aiLL June 4, 2023 - The Democratization of AI

Description

Chapters

Transcript

0:13 mm-hmm
0:31 hey hey it's Kyle Shannon AI learning
0:34 lab hang on
0:36 putting on the Nuggets game on my iPad
0:37 off to the side so I can pay attention
0:40 to to that I'm in Denver Colorado
0:42 watching the uh the Nuggets
0:45 um this is the AI learning lab I was
0:46 live a little bit earlier I had a little
0:48 time to kill so I figured I'd pop back
0:50 on for a bit more
0:52 um
0:52 this channel is all about learning about
0:55 AI I'm learning about AI as I'm learning
0:58 about it I'm sharing what I know so if
0:59 you have questions about AI just pop
1:01 them in the uh
1:03 in the what do they call them the
1:05 comments and uh share this live if you
1:08 think other people would benefit from
1:10 this a little bit about my background
1:13 thoughts on AWS sagemaker is that new is
1:19 that the same thing that AWS announced a
1:22 few weeks back hang on let me go look
1:27 AWS Age Maker
1:41 when was this announced
1:44 today
1:49 getting started with sagemaker
1:51 um
1:53 I think that uh I I don't know I don't
1:55 know enough about sagemaker to give
1:57 anything
1:58 um specific about that but I think in
2:00 general think you know the cloud-based
2:02 service is making these large language
2:04 models accessible to businesses really
2:07 really easily is is going to be it's
2:10 gonna essentially be a requirement if
2:11 you're doing anything in cloud services
2:13 you're going to have to have these kind
2:15 of foundational models available I think
2:17 one of the biggest things that you know
2:20 even in in my business I've got a small
2:22 video technology business
2:25 um the ability for me to leverage my own
2:28 data in all sorts of powerful ways if
2:31 these uh cloud services can make that
2:33 really really easy then I think it's a
2:35 good thing
2:36 um I think AWS is a little behind the
2:38 ball
2:39 um Google's Google's a little behind
2:41 right now and uh openai is screwed some
2:45 people up
2:49 okay uh if you have questions about AI
2:52 oh there's lots of people here so great
2:54 this is my name is Kyle Shannon this is
2:55 the AI learning lab
2:57 um if you're new to generative AI if
2:59 you've not played with chat GPT I want
3:01 you to do that I want you to do that
3:03 while you listen to me ramble so these
3:06 uh the reason I have these URLs up in
3:09 this little window here and look I'll
3:11 even make them bigger boom
3:14 um the top one is uh Chachi BT's
3:17 official website chat.openai.com Bing is
3:20 Microsoft search engine but there's a
3:22 chat button there now if you click that
3:24 that's gpt4 and you can use it for free
3:26 and then po.com lets you use their own
3:30 cert their their own large language
3:32 model anthropics Claude different
3:34 versions and chat CPT 3.5 and 4.
3:39 um
3:39 some of the models you have to have paid
3:42 access to use but pose really cool
3:44 because one of the models that you can
3:46 use is Claude 100k
3:49 and Cloud 100K is a hundred thousand
3:52 token chat gbt basically where you can
3:55 paste in up to 75 000 words into your
3:58 prompt so for example you could take a
4:02 transcription from a three-hour meeting
4:04 and just paste it in there
4:07 um you could take your great American
4:10 novel and dump it in there and ask it
4:13 for notes
4:15 have it do a rewrite
4:18 if you uh if you have questions about AI
4:22 pop them in I will do my best to get
4:24 through and and answer them if I can't
4:27 answer them I will tell you uh I'm a
4:30 smart ass I'm a gen xer I'm a long time
4:33 entrepreneur
4:34 and uh yeah and I like to I try to have
4:37 fun here try to keep it keep it light
4:39 keep it alive what I want to get too
4:41 heavy we don't want to get too heavy
4:43 people
4:44 uh
4:45 um can we invest in what in you can
4:49 invest in Microsoft for Bing and for
4:51 what they're doing getting under Google
4:53 skin
4:55 um I don't think you can invest in open
4:56 AI right now they're not public
5:00 um
5:01 Nvidia just hit a trillion dollar market
5:04 cap after after doubling the previous
5:06 year I heard someone say that SEO is
5:09 dead because of AI I'm trying out some
5:11 and wondering what you think
5:16 I think SEO is going to be forever
5:19 changed I don't think it's going to be
5:21 dead but I think there will be
5:25 um
5:29 llmo large language model optimization
5:33 um how do you end up showing up in in
5:36 results when someone asks about your
5:38 subject is similar to SEO but with SEO
5:41 you can have all these sort of you know
5:43 fake and artificially padded sites and
5:46 things like that the the I feel like the
5:50 SEO game got really toxic and the reason
5:52 it got toxic is that Google had 92 or 94
5:56 market share when chat GPT came out
6:00 and they essentially it was effectively
6:02 a monopoly so
6:04 um so everyone sort of figured out how
6:07 to game the system and you know they
6:08 would change the algorithm and change
6:10 things up but for the most part
6:11 everyone was doing the same thing when
6:13 it came to that kind of marketing I
6:15 think that as more and more people get
6:18 used to getting answers from these
6:20 engines rather than getting a list of
6:21 options that many of which are paid for
6:24 it dramatically changes that game and in
6:27 fact you know Google announced that
6:29 they've got a new
6:30 search result that includes the answer
6:32 to your question as well as links
6:35 um if it gives me the answer that I need
6:37 I don't need to go look at those links
6:38 right so I think it I think search
6:41 engine optimization has changed uh for
6:43 forever it's not going to be apparent in
6:47 the next six months to a year but as
6:50 usage of large language models goes up
6:53 and search goes down and maybe there's
6:56 some shifting and balancing from Google
6:58 maybe to Microsoft maybe to other
7:00 players
7:01 I think you're going to see The Rules
7:04 start to change and and what's required
7:06 start to change
7:08 so hope that helps you think
7:12 you think it's gonna be well e reality
7:15 soon
7:16 um yeah I don't really remember Wally
7:19 well enough to know but but
7:23 robots walking around helping us uh I
7:26 don't know enough about that movie oh
7:27 you know what we can do
7:31 watch this
7:34 over to the chat JPay Thai
7:39 yeah okay
7:42 what's going on
7:44 come on
7:45 okay so we're gonna do a new chat I
7:48 don't think we need to connect it to the
7:50 internet oh yeah we do because we okay
7:52 so so here's what we're gonna do we're
7:54 doing gpt4
7:56 and we're gonna say
7:58 um
7:59 what are some of the
8:04 um
8:05 sci-fi
8:06 elements of the movie
8:10 Wall E
8:17 um
8:19 that are related to generative AI
8:26 create a table
8:30 of the element
8:33 in one column
8:38 where we are today and what year you
8:44 predict we'll
8:48 get there let's see if it
8:52 let's see if it goes off to the internet
8:53 yeah it says browsing the web so that's
8:55 interesting
8:57 Wally movie summary so it's going out to
9:00 the web and getting a movie summary from
9:01 wellie
9:03 I wonder where it went let's go find out
9:05 where it went it went to Wikipedia okay
9:08 solid
9:09 smart clever now it's scrolling down
9:12 it's thinking oh wait let me that's a
9:14 little too bright there you can there
9:15 you should be able to see it
9:18 um thinking it's thinking
9:21 reading content good this is good
9:24 thinking
9:28 all right
9:29 doing a little more thinking
9:33 oh okay so so the first one was go get
9:36 the the Wally movie summary the second
9:38 thing it did was what's the current
9:39 state of AI sentience and object
9:41 recognition
9:44 click failed so I don't know where it
9:46 went
9:50 um
9:53 very possible this is going to be
9:54 largely disappointing but you know hey
9:56 this is why we do this we're all
9:58 learning we're all just figuring this
10:00 crap out
10:02 bam
10:09 boy it's doing a lot
10:11 it's searching all over the place
10:12 predictions for when AI will achieve
10:15 sentience
10:18 research.ai multiple.com reading content
10:21 failed sometimes I think it goes to
10:24 sites that are um that require a login
10:28 and then it fails so
10:30 all right sure let's discuss some of the
10:33 Sci-Fi elements from the movie Wally
10:36 all right and now it's created a table
10:38 so sentient robots as of 2023 we don't
10:41 have them
10:43 predicted year given the current pace of
10:45 AI and research it's likely to be a
10:48 significant time for sentient robots
10:50 okay fine
10:51 Advanced computer vision
10:54 computer vision systems have greatly
10:56 Advanced and are highly effective it's
10:58 likely we will see further improvements
11:00 over time natural language processing
11:02 and understanding
11:04 so it didn't really give me a year
11:06 prediction and then autopiloted
11:09 spaceships
11:11 so yeah so it's like we've got some
11:14 elements I you know I kind of feel like
11:15 we're we're living in this we we are
11:18 quickly approaching a
11:21 what I'm calling a post-sci-fi world
11:24 where where you know just little
11:25 elements of of you know movies that we
11:28 were like that's Fantastical when we
11:30 were kids
11:32 um there's a lot of stuff I remember
11:34 from when I was when I watched Star Trek
11:36 as a kid where I'm like that'll never
11:38 happen and now we have like you know
11:39 phones that are
11:42 like what they had there and pretty
11:44 amazing stuff
11:45 um what can AI do some examples I just
11:47 showed you some
11:50 um AI can
11:51 here's a fun one here's a good way to
11:54 start so if you've never if you haven't
11:56 played with with chat EBT that's why I
11:58 have these URLs up go to chat.openai.com
12:02 or Bing go to bing.com and click on the
12:04 chat button
12:06 and start playing with these tools and
12:10 where where I would start is with what I
12:14 call parlor tricks and parlor tricks are
12:16 just cool things
12:18 that kind of demonstrate what the using
12:21 chat gpt4 that's connected to the
12:23 internet it went and found you know a
12:26 movie synopsis and then kind of compared
12:29 features of that movie with the future
12:31 and it actually wrote it put it into
12:33 table format like it can do that it
12:35 formatted that table it did all the
12:36 research format of that table like that
12:39 would probably take you a solid hour to
12:41 go research all that make a table do
12:43 that sort of thing
12:44 but let's do some let's do some fun
12:46 things and I'll just do it with GPT 3.5
12:48 since that's the free one so if you if
12:50 you go to chat GPT you can use it for
12:52 free but you only get to use 3.5
12:56 the the downside is it's not quite as
12:59 magical at the writing stuff but it's
13:00 super fast so let's say
13:04 um
13:05 my dog is
13:08 black and white
13:12 and
13:14 football shaped
13:18 and fat
13:20 and Barks at
13:24 anyone that walks by
13:29 um
13:32 right a country song about
13:37 my dog champ
13:41 and so zip bang it writes you
13:47 um a song and it knows things like verse
13:52 verses and choruses and it Rhymes so let
13:55 me tell you about my about A Pup Named
13:57 champ a black and white Wonder with a
14:00 football shaped stamp he's got a bark
14:02 that's loud and he's quite a sight when
14:05 folks walk by he puts up a fight my dog
14:08 champ he's a legend in town with his
14:11 round little body
14:13 um and his bark so loud he's black and
14:15 white like a piano key jump uh oh champ
14:18 you're the best friend to me so right it
14:20 wrote a cute chorus
14:22 um it didn't give me a title though
14:23 please
14:26 um give me 10 titles
14:29 so
14:31 so it understands that it just wrote a
14:34 song about a dog named champ so I don't
14:36 have to say give me 10 titles about a
14:38 song
14:39 with about a dog named champ I don't
14:41 have to repeat that so so one of the
14:43 things that these chat chibi teeth
14:45 chat GPT things do is they have memory
14:47 of what you can do so
14:50 champ The Barking Ball of love black and
14:53 white football shaped that's pretty good
14:55 Legends of champ the town's best friend
14:57 chubby champ the protector's Hal piano
15:00 keys and puppy love The Ballad of champ
15:02 so
15:03 um so I did that okay so so that's a
15:05 thing you can do so that's pretty slick
15:07 now I can say
15:10 um write me a professional
15:15 business email
15:18 um requesting
15:22 a partnership
15:25 with a
15:28 um
15:30 uh what's it called a
15:35 with a a company I admire
15:41 all right and so it understands I didn't
15:44 give it enough specific things to give
15:46 it a specific name
15:49 and I didn't really tell him anything
15:50 about my company so that would improve
15:52 it but dear company name I hope this
15:54 email finds you well I'm writing to
15:56 express my admiration for the company
