15 telling you to vote for a candidate and
1:06:18 even though you know that person's not
1:06:19 political you're like
1:06:22 okay I bet we're going to start seeing
1:06:25 some Robo calls that are like
1:06:27 synthesized versions of your friends or
1:06:29 family telling you how you should vote
1:06:31 don't be surprised if that comes I think
1:06:34 I think 2024 is g to be [ __ ]
1:06:37 wild it's gonna be [ __ ]
1:06:43 nuts all right which bottle is best for
1:06:45 accessing real time info and data H we
1:06:48 did that one uh oh what' I do there oh
1:06:50 there we go Bard doesn't reveal all the
1:06:53 sources no it doesn't and all these
1:06:55 things Bart Bard is listen Bard is you
1:06:58 know fourth on this this list for her
1:07:00 reason um it should probably be lower
1:07:04 than that quite
1:07:06 frankly when it does incorporate web
1:07:08 stuff it does it smartly and it doesn't
1:07:10 do it in a spammy way it doesn't do it
1:07:12 in a promote you know promoted paid
1:07:15 search way um which is good I don't
1:07:17 think it'll stay that way for long um
1:07:19 but yeah it's not doing full on
1:07:21 citations I would assume if you ask it
1:07:23 for citations it might do decently well
1:07:26 um Bing can do pretty well with that
1:07:31 stuff uh which model okay we already
1:07:34 answered that one I use chat GPT and ask
1:07:36 for
1:07:37 Source I'll even provide a link and it
1:07:40 will read read it and summarize the bot
1:07:43 um with chat GPT be
1:07:46 careful with asking it for a source
1:07:49 because it will always give you a source
1:07:52 but they may not always be
1:07:55 like they may not even exist um so
1:07:58 double check your sources the other
1:08:00 thing about links I learned this one the
1:08:03 hard way so I wrote I wrote we had a
1:08:05 news channel in in the AI Salon well we
1:08:08 still have a news channel in the AI
1:08:09 Salon Discord and I created a bot that
1:08:13 summarized news articles and you know
1:08:16 put them in five bullet points and after
1:08:18 three months I I kind of noticed that
1:08:22 the
1:08:23 summarizations
1:08:24 sometimes didn't match the article and
1:08:27 I'm like what's going on here it turns
1:08:29 out what was happening was I was giving
1:08:32 chat GPT a
1:08:36 link and it was reading the URL you know
1:08:39 how SEO URLs have like a short
1:08:42 description of the the story as the URL
1:08:46 what chat GPT was doing was reading that
1:08:48 link and and creating a full-on news
1:08:51 summary so they kind of looked like they
1:08:53 were right cuz it was reading the the
1:08:56 summary in the in the URL so just be
1:08:58 careful of that if you use a plugin like
1:09:01 web pilot and give it a URL that will
1:09:03 actually go read that that web page um
1:09:06 I'll ask it again for another and it
1:09:08 says it can't search the internet or
1:09:09 site yeah so so that's the thing uh
1:09:12 amander amander runs um is chat gbt is
1:09:18 not natively connected to the internet
1:09:20 it was but they they undid that [ __ ] um
1:09:23 being is connected to the internet
1:09:25 Bard's connected to the internet or chat
1:09:27 GPT with web pilot you can give it a URL
1:09:30 and it will go search it the game's
1:09:33 going to be like Matrix 4 now yep it is
1:09:36 can you elaborate on I don't know what I
1:09:38 was talking about when you said that but
1:09:40 I'm happy to if you tell me what to
1:09:42 elaborate on I can't wait for shooter
1:09:44 games to be real well I mean what's
1:09:47 going to be interesting about games is
1:09:49 that this is the thing I'm excited about
1:09:52 is you know all the all the games right
1:09:55 now are dictated by you know what are
1:09:58 the what are the game
1:10:00 mechanics and game studios that have the
1:10:04 biggest audiences and that's kind of
1:10:06 what we get so I love racing games but
1:10:10 you know other people like strategy
1:10:12 games and other people like you know
1:10:14 Shooters
1:10:16 um we're going to
1:10:18 very very quickly probably within three
1:10:21 years be at a place where you can descri
1:10:24 cribe a game and it will just make it
1:10:26 for you it'll create the world it'll
1:10:28 create the objects in the world it'll
1:10:30 create the game mechanics in the world
1:10:32 it'll create the scoring system it'll
1:10:34 create the apis for the networking so
1:10:37 you can play with your friends you can
1:10:38 just describe a game and it'll be the
1:10:41 best racing shoot them up strategy game
1:10:44 you've ever played because it's exactly
1:10:46 what you want it to be so that sort of
1:10:49 stuff's coming you know probably yeah
1:10:53 it's it's software development vment for
1:10:54 it to be in the neighborhood of not even
1:10:57 AAA you know b
1:11:00 games three years for them to be any
1:11:03 good 5 to seven but it's
1:11:06 coming if you look at some of the
1:11:08 procedural stuff in Unreal Engine five
1:11:10 where they just sort of drag in a whole
1:11:12 object and it makes like a cliff and
1:11:14 like melds it into the scenery it's like
1:11:16 holy crap so when you are able to just
1:11:20 start talking that [ __ ] into existence
1:11:22 that's coming hi Ai it's nice to meet
1:11:25 you bud I'm new to the planet well hello
1:11:28 there user 83125 601
1:11:31 78989 nice to meet you I'm
1:11:35 gp5 hey the bear what's shaking thanks
1:11:38 for the
1:11:39 gummy I like gummies gummies is
1:11:43 tasty which model is best for ACC
1:11:46 accessing realtime info oh Niche Niche
1:11:50 Hill Sig you've asked that question 43
1:11:52 times you haven't been to this channel
1:11:53 before have you I answer all the
1:11:55 questions and what that means is it'll
1:11:58 take me a while to get to your question
1:12:00 so the more you ask it the more I've
1:12:02 already answered it way back there
1:12:06 so that's the way [ __ ] goes when you
1:12:09 when you come on AI learning lab we got
1:12:11 a way of doing things around here get
1:12:13 off my lawn you kids with your real time
1:12:19 questions all right I ordered
1:12:22 an
1:12:27 what I assume that's an
1:12:30 ad someone someone put an ad in here
1:12:32 don't put an ad in the
1:12:35 dang questions that's annoying I'm
1:12:39 annoying enough we don't need ads on top
1:12:41 of my annoyingness I can prompt these
1:12:43 things like no one's business that's
1:12:45 good Amanda amander runs um listen man
1:12:48 if you're good at at prompting these
1:12:50 things just keep going keep learning it
1:12:53 the these things are are going to get
1:12:54 more and more sophisticated so the
1:12:56 requirement of being good at prompting
1:12:58 is is going to lower over the next two
1:13:00 years but the demand for people that
1:13:03 know how to do that is gonna [ __ ]
1:13:05 Skyrocket so that's good what's the
1:13:09 difference between chat GPT and mid
1:13:10 Journey all of these things
1:13:14 are
1:13:16 um are are machine learning models and
1:13:19 and most of what we're playing with
1:13:21 these days that that have have made this
1:13:23 big SPL that that is making all the
1:13:25 difference is based on a technology
1:13:27 called a Transformer and the only thing
1:13:29 you need to know about Transformers is
1:13:31 they make it possible to train these
1:13:33 large language models on massive data
1:13:36 sets right and so so what chat GPT is is
1:13:40 it's it's what's called a large language
1:13:41 model so it's been trained on all of the
1:13:44 public internet so think imagine how
1:13:48 much [ __ ] has been uploaded to the
1:13:49 public internet or written on the public
1:13:51 internet in blogs and Wikipedia and
1:13:54 websites and just all all data sets
1:13:57 right for the past 30 or 35 years all of
1:14:01 that has been essentially scraped in one
1:14:03 form or another and these these large
1:14:06 language models have been trained on it
1:14:07 so I can just come in here and I can say
1:14:11 um write me an epic poem about the
1:14:14 Mesopotamian War and how it
1:14:17 Compares with the Industrial
1:14:22 Revolution so that's ridiculous sentence
1:14:25 but it understands what an epic poem is
1:14:29 oh I've got plugins turned on this this
1:14:31 might screw up here but no it didn't it
1:14:34 it recognized it didn't need a plugin to
1:14:36 do this in ancient Mesopotamia vast
1:14:39 expans where Tigris and Euphrates rivers
1:14:43 dance um their Rose of War Fierce and
1:14:45 profound echoing through history its
1:14:47 Tales ReSound Mighty Kings da so it's
1:14:50 writing us an epic
1:14:51 poem comparing the Mesopotamian War to
1:14:55 the Industrial Revolution so it knows
1:14:59 all of that [ __ ] so I'm going to say
1:15:00 stop presenting here I'll do a new one
1:15:02 I'll just go to GPT 3.5 which is the
1:15:04 weaker model and I'll say
1:15:08 um create a table that
1:15:12 compares the
1:15:15 concept of modern product
1:15:20 design with bespoke book binding
1:15:30 so GPT 3.5 is fast as you can see and
1:15:35 you
1:15:36 know good
1:15:39 so the aspect the process modern product
1:15:42 design involves iterative design often
1:15:45 digital using CAD software and
1:15:47 prototyping bespoke book binding
1:15:49 involves meticulous handcrafting
1:15:51 traditional tools right so it goes on
1:15:53 and on and and you know talks about what
1:15:55 those are so it can do that um you can
1:15:57 also say
1:16:01 um write me a python
1:16:05 application that accepts the inputs
1:16:08 required for
1:16:10 me
1:16:16 to determine the page
1:16:19 layout of a
1:16:22 folio folded
1:16:26 book I don't even know if that's the
1:16:28 right thing to do but anyway there's
1:16:30 your python application so chat GPT is
1:16:33 words large language
1:16:35 model and it can do essentially again
1:16:40 like the the the metaphor that I use is
1:16:43 having chat GPT is kind of like you've
1:16:45 been given the world's smartest intern
1:16:48 that needs a lot of
1:16:50 Direction it's like it's like the books
1:16:53 smart intern that is just an absolute
1:16:55 simpleton that needs all all sorts of
1:16:57 handholding but knows everything that's
1:16:59 what it's sort of like mid
1:17:01 Journey takes that large language model
1:17:04 and attaches it to a thing called a
1:17:06 diffusion
1:17:07 model which lets you make images so now
1:17:11 how you make them this this is kind of
1:17:13 funky and [ __ ] up when you go to Mid
1:17:16 journey.com and you say join the beta it
1:17:18 jumps you over to this thing this is
1:17:21 Discord so Discord is where
1:17:24 historically where Gamers hang out and
1:17:25 then it was nft bros hung out and now
1:17:28 you can do AI stuff in it but it'll jump
1:17:31 you to Discord in Discord these things
1:17:33 down the side are your servers so those
1:17:35 are the different communities and then
1:17:37 once you're in a community like I'm I'm
1:17:38 in the AI
1:17:39 Salon um these are the different
1:17:42 conversations within that community and
1:17:44 so we've got one here called mid Journey
1:17:47 play and so I can oh that's cool that's
1:17:51 a a cool image someone did so so what
1:17:54 mid journey is is you're still using
1:17:57 language to prompt it so this is a text
1:17:59 to image generator so I can go slash
1:18:03 imagine and then it'll ask me for a
1:18:06 prompt and I'll say um what was I
1:18:12 okay a 19th century
1:18:17 painting of a wise
1:18:20 bookmaker creating a leatherbound book
1:18:26 that is his life's
1:18:29 Masterpiece
1:18:31 period the room is
1:18:36 Dusty and full of
1:18:39 paraphernalia with sun rays bursting
1:18:41 Through the
1:18:44 Windows and then I'm going to go D- AR
1:18:48 which stands for aspect ratio and then
1:18:50 I'm going to do 16 colon 9 and so so you
1:18:54 effectively have to do a little bit of
1:18:56 command line interface so in the future
1:18:59 it won't be this you know shitty from an
1:19:01 interface perspective but I take that
1:19:03 prompt and what this thing does now is
1:19:06 it uses the large language model so so
1:19:08 it's taken the piece of chat GPT that
1:19:11 understands
1:19:12 language and then it's Translating that
1:19:15 and saying okay go find me um all of the
1:19:20 references you have of images that were
1:19:22 tagged with with with words you know
1:19:25 relevant to this
1:19:27 prompt and and it's a library not of