15:57 company name outstanding achievements
16:00 and Innovative approach in your industry
16:02 field as the blank of my company right
16:05 so so it put variables in there and then
16:08 it's it's a decently long letter so I
16:10 can say things like make
16:13 it's shorter and more
16:17 conversational
16:19 and so it will just do that
16:24 boom boom boom so it starts out now
16:28 I hope this email finds you well
16:30 exclamation point I wanted to reach out
16:32 and let you know how much I admire you
16:33 your impressive achievements and Forward
16:35 Thinking approach
16:37 we specialize in blank our values and
16:40 vision strongly align with yours now I
16:42 can do this now I can say
16:45 um please
16:47 um format this email in HTML
16:53 that is safe
16:56 to send in and email uh in in an in a
17:02 bulk
17:04 email
17:06 program
17:10 and now it writes the code so so it's
17:13 basically rewriting the email that it
17:16 just wrote in HTML code
17:19 right so
17:22 so those are just like a handful of
17:24 examples like just just you know off the
17:27 top of my head and what I would say is
17:29 this
17:30 start playing with these tools and start
17:32 out with these kind of parlor tricks
17:33 have it write a song have it write a
17:34 poem have a I don't know
17:37 um look you can copy and paste your
17:39 LinkedIn profile just all the elements
17:41 from your LinkedIn profile paste it in
17:42 there and have it write you a resume
17:44 or have it write you a cover letter uh
17:46 if you're trying to get a job
17:48 and then what I would say is start to
17:50 think about projects in your life either
17:53 for you personally or for someone else
17:55 or for your job
17:57 where you want to accomplish something
17:58 and then try to figure out how to use
18:00 these tools to accomplish that a good
18:03 way to think about
18:06 is like an incredibly smart assistant or
18:10 an incredibly smart collaborator
18:12 Microsoft has named their chat GPT thing
18:16 co-pilot which I think is a really good
18:19 name and it's it's a very app
18:20 description you're the pilot you figure
18:22 out what you want to do and then think
18:24 of chat GPT as your co-pilot
18:27 um what if I want to talk to AI about
18:30 numbers not language well you can so let
18:32 me just do a new chat here I'll just pop
18:34 back over there
18:36 um thank you for the Roses AJ that's
18:38 cool
18:39 um so let me say uh because I don't know
18:43 [ __ ] about math so let's say
18:45 um uh please act as a
18:51 um advanced mathematics
18:56 professor
18:57 and tell me the 10 most important
19:04 equations
19:06 I should learn
19:10 thank you for the little thingy there
19:13 learn
19:15 all right as an advanced mathematical
19:18 dude hear hear your things the
19:20 Pythagorean theorem Euler's formula
19:22 foundational theorem of calculus
19:24 Maxwell's equations okay so now I can
19:27 say
19:29 um
19:30 please give me
19:35 five examples
19:38 um
19:39 that utilize Max
19:44 Wells
19:45 equations with
19:50 relevant data
19:55 certainly here you go
19:58 electromagnetic wave propagation
20:00 Maxwell's equations can be used to
20:01 analyze the behavior of electromagnetic
20:03 waves I think everybody knew that it's
20:05 fairly obvious isn't it let's consider a
20:08 scenario where the electromagnetic wave
20:10 with a frequency of three gigahertz
20:12 three ten to the ninth Hertz propagates
20:14 throughout space frequency F network
20:16 error regenerate response
20:19 sometimes chat gbt gets a little it gets
20:22 a little tired
20:23 too many people are using it they're
20:26 asking it for big equations so now it
20:29 went and solved it capacitor charging
20:33 magnetic field due to current so anyway
20:36 um so we can do math now
20:38 let me show you another cool thing
20:41 if you pay for
20:44 um
20:45 chat GPT plus and you get access to gpt4
20:51 then you also get access to these things
20:53 called plugins and in the plugins I
20:57 don't have any enabled right now but I
20:59 have a bunch installed one of them that
21:01 I have installed is
21:03 the Wolfram Alpha
21:06 um
21:07 uh plug-in so now chat GPT is connected
21:11 to wool from alpha
21:12 so please
21:15 um
21:17 show me all of Max Wells
21:23 equations
21:26 as a stand alone
21:31 formula
21:33 followed by an example
21:38 and that should now
21:41 fire off the plug-in
21:44 for wolf from alpha but maybe not maybe
21:47 it knows enough that it doesn't think it
21:48 needs to go out there and do that
21:52 um I'm trying to think what else
21:58 well let's we'll see if it we'll see if
22:00 it fires off the um
22:10 come on it's a little slow tonight am I
22:13 using oh I'm using GPT for gpt4 is
22:15 slower than 3.5 so if you want speed use
22:18 3.5
22:19 if you want quality use for
22:23 four is better at coding gpt4 is better
22:26 at writing it's like it's really good at
22:29 writing
22:32 can AI makes investment suggestions
22:35 Based on data available publicly online
22:38 so yeah so some of the plugins there are
22:41 Financial plugins
22:43 um so so you can do things like you can
22:45 have chat GPT write you an investment
22:48 bot you can take code from two different
22:51 investment Bots
22:52 and and put them together
22:55 it says using unknown plugin I wonder if
22:58 well from alpha just screwed up this
23:02 wolf from alpha could not understand
23:07 huh I don't know what it did
23:10 um anyway so I don't know why it didn't
23:12 use wolf from alpha there so it kind of
23:14 screwed up it looks like their servers
23:15 might be overloaded right now
23:17 um but anyway you can also
23:20 um some of the plugins are investment
23:22 like like there's ones that pull in
23:24 stock data real-time stock data and and
23:27 so so yeah you can analyze things with
23:29 chat GPT and use the plugins to go get
23:32 the real-time data or just use the
23:34 internet connected chat GPT
23:36 um but you can have it you can copy the
23:38 code from two different trading Bots and
23:41 paste it into chat GPT and tell it to
23:43 create you a hybrid bot that takes the
23:45 the you know combines the two approaches
23:48 um it's it's it it's really remarkable
23:52 all right so hello everybody if you're
23:55 new here my name is Kyle Shannon this is
23:57 the AI learning lab
23:59 these um what are they called URLs these
24:02 links right here that well they're not
24:03 links but um
24:05 if you haven't used chat GPT the top URL
24:08 there is chat apt's official website
24:11 chat.openai.com bing.com is Microsoft's
24:14 implementation of chat gpt4 and then
24:17 poe.com is another thing that gives you
24:20 access to chat GPT and Claude and their
24:23 own model sage and and some different
24:26 things the reason I put these here is I
24:28 want people just to try the real deal
24:29 there's a lot of scam sites out there
24:31 and scam apps that are trying to rip
24:33 people off so I'm trying to have you not
24:36 be ripped off if you have questions
24:38 about generative AI That's what I'm here
24:40 to do answer this
24:42 and uh yeah ask away and I'll do my best
24:45 to answer it and
24:47 have fun along the way
24:49 and I'm watching the Nuggets game
24:53 nuggets are down 89 to 85 right now the
24:57 heat are
24:59 picking it up they're picking it up
25:02 I live in Denver by the way so that's
25:04 why I'm looking I don't really give two
25:06 shits about basketball but you know
25:09 but they're in the finals so that's kind
25:11 of cool I'm I'm definitely a fair
25:13 weather sports fan unless it comes to
25:15 racing F1 and NASCAR both of those I
25:18 follow
25:19 uh not that you care oh the reason I'm
25:21 just rambling right now is I'm trying to
25:23 figure out where where I got lost in my
25:25 questions because Tick Tock reset my
25:27 questions and so I'm just trying to
25:29 figure out where I was so if it seems
25:30 like I'm just sitting here talking to
25:32 myself that's because I'm just sitting
25:34 here talking to myself how old am I I'm
25:36 58
25:38 um you think oh yeah okay so I'm I'm
25:42 sort of back to where I was hold please
25:44 I'm almost there
25:45 I have come to the conclusion a couple
25:47 of days ago that I am an AI it's
25:50 possible
25:52 that's awesome what are you some of your
25:54 favorite prompts
25:56 um
25:59 well so prompts.chat the the little
26:02 lower
26:03 thing here prompts.chat is is a thing
26:06 called awesome GPT
26:08 prompts
26:09 and and basically what it is is it's a
26:12 long long list of prompts that tell
26:14 Chachi PT to act as a certain kind of
26:19 person basically or a certain kind of
26:21 robot
26:23 um you know act as a stock Trader act as
26:25 a movie critic act as a
26:28 that actually works really well with
26:30 chat GPT because of the nature of how it
26:33 works it basically Narrows the sort of
26:37 you know they call it sort of the
26:38 semantic clusters or the vectors
26:41 um when you tell it to act as a stock
26:43 Trader it's not going to look at all of
26:46 the times someone mentioned stock
26:48 trading even if they're not experts if
26:49 you say act as a stock Trader it's going
26:52 to look in the in the in the latent
26:54 space it's called for the area where
26:57 people are stock Traders talking about
26:59 stock trading so
27:01 um so part of my favorite prompts are
27:04 are act as a
27:07 um
27:07 so that's a good one there's a there's
27:10 another good thing that you can do I
27:11 here's rather than a specific prompt
27:14 there's there's more like a prompt
27:15 philosophy
27:17 so Google for years has trained us that
27:20 you write the perfect prompt and you get
27:22 back an answer right and if they've done
27:24 their job right you get back this long
27:26 list of things and ideally the first two
27:29 or three links are the ones that you
27:31 need and then because of SEO and
27:34 advertising and things like that over
27:36 the past I don't know five or seven
27:38 years it's gotten pretty shitty where
27:39 it's just there's a lot of spam in there
27:41 and it's it's it's been over seo'd right
27:44 um
27:45 so so
27:48 with chat GPT because we've been kind of
27:51 trained to do this perfect prompt in
27:53 single answer out I think that's kind of
27:54 the expectation is that you know write
27:56 me the Great American novel and out
27:58 comes the Great American novel
28:01 what happens though is the
28:04 the more generic your prompt the more
28:07 generic the answer and so so a lot of
28:09 people try chat EBT and they're like
28:11 well it sucks it gave me a stupid answer
28:14 so rather than think about it like a
28:16 search engine think about it like a
28:18 conversation and you say well I want to
28:20 write the Great American novel where
28:21 would I start
28:23 hey Chachi PT can you act as a a writing
28:27 consultant that specializes in helping
28:29 people write the Great American novel
28:31 sure I'll do that and it'll probably
28:33 give you a little bit of advice then you
28:35 say I have an idea for a novel about you
28:38 know a guy that lives with a fat dog and
28:40 his two kids and his wife and I don't
28:43 know it's raining out whatever it is
28:46 um
28:48 what might be some other elements that
28:50 would make that story interesting and it
28:52 will give you some and then maybe you
28:54 like some of those and then you're like
28:55 I don't know how to organize a novel
28:57 what what are some different story
28:59 Frameworks that
29:01 um I might use to to make this novel
29:03 well there's save the cat and there's
29:05 this and it'll give you a bunch of story
29:06 Frameworks and then you say oh tell me
29:08 more about save The Cat and it'll tell
29:10 you and you're like can you help me
29:12 write an outline for my book using the
29:14 save the cat story framer sure here you
29:17 go out it comes right and and within you
29:21 know 10 minutes of that back and forth
29:23 interaction
29:25 you'll be more educated you'll
29:27 understand prompting better you'll
29:30 probably have a pretty solid outline to
29:32 your great American novel right and you
29:34 might be inspired to do other things
29:36 right and so so I so so for me the the
29:39 prompting game is is really about
29:42 prompting philosophy right and and it's
29:45 sort of break your conversation up into
29:48 small little chunks and give it a small
29:50 specific chunk that it can do well after
29:53 you provide it with a lot of context I
29:55 want you to be this consultant and
29:57 here's the idea I have and and you know
30:00 here's what I want my book to do and
30:02 then you ask it a simple little question
30:04 and it'll give you an answer based on
30:05 that context so that's that's how I
30:07 would approach it because all of the
30:10 specific prompts that people tell you oh
30:11 here's the 10 prompts that'll make
30:13 Chachi Pate beautiful you can make money
30:15 which I pay tag
30:17 um
30:18 those might work for them they might not
30:20 work for you you know I would say start
30:21 with what are you trying to accomplish
30:23 and then just think of it like if you
30:26 had an intern that just happened to be
30:28 smarter than anyone else on the planet
30:32 um how would you ask it how would you
30:34 ask that person to do something for you
30:36 I don't know you'd probably tell them
30:38 who you are and what you're trying to
30:39 accomplish and you'd give them some
30:41 instructions and see if they understood
30:42 it and then get an answer so that's what
30:45 I do software engineer a little worried
30:48 about AI yeah
30:50 um what should I study for job security
30:52 in the future I I mean I would lean in
30:55 heavy into
30:57 um
30:58 generative AI for
31:01 um I don't know where you are if you're
31:03 a software engineer you're probably
31:04 doing coding so if you're not using
31:07 github's co-pilot you probably should
31:11 um if I were if I were a software
31:13 engineer right now I'd be learning about
31:15 you know all of the open source large
31:17 language models I'd be installing them
31:19 on local machines and learning about
31:23 security
31:24 um I think there's a there's going to be
31:26 a big new area of
31:28 um of AI called AI Ops which is
31:33 there's going to be lots and lots of AI
31:34 tools and companies are going to want
31:36 their data manipulated in all sorts of
31:39 crazy ways and they're going to need
31:40 it's going to be like an I.T position
31:41 but it's going to be people that
31:43 understand what these AI tools are how
31:46 to connect them together how to connect
31:47 them to apis
31:50 um
31:50 I would start learning things like Lang
31:53 chain and auto GPT and baby AGI the
31:56 autonomous agents those things are all
31:59 kind of sophisticated you know
32:01 co-mingling of different tools on the
32:04 back end so I just start digging in and
32:06 learning learning learning
32:09 um
32:13 creators will never be replaced your
32:15 ability to imagine is beyond technical
32:17 skills yeah well so I I tend to agree
32:21 with that I think that what goes up in
32:24 value in a world where all the Tactical
32:26 stuff can be executed by the machines is
32:28 critical thinking creative thinking
32:31 curation taste
32:33 refinement
32:35 um understanding your audience
32:37 understanding strategy things like that
32:41 um yay so glad you're on thank you I
32:43 appreciate that
32:45 all right
32:47 um I can be in the university class and
32:49 have my knowledge project on my
32:51 classmates and be the dumbest wait I can
32:54 be in the university class and have my
32:56 knowledge project on my classmates and
32:58 be the dumbest I don't quite know what
33:00 that means oh dumbest student in the
33:02 class
33:04 well yeah I think what you're saying is
33:06 with access to these with access to this
33:10 immediate access to knowledge you don't
33:12 need to be book smart to be able to get
33:14 access to this knowledge that's true
33:17 um and and the reality is is that
33:19 everyone's going to have access to that
33:21 knowledge so I don't quite know the role
33:23 of universe universities as they
33:25 currently are are kind of broken in this
33:28 model right because a lot of what
33:29 universities are about right now is
33:31 giving you information in such a way
33:33 that you gain enough knowledge to be
33:35 able to be competent well with these
33:37 large language models you've got instant
33:40 access to that knowledge and and even a
33:42 level of competence like it can write
33:44 computer code for you even if you don't
33:46 know how to do that
33:48 um you still need to know how to put all
33:50 put it all together but um but yeah it
33:53 definitely does that
33:56 all right when you ask AI what
33:58 information do you need from me
34:02 how do you answer like 10 questions in a
34:05 row oh that's a really good question
34:06 here watch this this is cool
34:09 this is a little hack that
34:12 um
34:13 I got from someone on my
34:18 um
34:20 on my Tick Tock I put out a tick tock
34:22 and I said
34:23 you can take the readme file of a GitHub
34:26 paste it into chat gbt and and tell it
34:29 I'm on OS X and please give me
34:32 step-by-step instructions to get this
34:34 GitHub repository up and running GitHub
34:37 is a code repository for those of you
34:39 that are not geeky I'm just geeky enough
34:41 to know what it is but I don't really
34:42 know how to do it so I can have chat GPT
34:45 teach me how to do that step by step
34:47 which is cool
34:49 um
34:51 and
34:53 um and then someone on my Tick Tock told
34:55 me to do this okay so
34:57 um give me step by step instructions
35:03 for
35:05 um
35:06 getting a GitHub
35:10 repository up and running it may not do
35:15 it because it's going to ask me which
35:16 GitHub repository oh no
35:19 sure here's the step-by-step
35:20 instructions so now it's going to give
35:22 you the step-by-step instructions
35:24 create a GitHub account install git
35:28 configure git then it gives you the code
35:30 oh sorry to configure git now create a
35:34 new repository right so this is this is
35:36 about if I wanted to create my own
35:37 reposit
35:38 repository how to get it going okay now
35:41 I'm going to say stop
35:42 so so now what I'm going to say is this
35:44 I'm going to say
35:46 I want you to act as my coding coach
35:53 and give
35:57 me
35:57 each instruction
36:00 one at a time
36:04 and not move on to the next
36:09 step
36:12 until I say
36:16 okay I got it
36:21 and then I'm going to say do you
36:23 understand question mark
36:28 and now it should
36:30 absolutely I can do that we'll take it
36:33 one step at a time
36:35 step one if you don't already have this
36:36 blah blah blah blah blah blah let me
36:39 know when you've successfully created
36:41 your account okay I got it
36:46 excellent now let's move on to the next
36:48 step step two install git
36:51 go to the website download the version
36:53 suitable let me know when you've
36:55 completed it now I can say I'm
36:59 on OS X
37:02 does that matter
37:05 or is there anything special I need
37:11 to know question mark
37:16 yes you can install git on OS X you can
37:18 use Homebrew a package manager for OS X
37:22 and now it tells me how to install
37:23 Homebrew
37:25 so open your terminal on your Mac right
37:27 so so
37:29 and now I can say okay let me just say
37:32 okay I didn't even say okay I got it
37:34 great job
37:36 now it's moved on to step three
37:37 configure git so yeah so you can do that
37:40 so
37:42 hopefully that helps
37:44 what the funk is this this is the AI
37:47 learning lab my name is Kyle Shannon I'm
37:49 just a dude I'm an entrepreneur I'm the
37:52 founder and CEO of a company called
37:54 storybine which is an 11 year old
37:56 automated video platform
37:59 um I was around in the early days of the
38:04 World Wide Web in 1994 I started one of
38:06 the first online magazines and I I
38:08 co-founded a company called agency.com
38:10 which was one of the first digital
38:12 agencies
38:13 and one of the reasons I started this
38:16 channel is that I see incredible
38:19 parallels from the early days of the web
38:21 to what's going on right now with
38:23 generative Ai and chat GPT and and the
38:26 generative image engines and things like
38:28 that and so I'm just here to tell people
38:30 what I what I'm learning along the way
38:32 because I'm I'm getting up to speed
38:36 um on this AI stuff kind of in real time
38:37 over the past year and I'm sharing what
38:40 I'm learning along the way so if you're
38:42 new to ai go to the the top three
38:45 websites they're the very top one
38:47 chat.openai.com that's the official
38:49 chatgpt website so if you want to play
38:51 with chat CPT just go there and start
38:52 playing and then the other things are
38:54 some other options and then prompts.chat
38:57 will teach you how to do some prompting
38:59 and and some philosophy about it and
39:01 then futurepedia.io is just like this
39:04 database of tools lots and lots of tools
39:09 um
39:11 all right
39:14 can AI tell me the Powerball numbers
39:17 someone on the on my last lives said
39:20 that that um chat GPT correctly guessed
39:24 the Powerball numbers but they didn't
39:26 play them
39:26 [Laughter]
39:29 play to win
39:31 I'm a gen xer I don't I don't know if
39:33 you remember that there were TV
39:34 commercials they're like you know for
39:36 the lottery you got to play to win
39:40 so that always Pops in my head when
39:42 someone's like I should have done
39:43 something what program are you using to
39:45 have two cameras displayed I am using
39:47 this application called Tick Tock
39:50 um it's just um there's a when you go
39:53 live one of the options is dual camera I
39:56 I assume if you have dual cameras I
39:58 don't I don't know if it's specific to
39:59 devices
40:00 but most devices these days have dual
40:03 cameras so you just click dual camera
40:05 and and that box is my camera facing
40:08 away from me just you know pointing at
40:10 my laptop
40:12 so that's all that is
40:15 all right there are tools that existed
40:18 for decades the converted text to HTML
40:21 okay DJ so all right
40:25 picks are just tiring fine
40:30 um I need oops I don't need to shop I
40:34 need help
40:36 writing a python application
40:42 I want it to take
40:45 three text inputs
40:51 and
40:53 then incorporate
40:56 those values into a prompt
41:02 send that prompt
41:06 to open AI
41:10 um
41:11 API
41:13 and then take the results
41:17 and send an email to the person that
41:23 filled
41:26 out
41:28 the form
41:30 this was one of the first things I did
41:32 with chat GPT
41:37 sure here's a basic basic python script
41:39 that will do what you ask note that you
41:41 will need to replace placeholders with
41:43 actual API keys
41:45 so then it writes me some python code
41:50 and there's my form inputs
41:56 um
41:56 there's the responses
42:00 there's the headers there's it's sending
42:02 the thing out to open AI now it's
42:05 creating
42:06 Json of that now it's creating an email
42:10 template
42:11 so this is way more than writing a
42:16 procedural manually formatted
42:20 email formatting tool that would write
42:22 some HTML this is writing code right it
42:26 understands code it understands the
42:28 logic if you haven't used
42:31 github's copilot it sounds like you're
42:34 decently technical I would go start
42:37 using it and start seeing what this
42:38 stuff can do not only did it write the
42:41 code for me it's explaining how it did
42:43 it it's explaining how I need to replace
42:45 the URLs
42:46 it can analyze your code it can debug
42:49 your code it will often get it wrong but
42:52 if you're a good programmer
42:54 you'll know how to fix that [ __ ] but it
42:56 will help a lot
42:58 so anyway hope that helps
43:05 all right
43:08 that's not how math works okay I don't
43:11 know how math works I would say listen
43:13 I've got a degree in acting not
43:15 mathematics so if you know more about
43:17 math go to chatgpt and ask it how to do
43:21 the math that you're talking about and
43:25 um you might also want to try
43:28 bard.google.com which I don't have here
43:30 that's Bard's version of this or
43:33 Google's version of catch EPT is called
43:35 bard
43:37 it's really shitty at language but
43:39 apparently it's pretty good at code and
43:41 math because you know Google's an
43:43 engineering kind of company
43:45 gpt4 is way better at math plus gives
43:48 you math plug-ins but
43:51 um I'll explain Concepts great yeah
43:53 perfect yeah I think gpt4 is the way to
43:56 go but you could try bard
43:57 I've asked chat gbt to only ask one
44:00 question at a time and wait for me to
44:02 say next question yeah perfect
44:03 I'm looking for an API to scrub podcasts
44:06 to create videos so a lot of the plugins
44:09 right now are not a lot they're a fair
44:10 amount of plugins so I'll show you how
44:12 to do plugins here
44:14 if you pay for GPT Plus
44:19 make a new chat here
44:22 that little yellow flag behind chat GPT
44:24 is the plus sign
44:26 one of the first things that you should
44:29 do if you want to try plugins you go to
44:31 your settings
44:32 and then in your settings the second
44:35 option down here is called beta features
44:38 and there's browse with Bing and there's
44:40 plugins make sure those are enabled
44:42 since they're beta features you have to
44:44 explicitly turn them on
44:46 and then when you roll over gpt4 you now
44:49 have three modes you've got default
44:50 which is not connected to the internet
44:52 you've got browse with Bing which is
44:54 connected to the internet and then
44:55 you've got plugins so I'm going to
44:57 choose plugins
44:58 and then you've got this list of
45:00 installed plugins
45:02 and there's lots and lots of them I've