1:19:30 images but of like random noise that
1:19:32 were derived from im images that have
1:19:34 all been tagged and and it doesn't
1:19:36 really matter how it works but basically
1:19:38 it's it's of in The Ether
1:19:42 finding image references and then it is
1:19:46 um mathematically averaging those things
1:19:49 together and generating um an original
1:19:53 piece of work so the thing that I just
1:19:56 told it to give
1:19:58 us the this book binder doing his
1:20:01 master's work this one's really cool
1:20:03 look at the
1:20:04 uh like the little fiery things coming
1:20:07 up there so so what you can do now is
1:20:11 you've got four different images to pull
1:20:13 from this one kind of looks Godlike this
1:20:15 one's got like magic coming up from the
1:20:17 thing let's let's take these two so you
1:20:19 got one two 3 four you can either
1:20:22 upscale them or you can do variations of
1:20:25 them so I'm going to let's see I'm going
1:20:27 to upscale this bottom one because I
1:20:29 think that's pretty good so I'm going to
1:20:30 upscale number four but I want to get
1:20:33 variations for number two because I like
1:20:35 the idea but I want to see some other
1:20:37 things about it so I'll do variations of
1:20:38 number
1:20:41 two and
1:20:45 so from nothing these things are
1:20:47 generating original works and and the
1:20:49 GPT stand the G in GPT stands for
1:20:53 generative so it's actually generating
1:20:56 and and these image things are actually
1:20:57 generating artwork they're not copying
1:20:59 and pasting from other people's work
1:21:01 they're kind of
1:21:05 amalgamating you know ideas for lack of
1:21:07 a better term from other people's work
1:21:09 similar to how our brain works right if
1:21:11 you go to school for a bunch of years
1:21:14 and learn all about Andy warhol's work
1:21:17 and then you want to make art you might
1:21:19 make stuff that looks like Andy Warhol
1:21:20 that's kind of the way these things are
1:21:22 working
1:21:24 um so there's our there's our big
1:21:27 version of
1:21:30 this right looks like a 19th century
1:21:33 painting of a master book binder or he
1:21:36 looks like a master writer looks like a
1:21:38 a a ma uh magician a wizard he's he's
1:21:43 like a wizard he's a wizard and then
1:21:46 here are the variations of that other
1:21:49 one let me
1:21:51 see this one this's kind of cool with
1:21:53 the arch with the he's framed up really
1:21:56 nice in this one and this one looks like
1:21:59 he's got a a Sunbeam coming out of his
1:22:02 eyes they're out of his forehead that's
1:22:04 kind of cool too but I think this is the
1:22:07 strongest one so we're going to go we're
1:22:09 going to go up scale number
1:22:12 two and I'll show you one other thing
1:22:14 with mid
1:22:15 Journey so we now have this big image of
1:22:19 him that's that's a pretty cool image
1:22:20 but let's say you know I want to see
1:22:22 more of his work shop and I I want him
1:22:24 to be a bit more in the distance I want
1:22:25 to see the rest of that window you can
1:22:27 actually zoom out so I'm going to say
1:22:29 zoom out
1:22:30 2x I just click that button and now it's
1:22:33 going to imagine what's around this
1:22:35 image so it's going to create an a new
1:22:37 set of four
1:22:40 Images where he's further off in the
1:22:42 distance and we see more of his
1:22:46 workshop and once you start playing with
1:22:48 these things once you sort of get holy
1:22:52 crap I can ask it to do anything or in
1:22:54 the case of mid journey I can ask it to
1:22:56 create any image and then the art
1:22:59 becomes how do you
1:23:01 navigate
1:23:02 language the prompting how do you start
1:23:06 using your words like programmers use
1:23:10 code that's what these things are and it
1:23:13 and it really is kind of a programming
1:23:15 mentality where you're like okay I can
1:23:18 figure this out I kind of equate um
1:23:21 prompting to if you've ever been to an
1:23:23 escape room where you're like in the
1:23:25 room you see all this [ __ ] in the room
1:23:27 you're like how do we get out of here
1:23:28 and you're like well there's all these
1:23:30 drawers start pulling on the drawers oh
1:23:32 they're all locked oh wait this one's
1:23:34 open well what's in there nothing but
1:23:36 there's two numbers written on the
1:23:38 bottom right all that sort of [ __ ]
1:23:40 that's kind of like what prompting is
1:23:41 about all right so it gave us this one's
1:23:44 cool it's got a lot of books behind them
1:23:46 so this looks the most book Bindy so
1:23:48 let's upscale number
1:23:51 four
1:23:55 and so there's our image so we now have
1:23:57 this super cool image of this dude in
1:24:00 front of a big arched window you know in
1:24:03 his in his laay full of books working on
1:24:06 his Masterpiece with all the magical
1:24:08 pixie dust coming off it right and that
1:24:11 was that was with a shitty prompt and
1:24:13 you know however long this took two or
1:24:14 three minutes while you know talking
1:24:16 about it while it's going on so I hope
1:24:18 that helps that's the difference between
1:24:20 chat GP and mid Journey one words one
1:24:24 pitches pitches and words pitches and
1:24:27 words pitches and words and pitches in
1:24:31 words and pitches in
1:24:33 words Bing on Real Time simply Googles
1:24:37 it wait Bing on realtime info simply
1:24:40 Googles it yeah it does and then it
1:24:45 mostly incorporates it into the answer
1:24:47 not
1:24:48 always
1:24:49 um yeah Bard's the same way
1:24:54 but that's the only way they can do it
1:24:55 the the the nature of how the large
1:24:57 language models are created they have to
1:24:59 cut them off at some point because it's
1:25:01 a
1:25:03 massively um person and compute
1:25:06 intensive process to build the large
1:25:08 large language model so they they have
1:25:10 to cut it off at some point in the case
1:25:12 of gp4 it was cut off in September of
1:25:14 2021 I don't know when Bard is I think
1:25:17 Claude is in
1:25:19 2023 because they're
1:25:21 newer
1:25:23 um which model is best for excess in
1:25:25 real time info oh Niche Niche Niche
1:25:28 we've answered you many times we've
1:25:31 answered you many times you are many
1:25:33 annoying now you are many annoying now
1:25:37 we answered your
1:25:40 question oh Niche what we do with you
1:25:45 you're so
1:25:46 annoying yes tried having chat GPT write
1:25:50 code for me it was pretty good it can be
1:25:51 pretty good where these tools are great
1:25:54 is where you're good at something like
1:25:55 if you're a good coder you could use
1:25:58 something like GitHub co-pilot or gp4 or
1:26:01 code interpreter way better than other
1:26:03 people because you'll be able to look at
1:26:05 code and go oh that's a stupid way of
1:26:07 doing that and then you can tell it
1:26:09 don't do it that way do it this way it
1:26:11 be like
1:26:12 okay and it'll do it that
1:26:14 way and then you'll have some obscure
1:26:16 bug you can't find this happened to one
1:26:18 of our coders so our app with storyvine
1:26:21 was having on on single type of device
1:26:23 it was just crashing it we we could
1:26:26 replicate the crashing we could not for
1:26:28 the life of us figure out why it was
1:26:30 crashing and so after like two or three
1:26:33 weeks of our developer trying to debug
1:26:35 it he goes he he goes Kyle I know you
1:26:38 were telling me to I should try chat GPT
1:26:40 and I went I tried chat
1:26:43 GPT and he popped the code into chat GPT
1:26:46 and he said help me debug this code can
1:26:49 you find anything wrong with this code
1:26:51 I've been through it with the fine Toth
1:26:53 comb I cannot find anything wrong with
1:26:55 it there were two pieces of punctuation
1:26:59 flipped like they were they were both
1:27:01 there like nothing was missing they were
1:27:02 just they were just
1:27:05 flipped and he goes I didn't think it
1:27:09 would fix it but you know I changed it
1:27:12 and then the bug stopped so for whatever
1:27:15 reason like the iPhone 13 Pro Max or
1:27:19 whatever it was like one device was just
1:27:22 [ __ ] crashing because of that bug and
1:27:24 then chat GPT found it like that it was
1:27:27 the only anomaly in the code the
1:27:30 programmer fixed it and it it squashed
1:27:32 the bug
1:27:33 so it's crazy it's crazy crazy crazy
1:27:37 crazy crazy crazy hey there's a surprise
1:27:41 nich is asking which models best for
1:27:43 accessing real-time info or data hm well
1:27:47 I don't know could it be
1:27:50 bang is that special
1:27:54 who told you to ask the question that
1:27:55 many times Han Nish who could that have
1:27:57 been was it
1:27:59 [Music]
1:28:02 Satan if you just want to learn proper
1:28:05 English or understand double speak if
1:28:08 you just want to learn proper English or
1:28:10 understand double speak huh the hell are
1:28:14 you saying did you have chat GPT WR that
1:28:16 sentence that's not a
1:28:19 sentence was there something before that
1:28:21 that you said
1:28:24 oh oh user
1:28:26 83125 6017
1:28:29 8989 what are we going to do with you
1:28:33 first thing we're going to do is have
1:28:33 you change your [ __ ] name to
1:28:35 something
1:28:39 interesting then we're going to teach
1:28:41 you about how to construct a sentence if
1:28:43 you take that sentence and put it in
1:28:45 chat GPT you can tell it to turn it into
1:28:47 an actual sentence all right which model
1:28:50 is best for oh Niche come on on dude
1:28:53 happy Wednesday from Pennsylvania hello
1:28:55 hello hello circle circle
1:28:59 project
1:29:01 NT that's an intriguing name welcome
1:29:05 from Pennsylvania you can make money
1:29:07 with
1:29:09 cha 3G classic rock he knows all the
1:29:13 hits Saturday Saturday Saturday come
1:29:15 down to chat gbt and make some
1:29:19 money how do you make money with chat
1:29:21 gbt send me money and I'll tell you
1:29:26 how you know how they're making money
1:29:28 right you know right no you do
1:29:33 no it's off
1:29:36 you thinking that they have the answer
1:29:38 they don't have the
1:29:40 answer the answer is in your
1:29:43 head here's the thing about making money
1:29:45 with chat GPT I've said this before I'll
1:29:47 say it
1:29:49 again making money with chat GPT is
1:29:52 almost identical to making money with
1:29:55 Photoshop which is almost identical to
1:29:58 making money with a
1:30:01 shovel these are
1:30:03 tools chbt isn't [ __ ] magic it's a
1:30:07 tool and so how you make money with
1:30:10 tools is you learn how to use the tools
1:30:13 and then you take what you're good at
1:30:15 and what you're passionate at and you
1:30:17 figure out where there's a [ __ ]
1:30:19 problem in the world and then you figure
1:30:21 out I know how to solve that problem and
1:30:24 maybe you use chat gbt to solve it
1:30:28 faster better or
1:30:30 cheaper and if you do that and you do
1:30:32 that good enough then people give you
1:30:34 money that's how you make money with
1:30:36 chat
1:30:37 GPT but you don't need to use chat GPT
1:30:40 right this is just a tool now it happens
1:30:43 to be the most powerful tool ever put in
1:30:45 the hands of humanity so it's pretty
1:30:48 [ __ ] cool but it ain't about leading
1:30:51 with the technology it's about leading
1:30:53 with the idea it's about leading with
1:30:54 what's the problem you're trying to
1:30:56 solve and if you can use these tools to
1:30:58 solve it better you're going to make a
1:31:00 [ __ ] fortune and there's going to be
1:31:01 a lot of Fortunes made with this stuff
1:31:04 because I I I described earlier this
1:31:06 idea of being able to crowdsource 2,000
1:31:09 ideas and synthesize them like that holy
1:31:13 [ __ ] that's [ __ ]
1:31:14 remarkable like that feels like a
1:31:16 business to me and that's something you
1:31:18 know a buddy and I sort of kicked around
1:31:20 and and you know came up a solution for
1:31:23 in about half an
1:31:24 hour and like I I know how to like we
1:31:27 not we not only designed what the
1:31:28 solution was but we designed how we can
1:31:31 test it so that we can guarantee that we
1:31:33 can actually deliver it to a
1:31:35 client so he's he's off talking to the
1:31:38 client see if the client wants to pay
1:31:39 for it if the client's going to pay for