45:04 got a fair amount installed but if you
45:05 don't have any installed you go to the
45:07 plug-in store
45:09 excuse me just had ice cream and
45:15 um
45:17 I'm gonna say oh we'll see how many they
45:19 have here there's 65 Pages 65 times four
45:22 plug-ins like 250 260 plug-ins
45:25 and in here there are plugins that you
45:30 can point it to YouTube or you can point
45:32 it to you know audio files or things
45:34 like that it will pull the transcripts
45:36 from them and then you can scrub things
45:39 now if you're looking for an API
45:43 um
45:44 then then you'll just go need to find an
45:47 API to do that where you where you pull
45:49 the transcript and then send that
45:52 transcript to open AI
45:55 um or you could use zapier which is a no
45:57 code
45:58 um tool that can help you create an
46:01 Automation and string things together
46:06 I didn't see who won the Nuggets game
46:08 but they're interviewing the heat so I
46:10 assume the Heat have won
46:13 I don't really give too much of a [ __ ]
46:16 but I think it's tied 1-1 right now
46:18 let's see
46:19 yeah he's smiling the guy from The Heat
46:21 smiling
46:23 I think that means the Nuggets done
46:25 tanked it all right you're not even
46:27 using the best feature well tell me what
46:28 the best feature is I would love to
46:30 learn
46:31 I would like you to replicate your live
46:33 broadcast setup
46:35 can you help hardware and software okay
46:37 I have a chicken foot monopod so it's
46:39 like three little feet in a monopod I
46:42 gotta stand
46:44 I have an iPhone
46:45 I have a
46:48 64 LED light on a cardboard box
46:52 there's my cardboard box
46:55 and I have a laptop that's it
46:59 that's it
47:02 love the education plugin show me show
47:05 me is an incredible plug-in yeah show me
47:06 we'll we'll create little charts and
47:08 graphs for you I don't know does show me
47:11 do anything I've only seen it do like
47:13 flow charts does it do anything other
47:15 than flow charts
47:21 um you need an MLs Major League Soccer
47:24 plug-in I think there's a football
47:26 plug-in like a soccer plug-in I don't
47:29 know if it covers the MLS though but
47:32 that might not be what you're talking
47:33 about
47:35 um how long before they'll be able to
47:38 give answers with real-time info right
47:40 now today you mean like because it's not
47:43 connected to the internet so
47:48 if you go here see this if you pay for
47:52 uh catch EPT Plus
47:55 you can say browse with Bing so if I
47:57 click on that I can say when you're
48:00 talking to it yeah it's funny like I
48:02 find myself saying please and thank you
48:04 a lot and and ex and I'll like give it
48:06 encouragement I'm like that was a great
48:08 answer and I I don't know it seems to
48:11 work uh chat GPT can be helpful for
48:14 someone oh can chat GPT be helpful for
48:16 someone starting a business yes
48:18 my friend Mackenzie Bose who he came and
48:21 spoke I've got an AI salon and we bring
48:23 speakers in every other week and he um
48:26 he's got a he's got a project that he
48:29 started called GPT boss so if you go to
48:31 gptboss.com it's a bunch of it's
48:34 essentially a bunch of pre-prompts
48:37 um in front of uh gpt4 but he's got like
48:41 you know a CFO and a social media
48:46 copywriter and a social media manager
48:49 and a business coach and he's got all
48:51 these people that you can hire right you
48:53 can virtually hire them and they all
48:56 have very specific expertise and and
48:59 it's very much designed and he's also
49:01 got generators like generate marketing
49:03 copy you generate I don't know a
49:05 business letter or generate things like
49:07 that
49:08 um you can do that on your own as well
49:11 right you can you can start to say you
49:13 could you could even go to chat CPT and
49:15 you could say hey I'm thinking about
49:16 let's do it right gdpt 3.5 census this
49:19 it looks slow tonight
49:21 so I'm going to say
49:23 um
49:24 please please Act
49:28 as a new
49:30 business coach
49:33 I want to start my
49:37 own
49:38 business selling whatever 3D prints
49:45 of
49:46 fidget toys which is probably a horrible
49:49 business because
49:51 you can get them from a nickel from
49:52 China but whatever
49:55 um
49:56 where do I begin question mark
50:02 business selling 3D fidget toys can be
50:06 an exciting Venture one research the
50:09 market to develop a business plan three
50:11 determine your product line set up a
50:14 workshop Source materials and supplies
50:16 build an online presence right so it's
50:18 given me broad Strokes all the things to
50:20 do I could say
50:22 um what are the risks
50:27 of this uh this business
50:31 and are there other ideas that might
50:38 be better
50:41 so notice I'm kind of talking to it like
50:44 like if I were talking to an actual
50:46 business coach you know I might ask
50:48 things like that or the business coach
50:49 might volunteer that and if you if you
50:52 get your initial prompt right you know
50:55 it might ask you questions you might
50:57 have it interview you about your
50:59 business and ask all the the important
51:01 questions and then you tell it all the
51:04 answers and then it will give you a
51:05 business plan
51:07 like any business starting selling 3D
51:09 prints of fidget toys comes with its own
51:11 set of risks here's some potential risks
51:13 Market saturation the market for fidget
51:15 toys may become saturated oh it's beyond
51:18 saturated intellectual property
51:20 infringement yep that's a good one
51:22 quality control manufacturing and
51:24 Equipment issues
51:27 um changing customer preferences price
51:29 competition here's some other ideas
51:31 customized 3D printing services so offer
51:34 personalized 3D printing Educational
51:36 Tools and models spare parts and
51:38 accessories 3D printing workshops and
51:41 training so you know within whatever
51:43 it's been a minute here
51:45 um I've got some clarity at least around
51:47 my business and then you can just keep
51:49 digging like that if you spent
51:51 two hours with your business idea just
51:54 doing this and then sort of pull out
51:56 copy and paste out the things that you
51:59 feel are right or important the other
52:01 thing to do is like like give it
52:03 reinforcement say
52:05 um
52:08 um I really like the
52:12 personalized person
52:15 lies angle
52:18 can you talk more about what I might
52:24 do
52:26 um
52:26 specifically
52:30 customized jewelry
52:32 personalized home decor customized phone
52:35 accessories customize gifts and
52:37 keepsakes right I did one on the last
52:39 live I was just live earlier today and
52:42 someone said can it write
52:44 um
52:46 can it write like t-shirts for moms that
52:48 are burned out and you know tired right
52:51 or whatever exhausted and tired and it
52:53 was brilliant actually I think I
52:55 probably still have it up hang on I'll
52:56 show it to you it's really good
52:58 [Music]
52:59 um
53:00 mom t-shirt sayings so the prompt that I
53:04 wrote was give me a Toms that are tired
53:06 and overworked they should be clever and
53:08 funny and short enough to read at a
53:10 distance the table should include a
53:12 column for the saying a column for the
53:14 age of the kids and a column for the
53:17 category and and so again it's a fairly
53:20 shitty simple little prompt but it it it
53:23 dumped out 30 t-shirt ideas in you know
53:27 like 30 seconds so tired as a mother
53:30 that's not bad mom B the tired mom
53:34 um fueled by caffeine and Chaos raising
53:37 wildlings which is great I run on CAF I
53:40 run I run on coffee sarcasm and kids
53:44 tears I would change that too I run on
53:46 caffeine sarcasm in kids tears but it's
53:48 brilliant
53:50 um super mom has left the building can I
53:53 get a can I get the manager of parenting
53:55 I think that one's [ __ ] brilliant
53:56 that one would sell
53:58 um I can adult today
54:00 um I child-proofed my house but they
54:02 still get in I mean these are good
54:04 they're good toddler Wrangler world's
54:07 okayest mom I make mom jeans look good
54:10 wine because I can't drink coffee at 5
54:13 pm
54:15 so
54:17 um yeah so so it can be an incredible
54:20 business coach business support
54:24 is there a hugging face chat GPT plug-in
54:27 I don't think so yet that I know of
54:31 um I don't think so but it's it's
54:33 possible
54:35 I haven't looked at all the plugins I
54:38 it's there's two there's shit's
54:40 happening so fast it's just unbelievable
54:43 um if if you're new here
54:45 um the the URLs I have on screen here
54:47 the the top one is the official chat GPT
54:50 website then there's Bing and Poe or
54:52 just other chat GPT options to try and
54:56 then prompts.chat will teach you how to
54:58 prompt and futurepedia.io is tools uh
55:01 and if you're wondering who the hell I
55:02 am my name is Kyle Shannon this is the
55:04 AI learning lab I'm an entrepreneur and
55:06 a CEO and
55:09 um a guy that is
55:11 um
55:12 weirdly
55:14 um
55:15 I was I was called adorably optimistic
55:18 I'm adorably optimistic about generative
55:20 AI I think it's gonna be incredibly
55:22 transformative it's incredibly powerful
55:24 and I think it's important for people to
55:25 learn it so that's what this is about
55:28 um I don't know much more than you I've
55:30 been kind of diving into this stuff for
55:32 about the past year I'm I'm not a
55:35 machine a machine learning expert
55:38 I'm more interested in how we use these
55:41 tools to
55:42 amplify Humanity I want to write a short
55:46 and feature screenplay how do I prompt
55:48 chat GPT to help let's go do that
55:51 so I've written seven screenplays and
55:53 it's the gpt4 in particular is really
55:56 good at this stuff
55:59 um
56:01 do we need I don't even think we need
56:03 the Internet we can just use the default
56:04 model we don't want to go out and talk
56:06 to the internet we can just pull it from
56:08 within here so I can say
56:10 what are the three
56:13 most popular well okay I want you to act
56:19 as a screen writing coach
56:25 um please tell me the top three story
56:30 Frameworks
56:34 uh
56:37 for screen writing
56:40 did you say a feature a short end
56:43 feature for a screen
56:45 writing
56:47 um
56:48 we'll start with the feature I want
56:50 to I have
56:54 a feature
56:57 idea
56:59 all right sure be happy to help here's
57:02 the three things the 3x structure
57:04 structure the classic 3x structures the
57:07 Mainstay of the film industry it's split
57:09 into the setup Act One the confrontation
57:11 Act Two The Resolution Act three the
57:14 hero's journey this structure is based
57:17 on Joseph Campbell's uh book the hero in
57:20 a thousand faces it's a cyclical
57:21 structure featuring stages like a call
57:24 to Adventure refusal to to the call
57:27 crossing the threshold trials and
57:29 tribulations save the cat beat sheet and
57:33 and so so there's the three that it
57:35 likes those are three classic ones I
57:37 like to save the cat one just because
57:39 it's it's kind of interestingly
57:41 organized so I'm going to say okay my
57:44 idea is about
57:48 let's see
57:50 um
57:53 a dog
57:57 that um
58:03 jumps out of
58:07 his owner's car
58:12 at a stop
58:16 light but they don't notice and it's
58:21 about his
58:23 adventure to get back home
58:27 that's a tearjerker this one's a tear
58:29 jerker
58:32 that's a great concept it's got an
58:35 immediate hook inherent conflict and
58:37 clear so act one setup introductions to
58:41 the dog's happy home life
58:43 act two confrontation the dog realizes
58:46 he's lost stop that I'm gonna say can
58:49 you
58:50 um
58:51 can you please use the save
58:55 the cat
58:57 frame
58:58 work instead
59:02 question mark
59:06 foreign
59:10 image snapshot of the dog's comfortable
59:12 loving home Life theme stated during the
59:16 play someone can mention how curiosity
59:18 can lead to trouble ah you set it up you
59:21 see the family along with the dog embark
59:24 on a road trip the dog jumps out of the
59:26 car at a stoplight to investigate
59:28 something the dog has to decide whether
59:30 to continue exploring or try to follow
59:32 the car by the time he decides the car
59:34 is gone
59:35 right and it goes on and on and it gives
59:37 you all of the detail so
59:40 like
59:42 I just gave it a shitty
59:46 you know paper thin idea and it's
59:49 fleshed it out in a way that like huh I
59:51 could I can actually see this turning
59:53 into something right
59:54 and then you can do things like
59:56 um the the
59:59 dogs
1:00:01 name is Champ
1:00:03 please give give me 10 possible
1:00:10 titles for an animated
1:00:14 feature
1:00:16 that will make kids and adults
1:00:21 want to see it boom
1:00:26 absolutely here are 10 Champs big
1:00:28 adventure a dog named champ journey of
1:00:30 the Lost pump pup Champ's tail champ a
1:00:33 canine Odyssey
1:00:35 um Champs great journey home one tale at
1:00:38 a time
1:00:39 t-a-i-l huh
1:00:43 the story of champ champ the hero Hound
1:00:45 Champs paw prints a journey all right so
1:00:47 so those are fine I mean I I think one
1:00:49 tale at a time is pretty clever
1:00:52 um now I'm gonna say
1:00:54 um
1:00:55 write me a
1:00:58 Hollywood
1:01:00 log line that will get the most
1:01:06 cynical
1:01:09 agent and producer to want to
1:01:14 buy this film
1:01:20 champ the hero Hound a devoted family
1:01:22 dog named champ is left stranded near
1:01:25 the Heart of the City after after an
1:01:26 impulsive leap from the family car leads
1:01:28 to an unexpected separation Unleashed in
1:01:32 the urban jungle and guided by the
1:01:34 wisdom of a street smart stray oh that's
1:01:36 good
1:01:37 he must use every ounce of courage and
1:01:40 wit to navigate the Hostile forces and
1:01:42 daunting challenges in a heartwarming
1:01:44 thrilling quest to find his way back to
1:01:46 the family that desperately misses him
1:01:49 oh champ where did Chip go how did we
1:01:53 not know champ he's here
1:01:57 he's good
1:01:58 it's good
1:02:00 so
1:02:01 I was just uh just pulled out some of my
1:02:03 acting chops there huh was that good did
1:02:06 you believe it was it believable
1:02:08 [Laughter]
1:02:10 all right thoughts on using copyrighted
1:02:12 materials in the training of llm data
1:02:14 sets um
1:02:17 if the people that created the
1:02:21 initial data sets
1:02:24 had played by the rules and not used
1:02:27 copyright material we wouldn't have
1:02:29 these things
1:02:31 um I I have a sneaking suspicion so
1:02:35 stability AI they're the company that
1:02:36 made stable diffusion they've also got
1:02:38 their own large language model that they
1:02:40 just open sourced they open source their
1:02:43 image model too now they're a for-profit
1:02:46 company and they're open sourcing their
1:02:48 their core models and and I initially
1:02:50 was like why the hell are they doing
1:02:52 that and then what hit me was
1:02:54 I suspect that they knew that they
1:02:57 trained these things on copywritten
1:02:58 material and they might get their asses
1:03:00 sued from here to Oblivion
1:03:03 but they want the technology to be out
1:03:05 in the world so
1:03:06 um
1:03:08 in the ideal world
1:03:11 you would ask everyone for permission
1:03:13 for everything they've ever written and
1:03:14 put on the internet in the real world
1:03:17 there's so much [ __ ] on the internet
1:03:19 copywritten and non-copy written and
1:03:21 proprietary and non-proprietary that's
1:03:23 just out there it's been out there for
1:03:24 30 40 50 years
1:03:27 um if you really want to train one of
1:03:28 these models to have a vast amount of
1:03:31 knowledge you kind of have to train it
1:03:33 wide
1:03:34 um
1:03:36 companies like Adobe are
1:03:39 um
1:03:40 like they train their Adobe Firefly
1:03:43 their their model that's now part of the
1:03:44 generative AI part of Photoshop they
1:03:48 trained it on public domain and Adobe
1:03:51 stock images that they own all the
1:03:53 rights to so they made the conscious
1:03:55 decision not to train it on copywritten
1:03:58 material for that very reason right
1:04:01 because they're they're a big commercial
1:04:03 entity and they want anything produced
1:04:05 in Adobe products to be able to be used
1:04:07 commercially right
1:04:09 um so I think it it was a necessary evil
1:04:13 and now that it's been done these models
1:04:15 are [ __ ] incredible they're
1:04:18 incredible and if you understand how
1:04:20 they work
1:04:22 what they're not doing is they're not
1:04:24 copying and pasting passages from
1:04:27 copywritten work the way it actually
1:04:29 works is they're predicting a single
1:04:32 word at a time
1:04:33 based on your prompt they're saying
1:04:35 what's the probability of the next word
1:04:37 that should follow this prompt and then
1:04:39 the next one after that and then the
1:04:41 next one after that so it's writing
1:04:42 original content it's just the training
1:04:45 base is on lots of stuff and some of
1:04:47 that is arguably copywritten it is a
1:04:50 very gray area
1:04:52 um IP attorneys are going to be suing
1:04:54 for the next bunch of decades I'm sure
1:04:56 they're happy about this
1:04:57 um but anyway yeah that's what I think
1:04:59 about that do you think the you they use
1:05:02 wait
1:05:04 do you think they use this to push out
1:05:06 news articles yeah they're like like a
1:05:07 couple of uh Publications have recently
1:05:10 just
1:05:11 um admitted that they're doing that and
1:05:13 they've been doing that and they don't
1:05:14 don't intend on stopping and some have
1:05:17 even let go of some of their reporters
1:05:19 um because they're like we're just going
1:05:21 to do it with AI
1:05:22 um one of the big ad agencies I forget
1:05:24 who it was
1:05:25 um
1:05:28 so publicist no it wasn't publicist
1:05:31 I think it was uh
1:05:36 I don't know it was the French one no
1:05:38 that's publicist
1:05:40 I don't know one of them just said
1:05:41 they're going all in on AI on content
1:05:44 generation
1:05:46 um there's a there yeah this is why
1:05:48 they're wait is great
1:05:52 iy has lots of awesome open source
1:05:55 projects out there there's a lot of good
1:05:56 open source stuff out there I want to
1:05:58 write a screenplay can help I think I
1:05:59 just showed that hopefully that was good
1:06:03 um some comments in this live were
1:06:04 filtered to protect the community's
1:06:06 experience please don't Troll and please
1:06:08 don't attack each other you can attack
1:06:10 me what what did I got so so in the past
1:06:14 two days I've been told I look like
1:06:16 Rosie O'Donnell fat Alec Baldwin
1:06:20 um
1:06:21 um a skinny John Goodman
1:06:24 someone told me that I should really
1:06:26 pluck my nose hairs which I did
1:06:30 someone else told me to check my thyroid
1:06:34 um what did someone say to get filtered
1:06:36 I have no idea oh [ __ ] it just jumped me
1:06:38 down I gotta go back and figure out
1:06:39 where it was hold please
1:06:42 um yeah I don't like don't troll it's
1:06:45 just it's so [ __ ] boring
1:06:48 and don't be don't be [ __ ] I'm just
1:06:51 I'm tired of people being [ __ ] don't
1:06:53 be [ __ ] oh I want is for you people
1:06:56 to be nice to me and nice to each other
1:06:59 isn't that too much to ask
1:07:01 is that too much to ask
1:07:05 is it is that like the T-shirt thank you
1:07:09 my wife bought that yeah I like this one
1:07:12 [Applause]
1:07:18 sorry I'm trying to go back and figure
1:07:20 out where it was on the questions I
1:07:21 don't like to miss questions because I
1:07:23 actually
1:07:24 I mean that's why we're here
1:07:26 and if you asked a question I want to
1:07:28 try to answer it but it also means that
1:07:30 I'm way behind on my questions so if
1:07:32 you're like why hasn't answered my
1:07:34 question it's because I'm slow and old
1:07:36 I'm a 58 year old on Tick Tock what do
1:07:38 you expect from me I think I'm doing
1:07:40 okay
1:07:43 I'm arguing with myself storyvine is the
1:07:45 best thank you very much
1:07:47 storybines my company if you're new here
1:07:49 I just it looks like a bunch of new
1:07:51 people just joined um my name is Kyle
1:07:53 Shannon this is the AI learning lab uh I
1:07:56 talk about things generative AI I'm not
1:07:58 a machine uh learning Specialist or an
1:08:01 AI specialist I am an entrepreneur that
1:08:04 has been through a number of these
1:08:07 technical Revolutions in my life I was
1:08:09 just too young to be part of the PC
1:08:10 Revolution
1:08:12 um I hit it right in The Sweet Spot of
1:08:14 the world wide web and I started one of
1:08:16 the early digital agencies and I think
1:08:18 this is the next big one um generative
1:08:20 AI I think it's bigger than the web it's
1:08:22 it's huge huge huge huge and that's why
1:08:26 I'm here to to D mystify this stuff and
1:08:32 try to find out where I was in my
1:08:34 question answering
1:08:36 there was just an explosion
1:08:39 oh that was probably because of the
1:08:40 basketball game just ending you're not
1:08:42 supposed to shoot off fireworks if we
1:08:44 lost Dum Dums
1:08:47 oh man I'm way far behind sorry about
1:08:49 this okay I would like to replicate your
1:08:52 live broadcast setup oh okay I did that
1:08:54 one already okay I I'm back to where I
1:08:57 was now let me see let me get to a
1:08:58 question I haven't answered
1:09:00 how dumb has computers made Americans
1:09:03 well
1:09:04 the grammar of your sentence
1:09:07 partially answers that question
1:09:09 apparently
1:09:10 um chat GPT can help you with that
1:09:12 grammar what about apps what about them
1:09:14 oh like like well
1:09:16 so futurepedia.io is a database of AI
1:09:21 apps um there are tons of them out there
1:09:25 there's music generation ones and video
1:09:27 generation ones and image generation
1:09:30 ones and there's all sorts of knockoffs
1:09:32 and what I call wrapper apps which are
1:09:34 basically just prompts sitting on top of
1:09:35 chat GPT so if you just want to explore
1:09:38 apps you can go do that there I think
1:09:41 the most important skill right now if
1:09:42 you're new to this stuff is just start
1:09:44 playing with chat GPT
1:09:46 start understanding what it makes
1:09:48 possible
1:09:51 um anyway there was an F1 thing
1:09:54 um
1:09:54 who's your F1 team hello from Orlando
1:09:57 hey Frankie
1:09:59 um
1:10:01 I mean verstappen is staggeringly good I
1:10:06 he's a douchebag and I can't stand him
1:10:08 but like he won the race today by
1:10:11 24 seconds and you know the total wolf
1:10:15 was like well we we kind of slowed it up
1:10:17 there so it was probably only 15 seconds
1:10:18 that he really beat us by he he's
1:10:21 remarkable I love Lewis Hamilton I think
1:10:23 he's an incredible driver it's been
1:10:25 painful to watch Mercedes struggling as
1:10:28 much as they had for the past two years
1:10:30 um I like Lando Norris a lot
1:10:33 um
1:10:36 I mean Fernando Alonso we have like
1:10:39 three or four podiums in five races am I
1:10:43 a software engineer no I have a degree
1:10:44 in acting I'm an entrepreneur I'm the
1:10:46 founder and CEO of storyvine which is an
1:10:49 automated video platform I'm introducing
1:10:51 my students to AI tomorrow good
1:10:53 they are 8th graders and community
1:10:55 college students
1:10:57 I love it I love it like yes
1:11:00 this here's the thing
1:11:02 these tools are going to be everywhere
1:11:04 they're going to get rolled into
1:11:05 Microsoft Office Suite they're going to
1:11:08 get rolled into windows 11. they're
1:11:10 probably going to get rolled into you
1:11:12 know Apple's operating systems
1:11:14 they're going to get rolled into the
1:11:16 Google Suite of tools right so Bard is
1:11:18 going to be within you know dot Google
1:11:20 Docs and sheets and all that sort of
1:11:22 stuff so this stuff's going to be
1:11:23 everywhere it's going to be a required
1:11:25 skill to lose it use it so so what you
1:11:29 can do with students now is you can say
1:11:30 hey this Tool's out there let's learn it
1:11:33 together let's learn what it's good at
1:11:35 let's learn what hallucinating is let's
1:11:37 let's learn how it gets stuff wrong
1:11:39 let's let's let's understand that the
1:11:42 nature of how we prompt this thing
1:11:45 um can actually dramatically affect the
1:11:47 quality of the result let's let's learn
1:11:50 critical thinking to understand you know
1:11:53 why are we using this tool in the first
1:11:55 place what are we trying to accomplish
1:11:56 you know what does success look like
1:11:58 right all that sort of stuff is going to
1:12:00 be critical so aniline yes I'm awesome
1:12:06 more of that
1:12:08 will AI write programs for me to do X Y
1:12:11 or Z and if so how would I get it
1:12:15 um it can you need to know a little bit
1:12:18 about programming if you like it will
1:12:20 write code for you if you don't know
1:12:22 coding what I would use is is what's
1:12:24 called a no code tool like zapier so let
1:12:28 me let me show you quick
1:12:29 I've written a number of content engines
1:12:31 with zapier and I'm not a coder and I
1:12:34 have 80d so I can't I just don't have
1:12:36 the attention span to do real real
1:12:37 coding
1:12:38 uh did I I think I sent him with Google
1:12:42 um
1:12:45 remarkable conceptually but very very
1:12:47 early and very very janky
1:12:50 um you know language like Lang flow is
1:12:52 the the front end in front of Lang chain
1:12:56 to make people like me be able to use it
1:12:58 but it's it's just really early and
1:13:00 really janky
1:13:02 um and kind of like spit and duct tape
1:13:03 together right now so is auto GPT all of
1:13:06 those tools are very very promising