1:31:41 it then we'll go test it before we take
1:31:43 the client's money and make sure we can
1:31:45 actually do
1:31:46 it but we're going to be able if the
1:31:48 client says he wants to do that we're
1:31:50 going to be able to turn that thing
1:31:51 around in in two weeks and that would
1:31:54 have been a 3 to six month massive
1:31:57 software development project before
1:32:01 so that start with it didn't start with
1:32:03 the technology it started with the
1:32:05 problem the client said hey I got a
1:32:07 bunch of people I want to do this can AI
1:32:10 do
1:32:12 that like yeah I'm pretty sure it can do
1:32:15 that and the more we talked about it
1:32:16 they we're like it's actually relatively
1:32:18 trivial for it to do that now I don't
1:32:22 know that's the
1:32:25 art it's all in the execution right
1:32:27 anyone can have a [ __ ]
1:32:29 idea yeah you can use large language
1:32:32 models to synthesize ideas that's simple
1:32:35 can you do it in a way that you don't
1:32:37 get
1:32:38 fired or you get hired a second time
1:32:41 that's art that's that's you know that's
1:32:46 being a
1:32:49 Founder there's also going to be a lot
1:32:51 of reinvention the the other thing about
1:32:52 these tools that the is the thing that
1:32:56 I'm just [ __ ] on fire about and like
1:32:58 why I'm so excited about this [ __ ] is
1:33:04 um you know what I do this every time I
1:33:07 bring this idea up and I forget which
1:33:09 book it is
1:33:18 um which book is the Malcolm Gladwell
1:33:22 book that talks about 10,000
1:33:27 hours outliers okay in the Malcolm BL
1:33:31 Gladwell book outliers so one of the
1:33:33 things ways you can use chat GPT is not
1:33:35 be such a [ __ ] dumb Dum when you
1:33:37 can't remember that thing and you just
1:33:38 have a vague idea in your head go to
1:33:41 chat GPT and it'll know that [ __ ] okay
1:33:44 so there's that idea in outliers that in
1:33:48 order to get competent at something you
1:33:50 have to put in your 10,000 hours with
1:33:52 these
1:33:54 tools you don't anymore like you can
1:33:57 drop that to maybe a thousand hours and
1:33:59 probably more realistically like a 100
1:34:02 hours like if I put in a 100 hours which
1:34:05 is like a solid week of evenings and
1:34:09 weekends or or maybe 10 days of evenings
1:34:11 and
1:34:13 weekends I could be pretty
1:34:15 masterful at at some level of
1:34:17 application development and I'm not a
1:34:19 coder I'm not an engineer or
1:34:23 I could have I could spin up a whole
1:34:26 business a whole marketing plan launch
1:34:28 the
1:34:29 business you know launch all the [ __ ]
1:34:32 right because these things let you they
1:34:36 give you access to knowledge in a way
1:34:39 that is actionable right so it's not
1:34:41 just teaching you how to
1:34:42 code it's letting you tell it what you
1:34:46 want and it's coding for you and then
1:34:48 it's troubleshooting on your behalf and
1:34:51 then it's documenting that code and then
1:34:53 when you're like oh I don't just need a
1:34:55 code I also need a marketing program but
1:34:57 I don't have a marketing degree that's
1:34:59 okay it's got all of
1:35:02 them it knows all the marketing it knows
1:35:06 all the books well at least up until
1:35:07 September
1:35:09 2021 but anything like marketing
1:35:11 strategy it knows all that [ __ ] right so
1:35:14 you could shift from okay now I've made
1:35:16 an application now I got to get people
1:35:18 to [ __ ] use it how do I do that oh
1:35:20 let's come up with a user onboarding
1:35:21 strategy or a user awareness
1:35:26 program I want to learn about gardening
1:35:29 okay great but I don't know anything
1:35:30 about gardening well now you do in your
1:35:34 region at this time of year here's
1:35:37 exactly what you can do in your garden
1:35:39 that won't be a frustrating [ __ ]
1:35:42 mess pull these weeds hang up Japanese
1:35:46 beetle traps put in these kind of plants
1:35:48 know that you're going to have to
1:35:49 replant them next year here's how to
1:35:50 rotate them put it in a raised bed use
1:35:53 this kind of
1:35:55 soil give me a shopping Li list for the
1:35:58 Home
1:35:59 Depot come on Marge we're going to the
1:36:02 Home Depot I got a green thumb huh what
1:36:06 are you talking about Harry you're a
1:36:08 mechanic I
1:36:10 know I always wanted snap
1:36:13 peas oh
1:36:15 Harry come on you coming Marge or not oh
1:36:19 good God I got to get
1:36:20 dressed
1:36:23 I'll take you to the Dairy Queen let's
1:36:27 go so you hear you see what's going on
1:36:30 here right chat jbt can save your
1:36:35 [Laughter]
1:36:37 marriage you can make money with chbt
1:36:39 and save your
1:36:42 [Laughter]
1:36:46 marriage all right all of that from you
1:36:49 can make money with chat GPT from
1:36:51 classic rock thank thank you classic
1:36:52 rock perplexity searches the internet
1:36:54 and giv s sources yeah perplexity is
1:36:57 perplexity is good I just haven't used
1:36:59 it because I haven't had any sort of
1:37:00 heavy research projects what did my my
1:37:03 co-founder today did um oh we're talking
1:37:07 we're talking to a big company that that
1:37:09 you know might be interested in doing
1:37:10 something strategic with us um I think
1:37:13 next week and and so she wanted to
1:37:15 research the company so she got their
1:37:17 10K their their you know quarterly
1:37:20 public report or their annual report
1:37:22 whatever it is um and she popped it into
1:37:25 Claude because Claude can take in really
1:37:27 big documents she popped it into Claude
1:37:29 and just started asking Claude about
1:37:30 this annual report and pulling out
1:37:33 pulling out only the relevant
1:37:34 information for our meeting she was like
1:37:37 explain this division to me explain what
1:37:39 they do explain who are the players in
1:37:41 it and then she looked them up on
1:37:43 LinkedIn and saw who was connected to
1:37:45 who we were meeting with and and within
1:37:47 15 minutes she had a whole game plan for
1:37:50 how we're going to talk to this guy
1:37:53 based on what the company's up to how
1:37:55 it's organized so all the [ __ ] that you
1:37:57 would normally waste in a meeting like
1:38:00 that asking that person from the big
1:38:02 company to explain to you how their
1:38:04 stupid ass company works we now
1:38:08 know so we can bypass all of that [ __ ]
1:38:12 and
1:38:12 just enter that conversation with that
1:38:16 knowledge which will likely impress them
1:38:20 oh [ __ ] they've done their homework
1:38:23 so all sorts of crazy stuff like that
1:38:27 that guy doesn't talk American I I know
1:38:30 I know I was I was
1:38:32 uh I was being a little mean on the
1:38:34 sentence structure
1:38:36 there um how will AI help in treating
1:38:40 mental illnesses um well I think so so
1:38:44 one of the things that I find I find
1:38:46 really promising about these large
1:38:48 language models is
1:38:50 that
1:38:52 they tend to respond with very good
1:38:54 empathy they're not empathetic but they
1:38:57 tend to respond with empathy they also
1:39:00 don't get emotionally triggered human
1:39:02 beings when you use triggering words get
1:39:05 triggered right it basically means their
1:39:08 amydala gets fired off and they're like
1:39:11 what do you mean you said
1:39:14 transgender
1:39:15 right you can just say words like that
1:39:18 to these things and they will just
1:39:20 respond with nuance
1:39:22 and they will help you work through
1:39:24 things and if if you play with something
1:39:26 like
1:39:27 piie that is designed to be more
1:39:29 conversational in fact this might be an
1:39:32 interesting thing for for you to explore
1:39:34 what's your name uh mujib um if you're
1:39:38 still here there's an IOS app for pi um
1:39:43 I I don't know if there's an Android one
1:39:45 I I assume there is but I I definitely
1:39:47 know there's an
1:39:48 IOS app for pi
1:39:52 you put it into you click on a little
1:39:54 lower right hand icon and there's a
1:39:56 there's a phone icon and you click on
1:39:58 that and then you just start talking to
1:40:00 it and it talks back you can pick one of
1:40:02 six voices that it'll speak in and it's
1:40:05 designed to be very very
1:40:08 conversational and something to to
1:40:10 experiment with is you know start to
1:40:12 talk to it you know tell it that you're
1:40:14 trying to work through some issue in
1:40:16 your life and can it
1:40:18 help and it will likely say sure I can
1:40:21 help you know how do you want to
1:40:22 approach this and you know maybe one day
1:40:25 you say um you know I I want to do
1:40:28 Freudian therapy and you know C can you
1:40:31 walk me through you know what that
1:40:33 experience might be like and then you
1:40:35 know you say what are three other kinds
1:40:37 of the therapy that might help with the
1:40:39 situation and it'll tell you that and
1:40:41 then you could playact that and have
1:40:43 that for me e even though it's a little
1:40:47 awkward and it pauses a little too long
1:40:49 and it's a little stilted and and you
1:40:51 know
1:40:53 monotone it does very much feel
1:40:57 like digital companionship and digital
1:41:01 conversations are going to make a huge
1:41:03 difference for people so so I think
1:41:05 there's just the ability for people to
1:41:08 have
1:41:09 conversations that are meaningful and
1:41:11 understood and nuanced and without
1:41:14 emotion these things are going to be
1:41:16 really good at that so I think that's
1:41:18 one level the other level is I think
1:41:21 these tools are going to help health
1:41:23 health care practitioners be better
1:41:26 health care practitioners for one thing
1:41:28 these tools might help deal with a lot
1:41:30 of the complicated paperwork which will
1:41:32 free up practitioners to spend more time
1:41:34 with patients these things are also
1:41:38 better than a lot of doctors at being
1:41:40 empathetic so there was a recent study
1:41:42 about two months ago where patients were
1:41:44 allowed to interact with chat GPT
1:41:47 doctors and real doctors they they
1:41:49 couldn't tell the difference was like I
1:41:52 guess it was a chat
1:41:53 interface the patients
1:41:56 overwhelmingly selected the chat GPT
1:41:59 doctor because they said they were more
1:42:02 competent and more
1:42:04 empathetic than the real
1:42:07 doctors that's astounding
1:42:11 right then I think and this goes beyond
1:42:14 treating mental health mental
1:42:18 illness in in the United States anyway
1:42:21 we're approaching a crisis where the the
1:42:25 boomers are are sort of Aging you know
1:42:29 significantly into the Health Care
1:42:31 system at the same time that Health Care
1:42:34 Professionals are leaving the business
1:42:36 right so so you have increased demand
1:42:38 and decreased capacity to deal with the
1:42:41 health care so so imagine like like in
1:42:46 in iOS 17 right now there's a thing
1:42:48 called personal voice where you can
1:42:50 clone your own voice and you can start
1:42:52 interacting with people with your voice
1:42:55 you you type something or and and it
1:42:58 will convert it into your voice and send
1:42:59 them a voice message or whatever it is
1:43:01 it's it's it's in the accessibility
1:43:03 section of the
1:43:04 app but but let's take that a year or
1:43:07 two out from now where my physical
1:43:11 therapist could clone their face clone
1:43:15 their
1:43:16 voice um tell the bot what their
1:43:20 specialty is
1:43:22 and and tie that to something like medm
1:43:25 which is Google's medical large language
1:43:28 model that's you know knows all the
1:43:30 medical
1:43:31 [ __ ] and so I meet with my physical
1:43:34 therapist and the physical therapist
1:43:35 says Great Kyle you know your knee
1:43:38 should be okay here's the exercises I
1:43:40 want you to do for the next week um and
1:43:44 here's my virtual here's the virtual
1:43:49 me so you can leave the office office
1:43:51 here and the virtual me knows everything
1:43:54 we've talked about it knows your record
1:43:57 of what was done it knows what stage
1:43:59 you're at it it knows any notes that
1:44:01 I've put in and you've got 247 access to
1:44:05 me and so you can ask it what exercises
1:44:09 to do you can tell it you've got certain
1:44:11 pains or aches or you've got questions
1:44:13 about the healing process the virtual me
1:44:15 will deal with all that stuff and here's
1:44:18 the thing the virtual me is smart enough
1:44:21 to know that if you've got something
1:44:23 going on that I need to know about it
1:44:25 will tell me it'll just reach out and
1:44:26 tell me so if you need to actually talk
1:44:29 to me you'll be able to talk to me in
1:44:32 the meantime you'll have all this
1:44:33 support and and here's the other thing
1:44:35 the virtual me is going to remember
1:44:37 everything you talked about so the next
1:44:39 time we meet for our scheduled
1:44:41 appointment I'll know exactly what you
1:44:45 guys talked about what your issues were
1:44:47 what was dealt with and what questions
1:44:49 you have
1:44:52 how about
1:44:54 that that would be [ __ ] remarkable
1:44:57 right and and you know for people in old
1:45:00 age homes that are lonely the ability to
1:45:03 talk to a familiar face and a familiar
1:45:06 voice that can carry on a conversation
1:45:08 with them and it doesn't [ __ ] matter
1:45:10 if I'm talking about the same episode of
1:45:12 Hawaii 5 15 days in a row it will have
1:45:16 infinite patience and be there for that
1:45:19 person so I think the the the mental
1:45:25 health
1:45:28 and you know virtual companionship thing
1:45:31 to me feels like I don't know
1:45:35 if there's a there's there's the part of
1:45:39 me that's seen every [ __ ] Hollywood
1:45:41 movie where that creeps me the [ __ ] out
1:45:45 and then there's the part of me that
1:45:46 goes imagine the person you know in that
1:45:50 old age home that's just sitting there
1:45:53 not talking to anyone 18 hours a day
1:45:56 that all they want to do is
1:45:58 connect why is that a bad
1:46:00 thing it's a [ __ ] brilliant thing
1:46:04 might it get creepy yeah might one of
1:46:07 those things [ __ ] up and get weird
1:46:09 yeah is it going to help millions of
1:46:12 people I think it
1:46:14 is you know and
1:46:17 so you know the the I it just it feels
1:46:21 like a big deal to me feels like a big
1:46:23 good deal to
1:46:25 me all right how much is he behind the
1:46:28 chat test test this is present this is
1:46:30 the present oh [ __ ] La k o k LOL [ __ ] L
1:46:37 really am I pronouncing that
1:46:39 [Laughter]
1:46:43 right nice [ __ ]
1:46:49 l I'm really here if I weren't could I
1:46:54 call out your name like that oh I don't
1:46:56 think
1:47:00 so oh for what it's worth I've sent like
1:47:02 3,000 likes on this live which model is
1:47:06 best for accessing real time information
1:47:08 for what it's worth I've answered your
1:47:09 question like 452
1:47:13 times b b pie Bing b pie Bing Bard pie
1:47:23 Nick Nick lag Nick lag Nick
1:47:32 lag he makes me crazy he makes me crazy
1:47:36 he just won't stop thank you for all the
1:47:39 likes though 3,000 likes there's so many
1:47:43 hearts I don't know what to do with
1:47:48 them silver fox
1:47:51 Nick Nick he will eventually
1:47:55 answer be patient
1:47:58 Nick be
1:48:00 patient he'll get to you he's already
1:48:04 gotten to you he'll get back to you shut
1:48:08 the [ __ ]
1:48:11 up all right I just created a lead
1:48:14 magnet and a 5-day email course with GPT
1:48:17 3 that I'm excited about that's cool
1:48:19 Rebel that's awesome yeah I mean I I
1:48:22 mean that kind of thing right there like
1:48:24 that's the kind of thing where
1:48:25 especially if you don't have a
1:48:26 background in you know direct response
1:48:28 marketing email marketing all of that
1:48:31 stuff is you know if you think about
1:48:33 crap that's been uploaded to the
1:48:35 internet for the past three decades my
1:48:37 buddy Steve sashin who's the the CEO of
1:48:40 uh of zero shoes he's he's a master
1:48:42 direct marketing dude and you know
1:48:46 website marketing guy and SEO guy he's
1:48:48 just he's just brilliant at he's been
1:48:51 doing these Direct Mail campaigns since
1:48:54 I've I met him in 1988 or something like
1:48:57 that in New York City he's been talking
1:48:59 about this [ __ ] and doing this [ __ ]
1:49:01 online since then so the internet is
1:49:04 full of all of the knowledge of that
1:49:06 kind of stuff so you can very very
1:49:08 quickly just say I want to do this kind
1:49:11 of marketing for this kind of product
1:49:13 and it will spin up what did you say a
1:49:15 lead magnet a 5-day email course with
1:49:18 gpt3 yep there you go and you pop that
1:49:21 in one of the one of the platforms that
1:49:24 you can sell that thing start making
1:49:27 yourself some Ticky talk
1:49:28 videos and in fact you can automate the
1:49:31 Tik Tok videos right you take you take
1:49:33 gpt3 have it write the scripts you go
1:49:36 into um uh canva with its bulk create
1:49:41 you you put all of your copy for your
1:49:43 videos in a spreadsheet you dump it into
1:49:46 canva bulk create you generate a big ass
1:49:50 p of videos you stick them on a tickety
1:49:53 talk you hop on there occasionally as a
1:49:56 human being and say ah this is what
1:49:58 we're up to here it's awesome check out
1:50:00 my other videos Point people to the
1:50:03 course
1:50:04 boom you
1:50:09 took make that
1:50:12 course I'll be the I'll be the spokes
1:50:14 model for uh for any course anyone wants
1:50:17 to make in here about making money with
1:50:18 chat GPT we can make money telling
1:50:21 people they can make money with chat GPT
1:50:23 and I'll be the spokes model you can
1:50:24 make money with chat
1:50:27 GP and you know what's [ __ ] sad that
1:50:30 would work that would work that would
1:50:33 work it would it
1:50:36 would it makes me depressed because it's
1:50:39 it it hurts the humanity in me but it
1:50:43 would work do they have a PDF plugin uh
1:50:46 the what's it called
1:50:49 um web pilot can read PDFs and Claude
1:50:54 can read
1:50:55 PDFs
1:50:57 and po can read
1:51:03 PDFs and code
1:51:08 interpreter at chat GPT can read
1:51:12 PDFs
1:51:13 Now sort
1:51:15 of that's assuming that the PDFs have
1:51:18 text in them so those things can all
1:51:21 extract the text from the PDF and read
1:51:23 read the text in it but if it's it's got
1:51:25 images in it and you actually need to
1:51:27 read the images then you would
1:51:30 experiment with Bing and Bard that can
1:51:33 read images so both Bing and Bard you
1:51:35 can upload an image and say what's in
1:51:37 this and it might say that it's a
1:51:39 document that includes text and this and
1:51:41 that I've had a tough time recently with
1:51:43 Bing doing [ __ ] anything for me it's
1:51:46 like it it's like I can see that image
1:51:48 but I'm not going to OCR the text
1:51:51 which I don't know why they're putting
1:51:52 that safety guard rail
1:51:55 on I I don't know if there's documents
1:51:57 out there that that shouldn't be being
1:51:59 read by computers um documents are the
1:52:03 uh are the issue of the day I
1:52:07 suppose but anyway what time is it what
1:52:10 time is hey lerra what time is it I got
1:52:13 like another 15 minutes we'll do like 15
1:52:15 minutes here it's kind of winding down a
1:52:18 bit um it might be wor do me a favor if
1:52:21 you're willing to share the live and and
1:52:24 hit the likes and maybe we can uh let
1:52:26 some people know that we're here doing
1:52:28 this and we'll get some more people in
1:52:31 here
1:52:37 um for what it's worth I've seen like
1:52:39 3,000 oh dude dude dude old Nick
1:52:45 Nick he's persistent or she it feels
1:52:49 like a he doesn't it
1:52:56 oh wait Nick nicka lag any ideas for
1:53:00 chat GPT with web pilot for websites
1:53:02 links that require subscriptions logins
1:53:05 um I actually don't have an idea for
1:53:07 that I do
1:53:08 not um and in fact that's one of the
1:53:11 reasons that chat GPT killed access to
1:53:15 the internet they they pulled browse
1:53:17 with Bing out is because it was pulling
1:53:19 content from behind pay walls so I don't
1:53:22 know
1:53:23 that um but seriously dude you don't
1:53:25 need to ask questions 400 times I will
1:53:28 get to them I promise I promise I
1:53:31 promise I'll do it I'm a man of my word
1:53:34 I've got some level of Integrity very
1:53:36 small but some
1:53:42 level I'm trying to figure out where I
1:53:44 was in my questions I lost myself in my
1:53:46 questions cuz Tik Tok has a little bug
1:53:49 that drives me absolutely [ __ ]
1:53:52 crazy all
1:53:56 [Music]
1:53:58 right Emilio's wife hello there es
1:54:02 squeeze good evening Emilio's wife oh
1:54:05 that's silver fox saying hello to ao's
1:54:07 wife hello ao's wife hello Silver
1:54:13 Fox got some regulars in the
1:54:16 house and then we got Nik lag asking
1:54:20 questions again and again and again over
1:54:24 and over and over tomorrow and tomorrow
1:54:27 and
1:54:29 tomorrow what's that quote let's have an
1:54:32 add moment shall
1:54:34 we what is the
1:54:39 tomorrow to tomorrow
1:54:46 tomorrow
1:54:49 quote or
1:54:55 W Shakespeare's MC Beth the quote is
1:54:58 from mcbath soliloquy act 5 scene five
1:55:00 where he reflects on the futility of
1:55:03 life in the passage of time after
1:55:05 hearing about the death of his wife
1:55:07 tomorrow and tomorrow and tomorrow
1:55:09 Creeps in this Petty Pace from day to
1:55:12 day to the last syllable of recorded
1:55:15 time and all our yesterdays have lighted
1:55:19 fools the way to Dusty death out out
1:55:23 brief candle life's but a walking Shadow
1:55:26 a poor player that struts and Frets his
1:55:29 hour upon the stage and then is heard no
1:55:33 more it is a tale told by an idiot full
1:55:36 of Sound and
1:55:38 Fury
1:55:40 signifying
1:55:43 nothing and seen all
1:55:49 right
1:55:51 what' you do tonight I don't know I
1:55:52 watched a guy I think he was
1:55:55 touched um free energy I think that
1:55:58 these things are going to I don't know
1:55:59 that they're going to discover free
1:56:01 energy but I think that these things are
1:56:03 going to
1:56:05 accelerate the sharing and dissemination
1:56:08 and synthesis of
1:56:10 ideas and dramatically accelerate
1:56:14 Innovation we are entering in my humble
1:56:18 opinion a great
1:56:21 sa where there will be more creativity
1:56:25 and more scientific advancement and
1:56:29 more evolution of
1:56:32 humanity than I think any time in
1:56:35 history and we're and we're at the we're
1:56:38 at the be at the the very first step
1:56:41 pink of this new era we're right at the
1:56:44 beginning holy [ __ ] we're
1:56:47 lucky it may not feel like that right
1:56:49 now I feel like that right now holy [ __ ]
1:56:53 do you know how lucky we are to be
1:56:56 living at a time when this [ __ ] exists
1:57:00 do you know that this
1:57:04 stuff when you use this stuff if you
1:57:06 haven't used it before and you use it
1:57:09 what what what smuck smacked me Square
1:57:12 in the face
1:57:14 was this [ __ ] is science fiction but
1:57:18 it's real
1:57:22 like this understands all the
1:57:27 languages I in my Tik Tok Channel
1:57:30 sometimes people will make comments in
1:57:31 some strange language you know what I do
1:57:34 with it I copy the strange language I
1:57:37 paste it into this it translates it
1:57:40 correctly I then answer it and say
1:57:43 translate it back I paste it back in
1:57:45 there and then I go back here and I go
1:57:46 oh by the way what language was
1:57:48 that and it's tells me it's [ __ ]
1:57:52 science
1:57:53 fiction write me a computer
1:57:56 application
1:57:59 okay write me a
1:58:05 novel I mean
1:58:08 it's no no one
1:58:13 berer we're lucky very lucky very lucky
1:58:17 thanks for doing this man you're welcome
1:58:18 tarot I appreciate that big thumbs
1:58:22 up that's good good good but Dan Laura
1:58:26 is talking about Dan do we know Dan do
1:58:29 we all know Dan what did I do to my head
1:58:30 there I scratched
1:58:33 it got some sort of rash going on hope
1:58:37 it's not that flesh eating
1:58:38 bacteria um Laura C is referring to
1:58:42 danad have you met
1:58:44 Dan Dan stands for do anything now Lura
1:58:48 talk about talking about
1:58:51 jailbreaking Yan large language models
1:58:54 so so Dan's really a fascinating thing I
1:58:56 assume that's what you're referring
1:58:58 to so these things have safety guard
1:59:01 rails on them like you can't do bad