1:13:08 they're just
1:13:10 um
1:13:12 they're just
1:13:15 they break a lot
1:13:18 um and and they don't necessarily tell
1:13:19 you why they're breaking or how they're
1:13:20 breaking and they're complicated sets of
1:13:22 tools and and how the prompting actually
1:13:24 works is a little is a little funky so I
1:13:28 would say if you're a developer I would
1:13:32 be leaning into those things hard and
1:13:33 what I would probably do is I would
1:13:35 actually join the open source
1:13:37 communities and start contributing to
1:13:39 those projects
1:13:40 um I would really learn them deep if I
1:13:42 were a developer because I I think I
1:13:44 think autonomous autonomous agents are
1:13:46 going to be a big big deal I think
1:13:48 Apple's going to be great at them do I
1:13:51 think prompt writing will continue to be
1:13:53 important as chat AIS get better
1:13:58 I don't think that you'll have I so so
1:14:00 there's already something
1:14:02 version of it that came out let me see
1:14:04 if I've got it installed
1:14:10 I'm going to turn off wool from alpha
1:14:14 prompt perfect
1:14:16 okay so what this thing is it's pretty
1:14:19 slick
1:14:20 so if in my prompt I type perfect as the
1:14:24 first word and then I say
1:14:28 um
1:14:30 right
1:14:31 a poem
1:14:34 that mimics
1:14:37 a famous poet
1:14:40 about
1:14:42 um
1:14:43 the blue sky
1:14:47 that's a crappy prompt but I wrote
1:14:49 perfect in front of it
1:14:52 and so
1:14:54 it's firing up that plug-in prompt
1:14:56 perfect
1:14:59 and then I think inside it just sends my
1:15:01 crappy prompt to prompt perfect they
1:15:03 probably have some prompt
1:15:05 that rewrites my prompt so it says sure
1:15:09 I'll write a pro poem
1:15:11 in the style of Robert Frost who is
1:15:13 known for his nature poetry and his
1:15:16 contemplative often melancholic tone so
1:15:18 if I open up this thing
1:15:20 what I typed was what I what I told you
1:15:22 here it rephrased it compose a poem that
1:15:25 imitates imitates the style and tone of
1:15:28 a renowned poet while focusing on the
1:15:30 topic of Blue Sky your poem should
1:15:33 capture the essence of the original
1:15:34 poet's work while also incorporating
1:15:37 your own unique perspective and voice
1:15:40 and then so the the thing responded with
1:15:43 that and then behold the boundless blue
1:15:45 I stand a solitary figure on life sand
1:15:48 the sky a canvas brought in Grand
1:15:50 painted by a Divine hand the Azure Dome
1:15:54 so vast so high mirrors of depth I mean
1:15:57 that's not bad so I think that
1:16:01 a lot of the tools are gonna get you're
1:16:04 gonna get assistance that happen behind
1:16:07 the scenes uh where you where you write
1:16:09 something crappy and it'll fix it for
1:16:11 you so
1:16:12 so I don't think that I I think it's
1:16:15 it's kind of like
1:16:16 why do you need to learn math if we've
1:16:19 got calculators and computers and
1:16:20 spreadsheets well because you should
1:16:22 probably understand the concept of like
1:16:23 you know addition and subscription
1:16:25 subtraction and division things like
1:16:28 that so I think it's very similar
1:16:30 because we're very early having that
1:16:32 skill and honing that skill of
1:16:34 understanding how to talk to these
1:16:36 things
1:16:37 you really get to understand what
1:16:39 they're good at what they're not good at
1:16:40 why they make mistakes and and you can
1:16:42 you can get better at it so I think it's
1:16:44 an important skill but I don't think
1:16:46 it'll be long term I think I think we'll
1:16:48 pretty quickly get to the point that you
1:16:50 just kind of ask it vague shitty stuff
1:16:53 and it gives you remarkable results
1:16:59 how do we get gpt4 to analyze real live
1:17:02 real-time MLS data
1:17:05 MLS is in multiple listing services like
1:17:08 real estate
1:17:09 there's probably a plug-in for that and
1:17:12 then gpt's chat gbt is also coming out
1:17:15 with this thing called code interpreter
1:17:17 that can take big data sets and analyze
1:17:20 the [ __ ] out of them go watch videos on
1:17:22 code interpreter in fact
1:17:27 and I will search for code
1:17:30 interpreter chat GPT it comes up as a
1:17:33 search
1:17:35 um
1:17:36 just go like any any of these top videos
1:17:38 here
1:17:39 um talk about code interpreter it's
1:17:42 insane it's insane
1:17:45 baby AGI baby AGI is just like Auto GPT
1:17:49 and I think there's a couple other ones
1:17:51 out there GPT for all or something like
1:17:53 that they're they're what are called
1:17:55 autonomous agents so rather than having
1:17:57 to just prompt and get an answer back
1:17:59 you give it a goal and it will go write
1:18:01 its own prompts and figure out what
1:18:03 tools it needs to use and it'll go
1:18:05 execute [ __ ] on your behalf it's a
1:18:08 little scary
1:18:10 um and they don't really work all that
1:18:12 well right now but
1:18:14 um that's that's what that thing is
1:18:16 uh I'm awesome welcome back after an
1:18:19 eight day vacation yeah I got um
1:18:22 I got put into tension for being a bad
1:18:24 boy
1:18:24 [Laughter]
1:18:27 all right I know a guy that makes a lot
1:18:29 more money than me he's a coordinator oh
1:18:32 interesting all right we will provide
1:18:34 Learning Community Support oh that's
1:18:36 cool
1:18:38 amelin
1:18:40 um can chat AI be helpful for someone
1:18:43 starting a business we covered that
1:18:46 um
1:18:48 we need to learn to be Visionaries I am
1:18:51 I a professor I'm not a professor I'm
1:18:53 just a dude
1:18:55 that's passionate about this stuff right
1:18:56 now
1:18:57 um
1:18:59 do we need to learn to be Visionaries I
1:19:02 don't know that we need to learn to be
1:19:03 Visionaries but I think I think those
1:19:06 people that
1:19:11 that can think more again it's not quite
1:19:13 Visionary but at least strategic I think
1:19:15 those people that can think more
1:19:17 strategically and
1:19:18 and think critically
1:19:21 um I think that's going to be the core
1:19:22 skill in the future is knowing what to
1:19:26 ask rather than how to ask it and right
1:19:29 now a lot of the work that we have is
1:19:30 how do you do the thing
1:19:32 and there's less people that do the what
1:19:35 should we be doing and I think I think
1:19:37 that's going to flip where where the
1:19:39 real value is going to be in you know
1:19:41 what should we be doing who are we
1:19:42 talking to what are we trying to
1:19:44 accomplish what does success look like
1:19:46 what tools can make that happen how do I
1:19:48 put them together
1:19:49 I think that becomes the new skill
1:19:51 inventors should be in prison totally
1:19:54 disagree but totally understand the fear
1:19:58 stop talking to Demons okay I don't know
1:20:01 that I am but maybe I am and I will
1:20:04 swear I I swear here so you know if you
1:20:08 don't like it sorry
1:20:09 um
1:20:11 are you not getting are am I not hitting
1:20:14 the query limit with gpt4 I'm not and if
1:20:17 I do it's only for three hours and
1:20:20 I'm not using it enough to really hit it
1:20:23 and um if I do I just flip over to 3.5
1:20:26 or I go to Bing or I go to Poe so
1:20:30 why aren't you using unrestricted Dalai
1:20:34 Lama because I'm too lazy to install it
1:20:39 um the best thing you can do is test it
1:20:41 and if chat GPT gets something wrong you
1:20:43 correct it exactly do you have a course
1:20:45 we can purchase I don't I was thinking
1:20:47 about putting together a workshop
1:20:50 um you know teaching a workshop so I
1:20:51 might do that
1:20:53 um here's the deal like
1:20:56 it is my strong contention that no one
1:20:58 right now is an expert everything's
1:21:00 changing so fast that
1:21:02 um
1:21:02 even if I put together the perfect
1:21:05 course and taught you here's exactly
1:21:07 what to do three months from now
1:21:10 half of that half of that is going to be
1:21:12 wrong so maybe what I'll do is I'll put
1:21:14 together a course that teaches like the
1:21:18 how to think about it rather than how to
1:21:20 do prompt engineering you know here are
1:21:22 the 10 perfect prompts that's all
1:21:24 [ __ ] that's all going to go away so
1:21:28 um maybe I'll maybe I'll think about
1:21:29 putting something together that's more
1:21:30 about the philosophy of of how to work
1:21:32 with these things rather than the the
1:21:34 Tactical piece because the Tactical is
1:21:36 just gonna go
1:21:38 um you have to stay up on it people that
1:21:39 don't will be left behind I I think you
1:21:41 need to get curious about this stuff I I
1:21:43 agree with that well Apple release an AR
1:21:45 VR device tomorrow yes they will
1:21:49 um they will probably in conjunction
1:21:52 with it release some sort of AI stuff
1:21:55 but from what I'm hearing from like I
1:21:57 just listened to a
1:21:58 Twitter spaces from Robert scoble that
1:22:01 was saying he thinks it's going to be a
1:22:03 year before they really do anything with
1:22:04 AI which is a drag because I think
1:22:06 Apple's gonna kill it I think they're
1:22:08 gonna do really good with this AI stuff
1:22:10 with the like autonomous agents
1:22:14 um and and
1:22:16 it sounds like their goggles are going
1:22:19 to be
1:22:21 interesting
1:22:23 um they're not ready for prime time I
1:22:24 think they're only manufacturing 300 000
1:22:26 of them they're only for the development
1:22:28 Community there's another version coming
1:22:31 probably in October like before
1:22:32 Christmas
1:22:34 um so I don't know
1:22:36 thank you for doing these lives and
1:22:38 being so patient and supportive you're
1:22:39 welcome
1:22:41 um
1:22:42 uh uh
1:22:43 I ask they say Okay
1:22:46 um how to persuade middle managers that
1:22:49 they must learn chat GPT
1:22:51 hmm
1:22:53 how does it do what you said wait how
1:22:55 does it do that I don't know what it did
1:22:57 um
1:22:59 I
1:23:02 I you know I I think how you persuade
1:23:04 middle managers that they must learn
1:23:06 chat GPT is that you don't tell them
1:23:08 they should go learn it I think you
1:23:10 evangelize it I think you get them in a
1:23:12 room and you say let me show you what I
1:23:14 learned I mean that's what this channel
1:23:16 is like this this channel is really just
1:23:19 me going oh [ __ ] I figured something out
1:23:21 who can I tell no one I know gives a
1:23:23 [ __ ] about this stuff what do I do uh
1:23:25 I'm 58. I know he'll start a tick tock
1:23:28 Channel
1:23:33 ah
1:23:35 but I did that
1:23:37 um
1:23:37 I would get him in a room and I would
1:23:39 show him what it does and I I would say
1:23:41 what's the hardest part of your job
1:23:45 or maybe you know the hardest part of
1:23:46 their job and you just show them hey you
1:23:48 know how you take like most of Friday
1:23:51 afternoon to to compile this little bit
1:23:54 of data
1:23:55 watch this
1:23:57 paste in the raw materials tell it to
1:23:59 generate the work that they do and watch
1:24:03 it spit it out in 30 seconds and then
1:24:05 when they pick their jaws up off the
1:24:07 ground say
1:24:08 this is why you need to learn this stuff
1:24:10 because if you don't someone's going to
1:24:13 come along that does what you do 10
1:24:15 times faster and cheaper
1:24:18 and
1:24:20 what are we supposed to do like I don't
1:24:22 know I would I would I would I would
1:24:25 demo this stuff I would evangelize it I
1:24:27 would get like look how exciting this is
1:24:29 look look what this can do like what if
1:24:31 this freed up time for you to do the
1:24:33 kind of work you really enjoy so you're
1:24:35 not doing so much of this you know
1:24:36 repetitive Soul Crush crushing work you
1:24:39 could do the more creative interesting
1:24:41 work that you're passionate about I I
1:24:43 would you know less stick more carrot
1:24:47 what if I asked to tell everything I can
1:24:52 about it my name
1:24:54 um
1:24:55 you can do that you can even you can
1:24:57 have it ask you a series of questions to
1:25:00 talk about yourself and then do
1:25:02 something with what you told it
1:25:04 any University recommendations I I'm
1:25:07 really reticent to do that the
1:25:09 University of Florida has
1:25:11 leaned in pretty heavy on AI Nvidia
1:25:15 built them a AI super computer and it's
1:25:17 rolled out across the University's
1:25:20 curriculum it's not just in the computer
1:25:21 science department which I think is the
1:25:24 right way to to think about this because
1:25:25 this is a business tool now it's not
1:25:27 just a it's not just for the I.T
1:25:29 Department this is a full-on business
1:25:31 tool and and will be increasingly
1:25:34 um
1:25:35 I would I would talk to the you know the
1:25:38 school leaders that you're looking at
1:25:40 and and ask them their position on AI
1:25:43 and if they're like well we've got
1:25:44 blockers that block that we're we're not