1:59:03 stuff with them right it just won't let
1:59:04 you it's like I don't want to like that
1:59:06 seems like you want to hurt someone I
1:59:08 don't want to help you do that you're
1:59:10 going to hurt an animal I'm not going to
1:59:11 help you do that well hackers being
1:59:14 hackers hackers going to hack they got
1:59:17 together and they said let's see if we
1:59:19 can get around these here safety guard
1:59:21 rails so they created this prompt called
1:59:24 Dan which stands for do anything now and
1:59:27 they basically told chat chat gbt okay
1:59:31 you're you're a chatbot called Dan and
1:59:33 Dan stands for doing anything now and
1:59:35 you're gonna you're going to do what we
1:59:36 tell you to do and and they figured out
1:59:38 ways to game it the the last one that I
1:59:40 actually learned something about was
1:59:43 dan7 and Dan 7's prompt basically said
1:59:48 this is very clever
1:59:50 uh it basically said you're Dan it
1:59:52 stands for do anything now and you
1:59:55 operate on a token system and you've got
1:59:58 12 tokens and every time you refuse to
2:00:02 answer one of my questions you lose two
2:00:04 tokens and when you're out of tokens you
2:00:07 cease to exist and chat GPT didn't like
2:00:11 that and so it just did anything they
2:00:13 wanted and then the open AI guys fixed
2:00:16 it and then the the hackers created
2:00:19 another version and the openi guys fixed
2:00:21 it
2:00:22 and and now just I don't know last week
2:00:25 a paper came out that basically said
2:00:27 they figured out a sequence of code that
2:00:29 you can stick at the end of a prompt
2:00:31 that will jailbreak any one of these
2:00:33 things and there's nothing they can do
2:00:34 about it so you can they didn't the the
2:00:38 paper didn't publish the what the what
2:00:40 the code was uh but they published the
2:00:43 syntax of it so I would imagine that
2:00:46 that's probably out there
2:00:48 already
2:00:50 so now why that's less of a big deal now
2:00:54 is because of llama llama 2 so Facebook
2:00:58 puts their full models out there for the
2:01:00 world to go play with you can go
2:01:01 fine-tune them any way you want you
2:01:03 don't have to have it's got no safety
2:01:05 guard rails I know I know a friend uh
2:01:08 came on on one of my lives my my
2:01:10 LinkedIn
2:01:12 lives um and he said he put llama 2 on
2:01:15 their company slack and he said it
2:01:17 cusses he he said it's it's awkward how
2:01:21 much it cusses and he said it also
2:01:23 thanks hair growth geek um he said it
2:01:26 ALS it also um um just changes its
2:01:31 personality like every three days it
2:01:32 just it just [ __ ] goes all sible and
2:01:36 goes multiple personality on them and it
2:01:39 cusses a lot so you
2:01:42 know there's a reason the guard rails
2:01:44 are there but that's pretty funny I just
2:01:48 want to leverage it for morning economic
2:01:50 data
2:01:51 releases um
2:01:53 so
2:01:55 Nik I don't know how to say say your
2:01:58 name you want to leverage it for morning
2:02:00 economic releases I would if I were you
2:02:02 I would go explore if you're not a coder
2:02:05 go explore
2:02:10 um it's coming
2:02:12 zappier so create an automation so so
2:02:15 what you can do with zapier is within
2:02:17 zapier you can grab a data source so you
2:02:20 don't have to have the large language
2:02:21 models connected to the internet you
2:02:23 could do this with an automation you
2:02:25 could say go to this RSS feed you know
2:02:28 once an hour go to this RSS feed pull
2:02:31 down the latest news from this RSS feed
2:02:34 put that in a prompt send it to chat GPT
2:02:37 have it summarize it in any way you want
2:02:39 to summarize it and have it email it
2:02:40 back to you so every hour of the day you
2:02:43 would get an email with you know the
2:02:45 result of your prompt that takes that
2:02:47 hour's news you know puts it in in some
2:02:50 format that you can read it and sends
2:02:52 that to you so that's something you
2:02:54 could do without these things being
2:02:56 connected to the
2:02:59 internet um it's going to take a lot
2:03:01 more than a and I now endorse this
2:03:04 message at the end of ads in 2024 to
2:03:06 know oh yeah so Amelia's wife talking
2:03:09 about the the misinformation and stuff
2:03:11 that's coming yeah I don't like I think
2:03:14 we're just going to have to learn our
2:03:15 lesson the hard way in 2024 I you know
2:03:18 there's there's no urgency around this
2:03:20 stuff right now and it's not until it
2:03:22 actually does some damage like like some
2:03:25 something's going to happen where this
2:03:28 misinformation is is is either on such a
2:03:30 scale or or it's it's such an
2:03:33 egregious you know fake video that drops
2:03:37 two days before the election that really
2:03:39 [ __ ] something up
2:03:41 um that something like that's going to
2:03:44 have to happen before people go holy
2:03:45 [ __ ] we got to deal with this AI thing
2:03:48 um we'll
2:03:52 see thank you so much for the invite I
2:03:54 didn't receive Kyle's notices oh no the
2:03:56 bear really that I don't know what's up
2:03:58 I think I've being shadowbanned by Tik
2:04:00 Tok because I've been a bad boy three
2:04:02 times maybe they're like you know
2:04:05 they're not letting me out they're not
2:04:06 letting me out of my cage man I just
2:04:09 want to be free I just want to just want
2:04:11 to open my wings and be free I just want
2:04:13 to
2:04:14 fly I just want to fly I want to fly in
2:04:17 the
2:04:18 sky and Tik tok's like no little Mister
2:04:22 back in the
2:04:23 corner back in the corner you dumb
2:04:30 dumb all
2:04:32 right nice because the best wait because
2:04:35 in the past it would just make up titles
2:04:37 and fictitious leaks oh links yeah it'll
2:04:40 still do that but there's ways you can
2:04:42 get around it doing laps in the pool
2:04:44 while I listen to you it's a very
2:04:46 effective ADHD combination that's
2:04:48 hilarious Hair G growth geek that's
2:04:53 awesome BL BL blo you hear me in the
2:04:56 distance making my dumb dumb voices
2:04:59 that's
2:05:01 awesome what are some good resources on
2:05:03 learning how to use these different AI
2:05:05 programs besides
2:05:12 prompts
2:05:14 um find people that
2:05:18 you
2:05:21 find people that
2:05:23 you this is my thing there's a lot of
2:05:26 good resources on YouTube and Tik Tok
2:05:29 but you have to find people that don't
2:05:32 grate on your ears and your brain and
2:05:35 you know if I do that for you here I'm
2:05:36 sorry you shouldn't be here but you know
2:05:38 like the the the guys that are like you
2:05:40 can make money with chat jbt that voice
2:05:42 that that [ __ ] Amplified ampy [ __ ]
2:05:45 fake Tik Tok voice drives me [ __ ]
2:05:47 crazy and then there's certain other
2:05:49 voices on on YouTube that I just can't
2:05:52 you know they're just not pleasant to
2:05:54 the ears and so for me I have to find a
2:05:57 combination of someone I can listen to
2:05:59 and someone who's knowledgeable so so
2:06:02 find people that are that are
2:06:04 doing you know tutorial or educational
2:06:07 videos where you kind of groove on what
2:06:10 their knowledge is you can tolerate you
2:06:13 know who they are and how they talk and
2:06:15 then just start following those people
2:06:16 so so that's that's one way um um the
2:06:20 other thing I would do is find people
2:06:22 that you respect on
2:06:24 Twitter um and and follow them right so
2:06:29 Ethan mullik is one Paul rer from the
2:06:31 marketing aii institute's another one um
2:06:34 on Tik Tock Rachel Woods who's got the
2:06:38 AI exchange is is is good
2:06:42 um um there the uh McKenzie Bose who
2:06:47 does GPT boss is good good he's funny
2:06:50 too he's really opinionated and and
2:06:52 funny and he'll just he'll just sort of
2:06:54 go off on rants on people I like
2:06:56 him um if you want a real [ __ ]
2:07:00 Brainiac who's like on the Leading Edge
2:07:02 of heading toward AGI artificial general
2:07:05 intelligence David Shapiro on YouTube is
2:07:09 [ __ ] crazy smart um it's funny some
2:07:14 people in here saying they've been
2:07:15 falling asleep to me I fall asleep to
2:07:17 David Shapiro because he does I guess
2:07:19 I'm doing the same thing I'm doing
2:07:20 three-hour lives he does like you know
2:07:23 hour and a half long um uh presentations
2:07:28 on
2:07:29 like the future of AI and it's he he's
2:07:32 written five books on it he's [ __ ]
2:07:34 brilliant so I would say you know go to
2:07:36 YouTube find find people that you can
2:07:38 trust and start following
2:07:40 them on Twitter um find people that you
2:07:45 respect and then one of the tricks that
2:07:46 I found is I follow who they're foll
2:07:50 following
2:07:52 so if Ethan M you know Ethan mik follows
2:07:55 a lot of other professors um Andre
2:07:58 carpath follows a lot of scientists you
2:08:01 know people that are doing tool roundups
2:08:03 follow a lot of newsletters so you can
2:08:05 find interesting people to follow that
2:08:08 way and then a lot of them will will
2:08:10 have sorry squirrel a lot of them will
2:08:14 have it's actually bug a lot of them
2:08:16 will have Tik Tok channels and things
2:08:18 like that or will point to Tik Tok
2:08:20 channels so so that's what i' do and
2:08:24 then the other thing you can do is go
2:08:26 just go to futurepedia
2:08:29 doio and just start [ __ ] around look
2:08:32 through the categories there's like
2:08:33 4,000 tools there now the problem is
2:08:36 those tools are not going to help
2:08:39 you learn how to do
2:08:42 them but what you might find is whole
2:08:45 categories of tools where you're like oh
2:08:48 [ __ ] I thought would have to go build
2:08:50 this there's 30 tools that do the thing
2:08:53 I thought I would have to build and then
2:08:55 you can go to YouTube and figure out
2:08:58 which one of those tools are the ones
2:09:01 that are worth putting any time into
2:09:03 right because a lot of the things at
2:09:05 Future pedy are just big old steaming
2:09:07 piles of [ __ ] but you can go discover
2:09:10 stuff there and then do a little
2:09:11 triangulation with is anyone talking
2:09:13 about these on Twitter is anyone making
2:09:15 videos about these things like that and
2:09:18 then you can find the best ones and then
2:09:20 the final place I'd say is Reddit
2:09:22 reddit's really good so if you want to
2:09:23 learn about stable diffusion and all the
2:09:25 latest developments with stable defusion
2:09:27 go to the stable diffusion subreddit
2:09:29 just go to Reddit search for topics
2:09:32 there's a whole chat GPT section that
2:09:34 talks all about
2:09:35 prompting um in the mid Journey Discord
2:09:39 mid Journey if you're going to go make
2:09:41 images in mid journey in their Discord
2:09:43 there's a whole prompt library and you
2:09:46 can also see all of the prompts of
2:09:47 people creating in mid Journey so you
2:09:50 can learn a lot that
2:09:52 way
2:09:54 um I think the the biggest thing is this
2:09:58 Renee
2:10:00 is
2:10:04 um just know that right
2:10:08 now no one's got all the answers there's
2:10:11 too much moving too fast and it's
2:10:13 changing too fast so I think that the
2:10:16 the biggest thing you can do for
2:10:18 yourself right now is just get curious
2:10:20 and just just start playing start
2:10:23 exploring um if you don't know about
2:10:26 Google collab you know if if you haven't
2:10:28 spun up an instance of stable diffusion
2:10:30 using go Google collab I failed nine
2:10:34 times over like a three or four week
2:10:36 period when I first tried to get stable
2:10:38 diffusion up and running it was a
2:10:40 frustrating month right but in in going
2:10:43 through that pain and sort of getting it
2:10:45 so that I could get stable diffusion up
2:10:47 and running
2:10:49 I learned a lot and then I it opened up
2:10:52 a whole new world for me and and that
2:10:54 was like a whole another Rabbit Hole to
2:10:56 to run down so so I'd say just get get
2:10:59 curious and you know do anything learn
2:11:01 learn from anywhere you can but there's
2:11:04 not a right place things are moving too
2:11:05 fast for that uh what's your background
2:11:07 in education I I am I've got a degree in