1:25:48 going to deal with that I would steer
1:25:50 far away from them I would I would you
1:25:53 know
1:25:54 what what probably makes sense is go
1:25:56 listen to a bunch of podcasts on
1:25:58 education where they're talking about
1:25:59 this and like that's how I heard the guy
1:26:01 the Provost of the University of Florida
1:26:03 I heard him on the Nvidia podcast and I
1:26:06 was like [ __ ] I would go to school there
1:26:09 um
1:26:10 so I would go find schools that are
1:26:12 taking this stuff seriously and they're
1:26:14 out there
1:26:15 there was in fact go go look for go
1:26:18 Google
1:26:19 um there's a New York Times article I
1:26:20 think back in January
1:26:23 um there was a bunch of heads of
1:26:24 universities saying we've got to take AI
1:26:27 seriously it's probably worth finding
1:26:28 that article
1:26:30 who programmed chat GPT
1:26:32 um well open AI is the company
1:26:35 and this is here here's a fun little
1:26:37 fact so chat GPT was launched on
1:26:40 November 30th 2022 it got to 100 million
1:26:43 users in six weeks six weeks
1:26:46 um the World Wide Web it took six years
1:26:48 to get to 100 million users chat EBT was
1:26:51 six weeks but here's the fun little fact
1:26:56 gpt3 came out like a year and a half or
1:27:00 a year and a quarter before chat GPT
1:27:02 came out and in the documentation of
1:27:05 gpt3 were instructions for how to create
1:27:08 chat GPT and Sam Altman the CEO of
1:27:11 openai assumed that someone would make
1:27:13 it no one made it so Microsoft was
1:27:17 getting ready to announce their billion
1:27:19 dollar investment in open Ai and Sam
1:27:21 Altman said to his Tech Team hey you got
1:27:23 two weeks to throw together chat gbt
1:27:25 they built that thing in two weeks
1:27:29 they really just did it as like a proof
1:27:31 of concept so that when they made this
1:27:33 announcement there was something for
1:27:34 people to go play with they had no idea
1:27:37 it was going to take off like it did
1:27:39 six weeks it it got to a million users
1:27:42 in five days and a hundred million users
1:27:44 in six weeks
1:27:46 um and it changed everything it changed
1:27:49 everything it it smoked Google Out of
1:27:52 the Shadows it you know meta went all in
1:27:56 on the AI stuff
1:27:57 um Apple's going to do something big AWS
1:28:00 just announced it like it it changed
1:28:02 everything so that that's who did it I
1:28:04 feel like you're talking to the master
1:28:06 from Tron uh okay concatgpt be helpful
1:28:09 for someone starting small business I
1:28:11 think I answered these I must be in a in
1:28:14 the wrong place in the questions
1:28:16 I think I am
1:28:19 probably got kicked off for drinking
1:28:21 water again I know right oh am I stuck
1:28:23 oh damn it am I still stuck am I good
1:28:26 if I'm good hit the little heart button
1:28:28 so I know I'm I'm being hurt
1:28:32 um
1:28:33 [Music]
1:28:35 uh okay good thank you
1:28:38 people put putting too much Fate on chat
1:28:41 GPT
1:28:42 too much faith on chat EBT
1:28:45 um
1:28:45 maybe so so listen chat GPT at this
1:28:49 point is kind of irrelevant it's it's
1:28:51 rolled into Microsoft so it's called
1:28:54 co-piloted inside Microsoft Google has
1:28:57 barred
1:28:59 um anthropic has clawed that's a
1:29:02 different company
1:29:04 um
1:29:05 um you know Poe has Sage there are now
1:29:09 20 or 30 open source large language
1:29:12 models out there so
1:29:15 um there you go
1:29:16 uh
1:29:18 people putting too much fate I assume
1:29:20 you mean faith on chat gbt you're not
1:29:22 too impressed okay have you used it
1:29:27 like what have you tried to have it do
1:29:30 chat GPT is a millennial wine Mom it is
1:29:33 it's also emansplating as a service why
1:29:35 can't I sleep why can't I stop doing
1:29:37 these lives the world is too interesting
1:29:39 it is it's very interesting right now
1:29:41 there are alternatives like to Amazon
1:29:43 like French wait
1:29:46 there are alternatives to Amazon like
1:29:49 French or Godfather 2 or one act
1:29:53 I don't know what that means
1:29:56 but cool what model are you referring to
1:29:58 I have no idea adobe's a big company
1:30:00 with Deep Pockets so they become a prime
1:30:03 target for being sued yeah but Adobe
1:30:05 Adobe knew that and and so that's why
1:30:08 they did their model that way they don't
1:30:10 you know they're trying to limit being
1:30:13 sued
1:30:15 um and they you know they want big
1:30:17 companies to feel confident that they
1:30:19 can use generative AI without fear of
1:30:22 getting sued it's not so much adobe's
1:30:24 afraid of getting sued adobe's afraid
1:30:27 that if they put something out there
1:30:29 that puts their customers at risk
1:30:31 that'll totally [ __ ] up their business
1:30:34 so they want to make sure that if you
1:30:36 generate something with Adobe products
1:30:38 they're they're fair game for
1:30:41 um
1:30:42 for commercial use
1:30:45 all right
1:30:50 I'm sorry I lost my place I'm finding it
1:30:52 again
1:30:53 laughing my ass off Rosie yes someone
1:30:56 told me I looked like Rosie O'Donnell I
1:30:59 am not fat Baldwin and I'm not skinny
1:31:02 John Goodman but I was called all of
1:31:04 those in the past two days
1:31:06 ah yes ah the trolls uh if you're new
1:31:10 here my name is Kyle Shannon this is the
1:31:11 AI learning lab go to those URLs at the
1:31:14 top there that top one is uh the
1:31:16 official chat GPT website just go there
1:31:18 play with it
1:31:20 it's important
1:31:21 for you
1:31:24 definitely a risky Vibe not sure what
1:31:26 that was about the perfect world is on
1:31:28 the table I know
1:31:31 um who do I need to cut Lolo
1:31:34 oh that's funny but someone oh that's
1:31:38 larisha works for me I don't know if
1:31:40 you're still here but hey there Rosie uh
1:31:44 I don't know you know the trolls [ __ ]
1:31:46 the trolls I'm right there with you
1:31:48 um uh is it terrible that I'm using chat
1:31:51 GPT to create a resume no okay here's
1:31:54 the deal Chachi PT is a tool it's a tool
1:31:57 just like Microsoft Word is a tool just
1:31:59 like your grammar checker is a tool just
1:32:01 like your spell checker is a tool just
1:32:03 like a calculator is a tool it happens
1:32:05 to be a really fancy [ __ ] tool but
1:32:08 it's a tool
1:32:10 so if you need to write a resume
1:32:13 and that would take you four hours
1:32:16 normally and it spits it out in
1:32:20 six minutes with back and forth
1:32:24 I'd say you're smart I'd say you've got
1:32:27 three and a half hours back in your life
1:32:28 that you can go out and hang out with
1:32:30 your dog so I'm new to AI I'm relatively
1:32:33 new to AI I spent about
1:32:36 um six months really diving deep into
1:32:38 the generative image stuff and then when
1:32:40 chat gbt came out um started learning as
1:32:43 much as I could and I started an AI
1:32:45 Salon I started a newsletter called
1:32:47 everyday AI I started this Tick Tock
1:32:50 Channel I started an art project called
1:32:51 Kyle Shannon dreams on Instagram and
1:32:53 tick tock
1:32:55 I started five or six other projects
1:32:58 that died on the vine because I've got
1:33:00 add and that's the [ __ ] that happens
1:33:01 when you get excited about something
1:33:06 um have I tried the
1:33:08 Randon Nautica which attempts to read
1:33:11 your mind oh no but I someone told me
1:33:13 about something that was like that uai
1:33:16 dot a i yesterday where you answer like
1:33:20 it gives you a bunch of questions you
1:33:22 answer them and then it figures out your
1:33:23 psychology and does [ __ ]
1:33:26 um I I would much rather have a tool
1:33:28 that I can just say here's my social
1:33:30 media history like here's all my
1:33:33 accounts go learn all the [ __ ] I said
1:33:36 put it in a database and then let's talk
1:33:39 about who I am and what we can do what
1:33:41 we can do together you make me giggle a
1:33:44 true Graphics guy thank you I don't
1:33:47 assume they're American
1:33:48 um have you played with Lion chain I've
1:33:50 played with Lang chain that's been set
1:33:53 up I haven't really set it up
1:33:55 um I got I had chat GPT
1:33:58 give me a tutorial on setting it up and
1:34:01 I got about 90 of the way through it and
1:34:03 then I just got bored
1:34:05 because the the problem with buying
1:34:07 chain right now if you're not an
1:34:08 engineer it kind of requires more of an
1:34:12 engineer mindset than I have right now
1:34:13 you really need to Tinker with it you
1:34:15 really need to kind of dig deep on it
1:34:17 it's quite simple to install it's quite
1:34:19 difficult to do something that doesn't
1:34:21 suck
1:34:23 so I'm just gonna wait for it to get a
1:34:24 little better better and you know some
1:34:27 Geeks will put a fancy front end on it
1:34:29 in fact I I just saw on Twitter today
1:34:34 someone built a whole thing
1:34:36 on top of Lang chain where it's like you
1:34:38 point your YouTube channel at it and
1:34:39 your Google Docs at it you point all
1:34:41 these different things at it and then
1:34:43 you can go interact with your [ __ ] so
1:34:45 that's what I'm kind of waiting for
1:34:46 someone just to build the front end for
1:34:48 it so I don't have to dick around with
1:34:49 it
1:34:51 um what do I believe chat EPT is the
1:34:52 next big thing what do you believe
1:34:54 Chachi PT is the next big thing yes I
1:34:57 well it doesn't
1:34:59 large language models are the next big
1:35:01 thing
1:35:02 chat GPT is just an instance sitting on
1:35:06 top of a large language model
1:35:08 but there there are already thousands of
1:35:11 tools built on top of chat GPT things
1:35:13 like Jasper and copy AI
1:35:16 um there's going to be image generation
1:35:18 tools video generation tools animation
1:35:21 generation tools PowerPoint generation
1:35:23 tools like
1:35:24 all of this stuff is all the same
1:35:28 technology just aimed at different
1:35:31 um like tool areas so yeah I think it's
1:35:34 a big thing I think it's bigger than the
1:35:36 world wide web and I like I said I was I
1:35:39 was a founder of one of the first
1:35:40 digital agencies back in the mid 90s so
1:35:42 I've been through one of these once I
1:35:43 think this is bigger bigger than that
1:35:46 what's been the Big Technical
1:35:47 Improvement in AI that caused this
1:35:50 explosion this is cool
1:35:52 so in I can actually show you a slide on
1:35:56 this
1:36:02 where is it where are you
1:36:07 so
1:36:10 cool wrote a paper called attention is
1:36:13 all you need
1:36:15 and they introduced this thing called
1:36:16 the Transformer and what the Transformer
1:36:19 did was prior to the Transformer you had
1:36:23 these neural networks and you have these
1:36:24 machine learning
1:36:27 um you know methodologies that take
1:36:29 advantage of these neural networks
1:36:31 but you could only train it on
1:36:33 relatively small data sets and you kind
1:36:35 of had to bring your own data sets and
1:36:37 what the Transformer allowed is that you
1:36:39 could train these neural networks on
1:36:41 massive data sets
1:36:43 and what openai discovered as they use
1:36:47 these things more and more is that the
1:36:49 more data and more processing power you
1:36:52 threw out them the better they got and
1:36:54 so they created GPT one and then gpt2
1:36:58 and then in 2020 they released gpt3 and
1:37:02 gpt3 was kind of the first one that was
1:37:04 people were like
1:37:06 um holy [ __ ] this is really good
1:37:08 and they did a couple of things there
1:37:11 they did
1:37:12 um
1:37:14 they trained it on English language and
1:37:16 and open AI uses this thing called human
1:37:19 reinforcement learning where they they
1:37:21 get the models to a certain point and
1:37:22 then they have human beings interacting
1:37:25 with the tools and then they'll say is
1:37:27 this good or bad
1:37:29 and then and then the it fine-tunes the
1:37:32 model and that's what got it really good
1:37:33 so that was one thing that they did and
1:37:35 then the other thing that they did was
1:37:37 they created this thing called uh
1:37:40 code something they basically taught the
1:37:43 large language model
1:37:45 like 60 computer programming languages
1:37:47 they taught it how to program right and
1:37:49 that became co-pilot which is in GitHub
1:37:52 what they found is that when they taught
1:37:54 this model to do the programming the the
1:37:57 English language or the language side
1:37:59 got dramatically better so there was
1:38:01 kind of a step increase in capability an
1:38:04 emergent
1:38:05 um an unexpectedly strong benefit from
1:38:09 teaching this language model how to
1:38:11 program and and uh and then that evolved
1:38:14 into gpt4
1:38:16 um but the Transformer is what's sitting
1:38:18 underneath mid-journey and stable
1:38:20 diffusion the image generation tools