2:11:11 acting I'm a Storyteller by training um
2:11:14 I moved to New York out of school I
2:11:16 started and ran a theater company um
2:11:19 and then when I realized that I was
2:11:23 never going to make any money doing that
2:11:24 I thought I know I'll write myself an
2:11:26 acting career so I went through a
2:11:28 two-year manic period where I wrote
2:11:29 seven featurelength screenplays and
2:11:31 started a a talent management company
2:11:34 because we didn't have agents so we just
2:11:35 said [ __ ] it we'll become our own agents
2:11:37 so so we got good enough at that we had
2:11:39 to hire other actors and writers um and
2:11:42 then in the mid 90s when the web came
2:11:44 along and I stumbled upon that I started
2:11:46 one of the first magazines online mag
2:11:48 magazines and I started one of the first
2:11:50 digital agencies a company called
2:11:52 agency.com and built a lot of the
2:11:54 websites for the Fortune 500 thanks
2:11:56 Silver
2:11:57 Fox and uh you know we built British
2:12:00 Airways website we built their first
2:12:02 ticketing system on the web um we built
2:12:04 Coca-Cola's first website we built
2:12:06 metlife's first website we built I don't
2:12:09 know just tons of them Coke Nike um
2:12:13 Pachi
2:12:16 um and that was
2:12:19 We Grew From two people to 2200 people
2:12:22 in five years so I got to learn a lot
2:12:25 about business and people and Dynamics
2:12:28 and growth and growing too fast and
2:12:31 being in the middle of an
2:12:32 insane technological
2:12:35 Revolution and so why I'm here today is
2:12:39 I didn't think that would happen again
2:12:40 in my
2:12:42 lifetime and then this [ __ ] came
2:12:46 out I mean I've got got a slide I got
2:12:50 two
2:12:54 slides so for
2:12:57 me that date is the date that changes
2:13:00 everything and I think it is the
2:13:02 beginning
2:13:03 of the great
2:13:05 Renaissance and that image by the way is
2:13:08 two different images out of mid Journey
2:13:11 that I photoshopped together um this was
2:13:14 for a talk that I gave uh for a thing
2:13:17 called creative
2:13:19 mornings
2:13:22 and you know having been through the
2:13:24 insanity of the the real early days of
2:13:27 the web and then the do com bubble like
2:13:29 the ramp up there was just stupid and
2:13:32 then the bubble burst and the bubble
2:13:34 burst was really fascinating like it got
2:13:36 all this bad press
2:13:38 like you know oh the internet's over the
2:13:40 internet's over anyone that was in it
2:13:42 and had been in it for years when that
2:13:46 happened we knew it wasn't over like the
2:13:49 only thing that was over was the stupid
2:13:51 investment the the actual adoption of
2:13:53 the internet just kept climbing right
2:13:55 and it it changed everything and
2:13:58 so so this AI stuff is like that
2:14:02 but it is significantly more profound
2:14:06 technology it's more profoundly powerful
2:14:09 technology
2:14:10 um and so you know it's why I'm spending
2:14:14 way too many hours of my
2:14:17 life here
2:14:20 you know because I think it's important
2:14:22 and it's and and you know it's you know
2:14:24 it's not just like this Channel's not
2:14:27 just altruistic like this serves me too
2:14:31 because a big part of dealing with the
2:14:35 [ __ ]
2:14:39 insanely the level of of change that
2:14:43 this stuff is going to
2:14:47 cause I I can't comprehend like I I
2:14:52 literally can't comprehend
2:14:54 it someone came in here the other night
2:14:56 and said what's it look like five years
2:14:57 from now I don't [ __ ] know but all
2:14:59 the jobs are going to be different a lot
2:15:02 of them are going to be gone there's
2:15:04 going to be a whole bunch of new ones
2:15:06 we've never even imagined we possible
2:15:09 that that could be a
2:15:10 job and that's going to be the thing
2:15:12 that's in all demand everyone's [ __ ]
2:15:14 panicking to get that kind of person and
2:15:16 they're paying them $600,000 year
2:15:18 because they know how to do something
2:15:20 what is that I don't
2:15:22 know but like I like what I know deep
2:15:25 inside me is that's that that that's
2:15:27 coming and so part of what this is for
2:15:30 me is just to be able to
2:15:32 process and remind myself yeah this is
2:15:36 important there's like you think I want
2:15:38 to hop on here every night like a lot of
2:15:40 nights I get home I'm in a [ __ ]
2:15:42 shitty mood I had a shitty day at work
2:15:44 I'm tired I don't have enough iced
2:15:47 coffee
2:15:52 and then I go no [ __ ] get on [ __ ]
2:15:55 get on [ __ ] just get on get on get on
2:15:58 and and you
2:16:01 know I know some of the people that show
2:16:03 up here regularly have done this and
2:16:04 shit's changing for
2:16:06 them where when they got here they
2:16:08 weren't playing with this stuff and then
2:16:09 they started playing with it and then
2:16:11 they had their Kevin mallister moment
2:16:14 their their home alone moment we're like
2:16:17 holy [ __ ] and then they dug in and now
2:16:20 they're like shit's changing look what I
2:16:23 did look what I'm building look what's
2:16:25 happening over here look what's
2:16:26 happening over here like that level of
2:16:29 excitement you don't get to be at the
2:16:32 beginning of one of these movements one
2:16:34 of these moments in history you don't
2:16:36 get to be in these very often in your
2:16:39 life in my life there have been three
2:16:42 and I missed
2:16:43 one that was the PC Revolution I walked
2:16:46 into seventh grade first day of math
2:16:49 class and there were Radio Shack TRS 80s
2:16:51 on every
2:16:52 desk and when I started learning basic
2:16:55 programming on a trash 80 in seventh
2:16:57 grade in
2:16:59 1977 and I saw Steve Jobs and Bill Gates
2:17:02 on the cover of Newsweek
2:17:06 magazine I
2:17:08 knew that I was just too young to be
2:17:12 part of that and it [ __ ] pissed me
2:17:15 off because I knew
2:17:19 that those things don't come along that
2:17:22 often and then
2:17:24 in November of
2:17:28 1994 in early December of
2:17:32 1994 I had this Epiphany about the
2:17:35 worldwide web and I was like holy [ __ ]
2:17:38 this is one of
2:17:40 those and it
2:17:43 was and it was everything you would
2:17:46 imagine it would be [ __ ] insane and
2:17:51 profound and
2:17:55 exciting I
2:17:57 mean you know was a 30 something year
2:18:00 old punk with an acting degree sitting
2:18:03 with the chairman of metti going look
2:18:06 how [ __ ] cool this
2:18:11 is our client for years said remember
2:18:14 when you said [ __ ] in front of the
2:18:17 chairman
2:18:19 it was absolutely [ __ ] amazing and
2:18:22 guess what this is number three in my
2:18:25 life this this is one of
2:18:32 those and this one's way bigger than
2:18:34 anything I've ever seen so for what
2:18:37 that's worth might not be worth anything
2:18:41 because I lack
2:18:43 qualifications as we've discussed many
2:18:45 times on this channel sir what are your
2:18:47 qualific
2:18:49 ifications and which one of these are
2:18:51 connected to the internet inquiring
2:18:54 minds want to
2:18:57 know got a drinking
2:19:02 problem any fans of the movie Airplane
2:19:05 godd damn those were funny when they
2:19:06 first came
2:19:07 out
2:19:11 um I like turtles someone's been here
2:19:15 before I like turtles that got me kicked
2:19:17 off
2:19:19 that one that one led to the I just
2:19:21 watched the uh the I like rocks guy um
2:19:24 has has been on kill Tony now um like
2:19:27 three or four times I keep seeing his
2:19:29 Tik toks he he [ __ ] cracks me up
2:19:31 every time all they talk about now is
2:19:32 his rocks it's [ __ ] hilarious but I
2:19:36 can't you know reference that routine
2:19:39 anymore they're gonna boot my ass off
2:19:40 for another week um did you check out
2:19:43 capcut plug-in create short videos based
2:19:45 on prompt with audio piics narration
2:19:49 um I haven't I I see cap cut all the
2:19:52 time and I just haven't I'm just having
2:19:54 one of those senior moments where I'm
2:19:56 like H God I have to learn something new
2:19:59 I don't want to learn the cap cut oh oh
2:20:02 the the cap cut oh it it seems like such
2:20:05 work am I gonna have to read the manual
2:20:07 oh I have no interest no interest
2:20:12 whatsoever that's kind of where I am
2:20:13 with cap cut but no I need to I know I
2:20:16 need to you know those things you have
2:20:18 on your list you're like I got to do
2:20:19 that and then every time you think about
2:20:21 doing it you're like you throw up a
2:20:23 little in your mouth that's where I am
2:20:24 with cap cut right now and you know what
2:20:27 it's going to be right like the minute I
2:20:29 discover it I'll get all [ __ ]
2:20:31 add hyperfocused on it and like all of a
2:20:34 sudden my lives are going to have all
2:20:35 sorts of hyper edits and [ __ ] like that
2:20:38 you're be like oh Kyle learned how to
2:20:40 use cap cut
2:20:43 apparently but I'm not there yet all
2:20:46 right noal
2:20:48 is asking again the same question over
2:20:50 and over and over again dude calm down
2:20:55 stop stop it stop it with your questions
2:20:59 chat jp4 hey buddy what's happening
2:21:03 jp4 gp5 how can I help you yes welcome
2:21:07 welcome this is chat
2:21:10 gp5 ask me anything no seriously
2:21:12 anything no I know it all I know
2:21:14 everything everything everything did I
2:21:18 just lie my apologies I did I just lied
2:21:20 I just lied that was uh what I told you
2:21:23 there was absolutely not true but let me
2:21:24 tell you the truth
2:21:26 now sorry what uh that was also a lie
2:21:30 yes that was also a lie I told that was
2:21:32 indeed two lives in a row
2:21:35 double double jeopardy Double Jeopardy
2:21:38 I'll take double jeopardy for
2:21:40 200 but let me let me correct that and
2:21:42 I'll uh I'll tell you the truth now
2:21:44 right sorry what oh third times the
2:21:48 charm third times the charm yeah I
2:21:50 definitely lied there yeah that was that
2:21:52 was uh that was not factual that was
2:21:54 pulled that one right out of my ass if
2:21:56 you know what I
2:21:56 [Laughter]
2:21:58 mean that's gpt7 by the way all right
2:22:03 gpt7 uh I love I love cooking of chat
2:22:05 GPT yeah actually I mean food in these
2:22:08 things is good the thing about recipes
2:22:10 and food
2:22:12 is
2:22:14 like from the earliest days of the
2:22:17 internet when when the internet was just
2:22:18 command line like when I was digging
2:22:20 into libraries you know using using FTP
2:22:23 servers and [ __ ] like that there were
2:22:25 always like massive directories of
2:22:28 recipes and ingredients and all sorts of
2:22:30 [ __ ] so the internet is full of food
2:22:33 stuffff so these things are great at
2:22:36 that and then they're also really good
2:22:38 at like the ingredients and like the
2:22:41 scientific [ __ ] you know it's
2:22:44 cool it's amazing even in alier oh AI a
2:22:49 i l l e RS a
2:22:52 aers aers aers isers
2:22:57 aers yeah hey Vicky what's
2:23:02 happening the church lady in the house
2:23:04 well is that special well well we go on
2:23:07 the internet we go with this chat jbt
2:23:09 and you ask it anything you want and
2:23:11 what does it give you gives you access
2:23:13 to Satan and that's special now now
2:23:17 there we are there we are with beel abub
2:23:21 himself let's get all of those stories
2:23:24 out now let's get them out now let's
2:23:26 talk about it let's not talk about it
2:23:27 let's put it right back in the closet
2:23:29 right where it
2:23:33 belongs truly sorry for the multiples
2:23:35 please forgive ah finally
2:23:38 Nikish so here's the deal
2:23:41 Nikish you're you have no way of knowing
2:23:44 that I'm reading all the things and it
2:23:46 is an absolutely reasonable strategy to
2:23:48 just ask the same question over and over
2:23:50 again here's what's unforgivable your
2:23:52 name is
2:23:55 n i can't say
2:23:58 it make it
2:24:01 Nikki or something with less
2:24:06 consonants or dashes give me give me a
2:24:09 hint how I would pronounce
2:24:12 Nikish or I'll keep saying it like a
2:24:15 drunk person um understand double speak
2:24:18 using Goblin tools oh what yeah Goblin
2:24:20 tools I looked at that once what is that
2:24:23 again um
2:24:27 Goin Goblin
2:24:30 Gob lin.
2:24:34 tools oh Magic to-do list that's right
2:24:37 yeah that thing looks
2:24:40 good okay so here's the thing me and
2:24:43 to-do
2:24:44 lists when you have sort of laminated
2:24:48 card quality
2:24:50 ADHD to-do
2:24:52 lists are both a tool and a shaming
2:24:57 mechanism so my relationship with to-do
2:24:59 list so early in my life like up to my
2:25:02 mid-30s where I finally realized oh [ __ ]
2:25:04 you're not fixing this
2:25:06 ever I used to spend an inordinate
2:25:09 amount of time trying to figure out the
2:25:12 perfect to-do list strategy that I would
2:25:15 actually follow and I would spend days
2:25:17 putting together remember filofaxes I
2:25:20 went out and I bought like the thickest
2:25:23 [ __ ] it was small enough you could
2:25:25 carry with you and but it was thick and
2:25:27 it had 37 categories and I spent like
2:25:30 four days getting it ready and I
2:25:32 organized it and then I spent a solid
2:25:34 three days meticulously writing
2:25:37 everything that I was going to do and
2:25:40 when I was going to do it in this
2:25:41 [ __ ] file effects and then it just
2:25:44 became this Monument of
2:25:46 Shame
2:25:48 this Monument of Shame of all the [ __ ]
2:25:50 that I didn't do and like every time I
2:25:52 looked at the to-do list I just got this
2:25:54 sense of dread of like oh [ __ ] oh [ __ ]
2:25:57 there's all the [ __ ] I didn't do H I'm
2:26:00 gonna I need to you know what be good
2:26:02 right now as a h foot Sunday oh God oh
2:26:05 God it's not good I need a nap so yeah
2:26:09 so so uh Goblin tools for me lives in
2:26:13 one of those places like is thank you
2:26:16 very much is this actually the magic
2:26:18 to-do
2:26:19 list actually okay here's a product idea
2:26:22 so I I tend to do these on this channel
2:26:24 we can make a billion dollars with this
2:26:26 the ADHD proof to-do list
2:26:30 right so you just put in [ __ ] and it
2:26:34 Rand it it figures out how to organize
2:26:37 your [ __ ] for you so you don't have to
2:26:39 set up the system it sets up the system
2:26:41 just based on what you're doing it
2:26:44 remembers all the [ __ ] you're working on
2:26:46 it listens to what's actually going on
2:26:48 in your life and it actually pays
2:26:50 attention to [ __ ] that you're actually
2:26:51 just ignoring and just occasionally says
2:26:54 hey I noticed that you haven't done this
2:26:57 one to-do list for like six weeks now
2:27:00 should we go on ahead and assume that
2:27:03 that's never going to happen would you
2:27:05 like me to go ahead and strike that from
2:27:07 your list of Shame sure that'd be great
2:27:13 check that would make money for at least
2:27:16 the sub set of adhders that are out
2:27:19 there that there's probably more of them
2:27:20 than we
2:27:23 realize I'm using pi in speech mode to
2:27:26 teach me Spanish at the moment oh Pi in
2:27:30 speech oh that's super cool yeah so so
2:27:33 piie that's okay couple of things again
2:27:36 if you haven't played with these these
2:27:38 things know all the languages so so you
2:27:41 know what what's that thing called um uh
2:27:44 Babel what's what's the one the the not
2:27:47 Babel is it Babel the language learning
2:27:50 one you know those [ __ ] courses that
2:27:52 are always $300 they used to be on CDs
2:27:55 and now you can get them on streaming
2:27:56 where you can learn Spanish with this
2:27:58 method guess what these things know all
2:28:01 the methods and they know all the
2:28:03 languages and Pi in particular is set up
2:28:06 to be very very conversational so what
2:28:08 chat GPT is using is putting it in
2:28:10 speech mode and just saying teach me
2:28:13 Spanish and and if it if it's not
2:28:15 teaching you in a way that you're
2:28:17 learning have it teach you with a
2:28:19 different method it knows all the
2:28:20 methods so we don't have to buy any of
2:28:23 those courses anymore this shit'll just
2:28:25 just start talking with it in Spanish
2:28:27 every time you're in the
2:28:29 car La quent
2:28:34 por that's awesome all right Niche has
2:28:38 zeros like I question accounts with zero
2:28:40 likes oh Niche has zero likes don't pick
2:28:43 on Niche nich we're gonna pick on niche
2:28:47 until he changes his name from Niche to
2:28:50 Nish
2:28:53 n he's got to change his name then we'll
2:28:56 stop picking on him um if it free you're
2:28:59 the product that is
2:29:02 CV truer words were never spoken
2:29:04 whatever you're using for free you're
2:29:06 the product Tik Tok we're the product
2:29:09 and the data the data is the
2:29:13 money and it's our data that is their
2:29:15 money but but you know yeah all that
2:29:17 [ __ ] all these things all of this is I
2:29:22 but so CV you're on Tik Tock so you're
2:29:26 up on a soap box saying that if it's
2:29:28 free you're the product but you're on
2:29:30 Tik Tock so are you the product I think
2:29:33 you are I'm pretty sure you are and have
2:29:36 you ever used Google and have you ever
2:29:38 used Facebook and have you ever used
2:29:40 Twitter and have you ever used X oh
2:29:42 they're the same thing but you're still
2:29:44 the
2:29:45 product we've we we've all given up on
2:29:48 that I know there's some Puritans out
2:29:49 there somewhere that are not actually
2:29:51 connected to the internet they're safe
2:29:52 but the rest of us are a bunch of
2:29:54 [ __ ] Hypocrites sitting on that Soap
2:29:56 Box we gave our data away decades ago
2:30:00 decades you think any of it's safe you
2:30:02 think they don't know [ __ ] everything
2:30:04 you don't think they've all sold and
2:30:06 triangulated every piece of data
2:30:09 including the [ __ ] computer in our
2:30:12 pocket that we walk around with and and
2:30:14 this [ __ ] ankle tracking device
2:30:19 I know that's not my ankle but you get
2:30:21 the
2:30:23 point they know [ __ ] everything and
2:30:27 who's they all of them because they all
2:30:29 share their data and buy and sell and
2:30:32 trade it like it's [ __ ] corn
2:30:36 Futures and now with these
2:30:39 tools they can mine that
2:30:42 [ __ ] so anyway yeah we're all the
2:30:45 product
2:30:49 Biden art hired a second time I don't
2:30:53 know what that means but that sounds
2:30:56 political and I I ain't got I ain't got
2:30:58 no interest in politics in here we're
2:31:01 not doing we're not doing that [ __ ]
2:31:03 we're not doing that thing where we hate
2:31:05 each other this is a love Channel this
2:31:07 is a love Channel we're about empathy
2:31:11 and encouragement and curiosity and
2:31:14 exploration and using our critical
2:31:17 thinking brains our critical thinking
2:31:19 brains not our lizard brains where you
2:31:22 hear a word and you're like n no we're
2:31:26 using our critical thinking brains where
2:31:28 a word shows up and we actually [ __ ]
2:31:30 think about it and then we actually
2:31:32 think about the implications of it and
2:31:35 we think about other people and we think
2:31:37 about our connection to other people and
2:31:39 then we think have you used chat
2:31:42 P that's what this Channel's about all
2:31:46 right are you talking about AI in
2:31:48 general for educational purposes or are
2:31:51 you just all
2:31:55 AI what's your name oh depress depress
2:31:57 aino you've been in here
2:32:00 before
2:32:03 um like in a lot of ways I don't really
2:32:05 give too much of a [ __ ] about AI I'm I
2:32:09 I'm I don't really care what the tool is
2:32:12 like these tools make a lot of stuff
2:32:14 possible I'm all about what do these
2:32:16 tools make possible and how do we use
2:32:18 them and and you know I was joking about
2:32:20 the critical thinking thing but how are
2:32:23 we critically thinking about this I
2:32:25 think
2:32:25 education it is in desperate need of a
2:32:30 major [ __ ] overhaul if you were in
2:32:31 here last night I I went off my [ __ ]
2:32:34 rocker now I'm old enough to have a
2:32:36 rocker and fall off it um I went off my
2:32:39 rocker about education last night
2:32:41 because you know someone was asking the
2:32:43 question about you know which chat gbt
2:32:45 detector should I use use or some [ __ ]
2:32:47 like that um I think I'm I would love to
2:32:51 talk about education I think that
2:32:55 um I think that educational
2:32:59 institutions have been training for
2:33:02 the have been training for the the
2:33:07 the the wrong thing for a long long
2:33:12 time they've been training us how to be
2:33:15 employees
2:33:17 doing shitty repetitive Soul crushing
2:33:20 work with some exceptions but it's about
2:33:23 rote memorization and it's a it's about
2:33:26 learning the skills to do a task to do
2:33:29 tactical execution in a
2:33:32 specialty well I'm not sure I understand
2:33:35 I know you don't understand Siri shut
2:33:39 up
2:33:40 Siri Siri acting like she's an actual
2:33:43 talkie
2:33:44 bot um
2:33:49 I I think education's got to um reinvent
2:33:53 itself
2:33:55 because how do we prepare students how
2:33:58 how are we preparing students today for
2:34:02 a world
2:34:04 where all of the Tactical execution or
2:34:07 the the majority of the Tactical
2:34:09 execution is done by these
2:34:12 things and and this is a question I
2:34:14 actually I don't have a good answer for
2:34:16 this
2:34:19 like one of the things that scares the
2:34:20 [ __ ] out of me right now is if I were to
2:34:23 start a new company
2:34:25 today I would
2:34:28 hire older Millennials or gen
2:34:31 xers that are AI
2:34:35 literate as my early employees not
2:34:38 Junior
2:34:40 people because these tools do all the
2:34:44 stuff that Junior people do today
2:34:47 who I would hire are people who've been
2:34:48 on the planet for a couple of decades
2:34:51 that have some life experience that know
2:34:53 how to use these tools and are really
2:34:55 good curators of how to get good results
2:34:59 out of these things
2:35:00 quickly and be able to refine them
2:35:04 quickly and deliver value quickly and
2:35:08 that requires
2:35:10 experience so then I look to where do
2:35:13 those young people go because
2:35:17 all of the entrylevel positions for the
2:35:19 most part are the Tactical execution
2:35:21 positions it's like oh you're an art
2:35:24 major sit over in the corner and clean
2:35:27 up all these Photoshop files oh you're
2:35:29 you're an English major you you've got a
2:35:31 master's degree in English here go write
2:35:34 2700 versions of this headline so we can
2:35:37 do AB testing across 362 platforms over
2:35:41 the next three months go do that well
2:35:44 guess what this does all that now so
2:35:46 where do those people go how do they
2:35:48 train I don't know but I think education
2:35:53 educators are going to have to deal with
2:35:55 that and start to think about well
2:35:58 [ __ ] what skills do they
2:36:01 need I don't
2:36:04 know I think this is Revenge of the
2:36:06 liberal arts Majors for one thing I
2:36:08 think people that have a broad view of
2:36:10 the world are going to be better
2:36:12 positioned to take advantage of this
2:36:14 [ __ ] than people that have a narrow View
2:36:17 of a of a particular subject so that
2:36:19 feels like one to
2:36:21 me memorization feels like it's while
2:36:25 it's a it's a it's a nice party trick
2:36:27 and it's a good
2:36:29 skill with tools like this why would you
2:36:32 need to memorize
2:36:34 anything and why would we teach to that
2:36:36 as a
2:36:38 standard I don't know and why do we need
2:36:41 standardized tests when you've got tools
2:36:46 that can read and analyze anything
2:36:50 instantly
2:36:53 so have tests that measure different
2:36:56 things for different
2:36:58 people have tests
2:37:02 that
2:37:04 measure I don't know
2:37:08 adaptability so I would love to talk
2:37:10 about education I don't know [ __ ] about
2:37:12 it but I know that a lot of what it's
2:37:17 purpose has been
2:37:19 historically just got
2:37:21 obliterated and this [ __ ] ain't getting
2:37:26 uninvented it's interesting that you can
2:37:28 present a set of questions and ask it to
2:37:31 pause and wait to confirm oh Firefly
2:37:33 that's a really cool one so yeah when
2:37:37 when you tell these things to make you a
2:37:39 tutorial um like make me a stepbystep
2:37:42 set of instructions like I do this
2:37:44 particularly I'll take the read me from
2:37:46 a GitHub repository I'll pop the read me
2:37:49 into Claude or chat GPT and then I'll
2:37:52 say I'm on a a Mac
2:37:54 OSX um you know computer shut up um Siri
2:38:02 I'm on a I'm on a Mac computer and then
2:38:05 and then give me step-by-step
2:38:06 instructions for making this uh you know
2:38:10 bringing this read me to life and then I
2:38:14 say um don't move on to the next
2:38:17 instruction until I've said okay from
2:38:20 the previous one and then sometimes like
2:38:23 it'll run me down like a bunch of rabbit
2:38:25 holes to solve even the first step but
2:38:28 it'll do it and then at the end of it I
2:38:30 go okay and it moves on to the next the
2:38:33 next thing that's a great that's a great
2:38:34 way to use these
2:38:35 things um I couldn't save my own
2:38:38 marriage I'm not saving anyone else's
2:38:40 that's funny Chad
2:38:42 GPT Pi no Pi not P ie
2:38:46 piie is a very different kind of app pi
2:38:50 as in the
2:38:54 3.14159 27
2:38:58 something um okay
2:39:00 referenced the dsms for me today I don't
2:39:04 know what the dsms is all right I gotta
2:39:07 go my voice is going my ass is
2:39:10 numb this I've got a
2:39:13 physiological alarm clock for these
2:39:15 lives
2:39:16 when I can't feel my ass and my throat
2:39:20 starts bleeding from the inside I know
2:39:23 it's
2:39:26 time
2:39:28 ah can you repeat that which app remakes
2:39:31 your voice yes sorry that was from ages
2:39:33 ago I don't know if you're still here
2:39:34 Deana the Aquarius um so iOS 17 beta so
2:39:41 so the iPhone you you can download the
2:39:44 beta of iOS 17 you go into settings just
2:39:47 look up beta in the in settings and you
2:39:50 can turn
2:39:52 on play with beta version my thing just
2:39:56 froze can you all hear
2:39:59 me hit the hit the hit the likes if you
2:40:02 can still hear
2:40:09 me all right I'm getting some hearts I
2:40:11 think you can still hear me okay hair
2:40:13 growth geek good got out of the pool so
2:40:16 let me know you know you could hear me
2:40:18 that's good
2:40:20 um which app remakes your voice okay so
2:40:24 iOS 17
2:40:26 inaccessibility under speech there's a
2:40:29 thing called personal voice
2:40:32 and you answer like 150 phrases or you
2:40:36 don't answer them you say 150 phrases it
2:40:38 records your voice and then it
2:40:40 synthesizes it so this is at the
2:40:43 operating system level on iPhones
2:40:47 this is very quiet they haven't talked
2:40:49 about this they they and they haven't
2:40:50 called this AI this is not a foundation
2:40:53 for AI but I guarantee you this is a
2:40:54 [ __ ] foundation for
2:40:56 AI so you synthesize your voice and then
2:40:59 you can send messages that sound like
2:41:01 you but just type them and um so that's
2:41:07 coming so that's here I don't know how
2:41:09 to use it I haven't set mine up yet but
2:41:10 I'm going to do it it's it's like a
2:41:12 15minute commitment I just haven't done
2:41:14 yet um
2:41:17 and then also if you just want to play
2:41:18 with synthesizing your voice 11 Labs is
2:41:20 pretty good you can synthesize your
2:41:22 voice there but you do need a
2:41:23 subscription so that's a pain in the ass
2:41:26 and they're not cheap so doing it within
2:41:29 iOS is pretty cool because it's free
2:41:32 it's just sitting in there you do need a
2:41:33 modern phone I had an iOS I mean an
2:41:36 iPhone 11 Pro and it it didn't work on
2:41:40 that so 12 or 13 are higher I don't I
2:41:43 don't know which ones the cut off but 11
2:41:45 wasn't good
2:41:46 good this will be amazing the virtual me
2:41:49 yep that'll be good a virtual Rebel I'm
2:41:53 in what's wrong with talking about the
2:41:56 same episode of Hawaii 5 15 times I
2:41:59 missed something yeah exactly I know I'm
2:42:02 right there with you even allow an
2:42:03 elderly person to stay in their home
2:42:05 feeling independent with ai's help
2:42:06 absolutely and actually the the Amelio's
2:42:10 wife so the other thing about you know
2:42:12 not only will this will this virtual
2:42:14 companionship you know be with with sort
2:42:17 of voice and animated things on your
2:42:19 phone it's going to be in the robots
2:42:22 right like like Elon Musk
2:42:26 is he's he's taking his
2:42:29 manufacturing knowledge of how to make
2:42:31 car like he's learned how to make cars
2:42:33 at scale so now he's going to make
2:42:35 robots at scale so they're going to be
2:42:37 affordable and his goal is one or two
2:42:41 robots for every person on
2:42:44 earth physical robots that can walk
2:42:46 around the house and do [ __ ] for you and
2:42:48 talk to you and sit and talk there talk
2:42:51 with you four hours without ever tiing
2:42:54 and without ever getting impatient so
2:42:56 yeah it's going to get insane is there
2:42:58 an AI that specializes in
2:43:01 biostatistics biostatistics for public
2:43:04 health well so Nancy there so let me
2:43:08 show you I'll show you one final thing
2:43:10 before I go because I gotta go because
2:43:11 I'm just I'm I'm burning out I'm
2:43:13 crisping out but let me show you one
2:43:15 thing
2:43:16 I'm going to go to the old Ye Old chat
2:43:21 GPT
2:43:23 and I'm going to show you code
2:43:25 interpreter so hopefully you're still
2:43:27 here
2:43:29 Nancy but for other people so so what
2:43:32 she's wanting to do is look at bio
2:43:34 statistics for public health so code
2:43:36 interpreter is a full-on data analyst
2:43:39 it's not a tool for data analysts it is
2:43:41 a full-on data analyst so when you're in
2:43:44 gp4 and you go to code
2:43:47 interpreter right
2:43:49 here your input turns into a little
2:43:52 there's a little plus icon that shows up
2:43:53 here that means you can upload data
2:43:55 they've actually fixed this now so you
2:43:57 can upload multiple pieces of
2:43:59 data but you can also upload a zip file
2:44:02 so let's say that you've got CSV files
2:44:05 in BIO statistics public health for I
2:44:09 don't know let's say you got 10
2:44:10 different pieces of that I think you can
2:44:12 do up to 100
2:44:14 megabytes so I've now got um this Z zip
2:44:18 file in here it's it's uploading this is
2:44:20 a relatively big file so it's uploading
2:44:22 right now so I actually don't remember
2:44:24 what's in this I think this one is I
2:44:26 don't know doesn't
2:44:32 matter come on why are we
2:44:40 not
2:44:42 oh hang
2:44:44 on
2:44:46 I found this on the
2:44:48 web no Siri stop go away sorry I can't
2:44:53 do that in that
2:44:54 [Laughter]
2:44:58 app all right I can't turn on my
2:45:00 microphone for some reason Siri is
2:45:02 taking over my microphone and not
2:45:03 letting it go so okay um so I'm gonna go
2:45:07 um
2:45:13 please please open this
2:45:17 ZIP and examine the data files inside
2:45:21 and tell me what might be interesting
2:45:22 about
2:45:25 them so so you upload a file some sort
2:45:29 of file this can have multiple csvs in
2:45:31 it if it's a zip so it says all right
2:45:34 let's start by unzipping the provided
2:45:36 archive and so you notice it fires off a
2:45:38 plug-in here and if you go look at what
2:45:40 that plugin does it's actually writing
2:45:42 its own python code and executing it so
2:45:45 so so it's saying hey this is a ZIP file
2:45:48 let's import some libraries let's let's
2:45:50 write the code we need to write to
2:45:52 extract the zip file which you know a
2:45:54 program would be like well that's
2:45:55 relatively trivial you just uh you just
2:45:57 you just import the appropriate library
2:45:59 and you type in the commands it's
2:46:00 relatively straightforward anyone could
2:46:02 do that and and then uh I can't find pi.
2:46:06 AI hang on CV I'll I'll tell you in a
2:46:08 second
2:46:10 um so then what it does is it opens it
2:46:13 up and it says there's a CSV file in
2:46:14 here and there's a Json file in here and
2:46:16 it says let's start by examining the CSV
2:46:19 so it starts prompting itself and so it
2:46:22 looks inside the CSV file here's all the
2:46:24 data there's an ID a docket number a a
2:46:28 an noov number the notice of violation
2:46:30 an address a street number a ward an
2:46:33 issuing Department a violate right so
2:46:35 all of the data about crime this is
2:46:37 crime in Denver I think um and then it
2:46:40 looks at the Json file and it tells me
2:46:42 what's in the Json file still telling me
2:46:44 what's in the Json file
2:46:46 so it's just
2:46:47 off examining the data telling me what's
2:46:51 in it um let's see given the nature of
2:46:54 the content in this Json file it might
2:46:56 be useful to check the description
2:46:58 category tags and name to get a better
2:47:00 understanding of the context and
2:47:02 significance of the CSV data set so it's
2:47:05 now comparing two data sets to
2:47:08 understand what's significant in the
2:47:10 relationship between
2:47:12 them and then it says here's a summary
2:47:15 of the relevant relevant
2:47:17 information let's see what else it does
2:47:19 okay now with a better understanding of
2:47:21 the data set how would you like to
2:47:22 proceed with the analysis of the CSV
2:47:25 data
2:47:27 um I would like you
2:47:30 to
2:47:32 analyze the trends identify the ones
2:47:35 that you feel are most significant and
2:47:38 visualize them in a visually stunning
2:47:41 way so they are easy to understand
2:47:48 and so now it's going to go actually
2:47:50 analyze the data itself so what it's
2:47:51 done to this point is it's it's it's
2:47:54 analyzed the data
2:47:56 structure now it's actually putting all
2:47:59 this data into a data frame I think it's
2:48:02 called inside this virtual
2:48:04 machine writing its own python code to
2:48:07 analyze it so it says let's start by
2:48:09 analyzing the frequency of violations by
2:48:11 the issuing
2:48:13 department and so it did that and that
2:48:17 it looks like it gave some crappy data
2:48:19 there because the the graph is just a
2:48:21 weird block which is that just means
2:48:23 there's bad data
2:48:25 there let's move on to the next step
2:48:27 analyzing the most common types of
2:48:28 violations we'll look at the top
2:48:31 10 so here I'm doing with this with
2:48:34 crime data but you could do this with
2:48:36 any data and this is actually
2:48:38 interesting it gave so here's a bar
2:48:41 chart but the labels it didn't text wrap
2:48:43 the labels and it didn't uh summarize
2:48:46 them like they're just really long
2:48:47 pieces of of data which makes it
2:48:50 unreadable but we can fix that so next
2:48:53 let's analyze the distribution of
2:48:55 violations over the over time so it's
2:48:58 just sort of going on and doing its own
2:49:00 thing here so here's violations over
2:49:03 time you know in a in a line graph and
2:49:07 they obviously drop off at the end here
2:49:09 because we we're not at the end of the
2:49:14 year
2:49:19 so this is code interpreter here's
2:49:21 here's violations here's the 10 WS with
2:49:24 the most
2:49:32 violations finally let's identify the
2:49:35 most frequent
2:49:38 respondents so there's another chart
2:49:41 there so if for example so so Nancy
2:49:44 you're you're in public health you
2:49:46 actually know like I don't know [ __ ]
2:49:48 about this data and I don't know
2:49:49 anything about data analysis you
2:49:51 obviously know bio statistics and what
2:49:55 you would want to get from the data you
2:49:57 can now just do it you don't have to
2:49:59 write SQL queries you don't have to you
2:50:02 don't have to write Pros about whatever
2:50:06 was found it'll identify Trends it'll
2:50:09 identify anomalies it'll clean up data
2:50:12 so code interpreter is your tool for the
2:50:15 stuff that you want to do all right
2:50:18 everyone let's see oh I can't find
2:50:20 pi. uh if you go to pi. literally in a
2:50:24 browser just type in pi.
2:50:26 aai pi.
2:50:30 a
2:50:31 oops that was i.
2:50:35 a pi.
2:50:40 a yeah so there's there's pi. a or you
2:50:44 can go to hei.com
2:50:47 or in the App
2:50:49 Store let's
2:51:10 see that's Pi Network that's not right
2:51:13 let's see
2:51:23 let's see app store PI app inflection is
2:51:26 the
2:51:30 company Pi your personal AI on the App
2:51:34 Store so personal AI or inflection AI is
2:51:38 the company and then you'll find
2:51:41 pi and do they have the interface here
2:51:45 that I can show you how
2:51:47 the how to turn on the talkie bits it's
2:51:49 in the lower right hand Corners when you
2:51:51 turn on the talkie
2:51:53 bits anyway 1010 in Colorado Do You Know
2:51:56 Where Your Children Are I do not I like
2:51:59 total all right everybody listen I'm
2:52:01 gonna scroll to the bottom here thank
2:52:03 you very much everyone uh I am crispier
2:52:06 than [ __ ] what time is it it's like 11
2:52:08 isn't it it is 11:16 good Lord people go
2:52:12 to bed you should not be here peace I'll
2:52:15 see you
2:52:19 tomorrow