1:38:22 it's the stuff that you know they're
1:38:24 reading people's brain scans right now
1:38:26 and recreating images that they're
1:38:27 seeing it's all based on this stuff so
1:38:30 so because you can train these models on
1:38:33 these massive data sets now you can
1:38:35 start to do all these other interesting
1:38:37 things with it so that invention by
1:38:39 Google in 2017 is is kind of the Tipping
1:38:42 Point that precipitated all of the stuff
1:38:45 that we're seeing and the ironic thing
1:38:47 is is that they invented it but then
1:38:49 they kind of held that technology back
1:38:52 I assume because they recognized that it
1:38:54 would be an existential threat to their
1:38:56 search business so they said well we
1:38:58 don't want to put ourselves out of
1:38:59 business and then Sam Altman went and
1:39:02 made a kick ass
1:39:03 large language model and then built chat
1:39:05 GPT that a hundred million people
1:39:07 started using within six weeks and that
1:39:10 changed everything
1:39:11 so what you're going to see over the
1:39:13 next two years is there's there the
1:39:15 entire development Community is just
1:39:17 pounced on this whole area so the
1:39:19 Innovation we're going to see over the
1:39:21 next two years is going to be
1:39:23 just more face melting more face melting
1:39:27 stuck what happened to him the signal
1:39:29 just went
1:39:30 am I still
1:39:32 out
1:39:33 I missed the Henry Ford wait I missed
1:39:35 out when Henry Ford manufactured a car
1:39:37 that's funny
1:39:39 um I missed that when we started nuclear
1:39:40 energy
1:39:42 uh
1:39:45 how can I make a career online with AI
1:39:48 um take what you know just start
1:39:49 learning AI
1:39:51 and then take things that you're
1:39:53 passionate about or you know a lot about
1:39:55 find a problem in the world that you
1:39:57 have a better solution for and have a AI
1:40:00 help you solve that problem it doesn't
1:40:02 ai's just the tool to help you get there
1:40:05 don't think like you can get react for
1:40:07 attaching pite
1:40:09 you can make money without GPD
1:40:13 all of those people are [ __ ] they're
1:40:16 trying to sell you something
1:40:18 how you make a business how you make
1:40:20 money is you solve a problem well enough
1:40:24 that someone is willing to pay you money
1:40:26 for that solution if you can use AI to
1:40:29 do that great
1:40:31 maybe you love kitties and you love
1:40:33 coffee mugs and you want to make kitty
1:40:35 coffee mugs well you could hire
1:40:37 illustrators on Fiverr to make those
1:40:39 kitty drawings or you could make those
1:40:41 kitty drawings or now you could use
1:40:43 mid-journey to make those kitty drawings
1:40:45 and then you go stick those kitty
1:40:47 drawings on coffee mugs and sell them on
1:40:49 Etsy I mean it the the the mechanism of
1:40:53 of how you start a business and how you
1:40:56 generate value in the world that doesn't
1:40:57 change the power and amplification of
1:41:02 these tools is what changes so maybe you
1:41:05 can do things at a very different scale
1:41:07 than you could have before or you can do
1:41:09 it faster or cheaper things like that
1:41:12 how do we agent GPT to do the activity
1:41:15 and then just spit lines of tasks
1:41:20 um
1:41:21 just tell it to act like a project
1:41:23 manager tell it what you're trying to
1:41:24 accomplish and have it spit you lines of
1:41:26 tasks and it will do that
1:41:30 um reshma Thomas how do we ask agent gbt
1:41:32 to do oh I just I just answered that
1:41:36 um right show them what you accomplish
1:41:39 using AI show your aoi correct
1:41:42 uh why would you actually show them then
1:41:45 you lose your job no I totally disagree
1:41:49 um
1:41:50 no they're gonna listen these tools are
1:41:54 going to be rolled into all of the work
1:41:56 software that we use anyway
1:41:59 um
1:42:00 I
1:42:03 I would encourage you with with this AI
1:42:05 stuff
1:42:06 try not to have a mentality of scarcity
1:42:08 but have a mentality of prosperity
1:42:10 meaning
1:42:12 there's more than enough to go around
1:42:14 like learn this stuff and get generous
1:42:16 with it because
1:42:19 even if it's positive change there's
1:42:22 going to be a lot of disruption so get
1:42:25 in a generated get in a generous mode
1:42:27 get curious about it and then teach
1:42:30 people as much as you can
1:42:32 even if you don't know anything teach
1:42:35 someone that knows more than you do
1:42:36 there's people out there that have never
1:42:37 heard of chat GPT there's a lot of
1:42:39 people out there who've never tried chat
1:42:41 gbt and are just making assumptions
1:42:43 about it and think they have three years
1:42:45 to deal with this they don't it's coming
1:42:47 fast so I would say
1:42:50 um just engage with them
1:42:55 you won't share the information unless
1:42:57 middle management asks
1:43:00 I'd rather get more lessons and skills I
1:43:02 don't know I would go the other way I
1:43:03 would I would get middle management up
1:43:05 to speed on this stuff University of
1:43:07 Florida was the one I was talking about
1:43:09 if you look up the Nvidia podcast on ai
1:43:13 nvidia's ai podcast
1:43:15 um it was back in January or February
1:43:18 just look up the University of Florida
1:43:20 Provost that that episode it's a it's a
1:43:23 really interesting episode can you tell
1:43:26 the AI is offering competing wait can
1:43:29 you tell what AI is offering competing
1:43:31 to accountants
1:43:33 um
1:43:34 all of it and code interpreter that's
1:43:36 coming from uh open AI is going to do
1:43:39 data analysis at a crazy crazy level
1:43:43 sorry you answered that there's a
1:43:45 bulletin board with these AIS uh yeah oh
1:43:48 sorry there is
1:43:51 um that bottom URL there futurepedia.io
1:43:54 that's a list of lots and lots of
1:43:56 different AI tools so if you want to go
1:43:58 geek out and blow four hours of your
1:44:01 life after you spent two wasted two
1:44:03 hours of your life with me
1:44:05 um go to futurepedia but what I would
1:44:08 really strongly encourage though is go
1:44:10 to chatgpt and just start playing start
1:44:14 learning it
1:44:15 share what you know yes I agree
1:44:18 um do you know about
1:44:19 mid-journey pricing tiers yeah well I
1:44:22 think the bottom one's 10 bucks
1:44:25 uh uh uh it's Runway where's mid Journey
1:44:33 mid-tierney's got the worst [ __ ] icon
1:44:35 in Discord
1:44:38 um rules welcome trial support
1:44:45 no
1:44:49 um
1:44:51 yeah do something I rarely do anymore
1:44:53 I'm gonna go Google that
1:44:56 mid Journey pricing tears
1:45:01 10 bucks a month gives you 200 minutes
1:45:03 30 bucks a month gives you unlimited GPU
1:45:06 time the Pro Plan is 60 bucks a month
1:45:09 what's nice about the Pro Plan is you
1:45:11 can create private areas so people don't
1:45:13 see your prompts
1:45:15 um
1:45:18 let me go back a little bit here I I got
1:45:20 I got lost into questions I got lost
1:45:24 is there a free AI for science
1:45:26 literature well
1:45:29 so here's the thing
1:45:34 the the base knowledge understanding
1:45:38 sitting underneath gpt4 is like all of
1:45:41 the internet
1:45:43 so you can just say
1:45:46 um
1:45:49 act as a science fiction literature
1:45:51 professor at an elite University like
1:45:54 start your prompt with that and then
1:45:56 start interacting with it and start
1:45:59 talking about sci-fi books and
1:46:01 and uh
1:46:04 characters and theories and things like
1:46:07 that
1:46:08 I asked gpt4 on Poe if it was gpt4 it
1:46:12 said it was gpt3 not 3.5 wait gbt4 on PO
1:46:19 if it was gpt4 it said well that may or
1:46:23 may not
1:46:24 be relevant so so
1:46:28 um GPT hallucinates GPT confidently will
1:46:31 answer you even if it's wrong
1:46:33 um I know for a fact that Bing is based
1:46:35 on gpt4
1:46:37 I assume that if Poe is labeling
1:46:40 something gpt4 that it's actually GPT
1:46:42 for
1:46:44 um
1:46:45 I'm like 20 minutes behind on comments
1:46:47 that was exactly my point on why I did
1:46:49 Adobe did that uh so yeah I know I'm
1:46:51 behind sorry about that
1:46:53 um someone said Iraq thank you Kevin I
1:46:56 appreciate that
1:46:57 the man my man the AI brain yeah it's a
1:47:01 it's a cloudy brain I just wrote an
1:47:04 obituary with it and it was really
1:47:05 heartfelt and exactly what I was trying
1:47:07 to say you know
1:47:09 like I I've said this before and it it's
1:47:12 this is this it's a really interesting
1:47:15 Nuance about these tools so
1:47:18 you know people ask you know is it
1:47:20 sentient and does it have Consciousness
1:47:23 and is it reasoning and is it doing all
1:47:25 this and does it have feelings and the
1:47:26 answer is no to almost all of that it's
1:47:29 it's a probability engine
1:47:32 but
1:47:34 the way they coded
1:47:36 gpt4 in particular it it very much wants
1:47:40 to please you and it very much wants to
1:47:42 interact with you as a human
1:47:44 conversation
1:47:45 and I find that it that it's very good
1:47:48 at empathetic language and it's very
1:47:51 good at Nuance you can ask it very
1:47:53 polarizing questions where if you ask
1:47:55 that in today's world you know you'd get
1:47:58 flamed out and canceled or whatever it
1:48:00 might be and
1:48:03 gpt4 well all the gpds really because
1:48:07 they don't have an emotional an
1:48:08 emotional component to it
1:48:10 they won't it doesn't get triggered
1:48:14 so you can just put in like facts and
1:48:17 say compare this with this even if those
1:48:19 are really highly charged things in our
1:48:21 society it will just compare them and it
1:48:24 does it with nuance and it does it with
1:48:26 Grace and it does it with empathy and
1:48:28 and they did they did a test a week and
1:48:31 a half or so ago now that they had
1:48:34 patients talk to chat GPT doctors you
1:48:38 know and they didn't know they were chat
1:48:40 gbt and real doctors and then they asked
1:48:43 them which doctors did you prefer and
1:48:45 the patient's overwhelmingly chose the
1:48:47 chat CPT doctors and they cited two
1:48:50 reasons one they were more competent and
1:48:53 two they were more empathetic
1:48:56 so so it's it's a it's it doesn't
1:48:58 surprise me at all that it that it you
1:49:01 know it that it wrote that obituary
1:49:02 obituary well and helped you gather your
1:49:05 thoughts it's it tends to be very good
1:49:07 at that
1:49:08 fellow A.D ADHD here yeah go ADHD
1:49:13 ADHD Club president that's hilarious
1:49:16 um
1:49:17 did I think they will merge AI VR
1:49:21 metaverse and Beyond yeah I think that I
1:49:24 think that um AI is going to drive
1:49:27 adoption in the metaverse I think
1:49:28 whatever Apple's going to announce
1:49:30 tomorrow
1:49:32 is going to be a stepping stone but I I
1:49:35 think ultimately their glasses flop but
1:49:38 I think probably the thinking beneath
1:49:40 them and how they're thinking about
1:49:42 World building and how Apple already
1:49:44 thinks about utility
1:49:47 um I think they're going to be really
1:49:49 good at
1:49:50 at some of the stuff that isn't so much
1:49:53 like going into VR like the virtual
1:49:55 world [ __ ]
1:49:57 I think it's it's going to be more about
1:49:59 augmented reality
1:50:01 and it's going to be more about
1:50:04 autonomous agents where you get to talk
1:50:06 to them
1:50:07 and maybe you even get to talk in your
1:50:10 with your synthesized voice and it will
1:50:12 do things for you and so yes I think AI
1:50:15 is gonna ai's gonna accelerate the
1:50:17 ability to build Virtual Worlds but AI
1:50:21 is also going to power a lot of the
1:50:22 augmented reality stuff and then AI is
1:50:25 gonna
1:50:26 um
1:50:27 is going to be the the kind of lubricant
1:50:30 of being able to communicate with these
1:50:32 tools without having to type stuff and
1:50:34 do prompt engineering and all that
1:50:36 [ __ ] I I think it'll get very high
1:50:38 level very quickly where you just kind
1:50:40 of ask your phone or your tools or your
1:50:43 computer hey I would like to start a
1:50:46 business what should I do that will come
1:50:49 everything like it won't be such a chore
1:50:51 to figure out how to use these tools and
1:50:54 and then yeah I think ultimately we kind
1:50:56 of naturally evolve into having glasses
1:50:59 that just kind of look like glasses like
1:51:01 this where I can you know meet someone
1:51:04 and it reminds me what their name is
1:51:06 because I have ADHD and I'm horrible
1:51:09 with names that's so and so here's when
1:51:11 you met them here's the last thing you
1:51:13 talked about oh that would be nice and
1:51:15 then at some point I can just you know
1:51:16 turn on the the blinders and go into
1:51:18 full virtual reality if I just want to
1:51:20 avoid people
1:51:22 so yes I think though I think AI is
1:51:24 going to drive all that stuff is there
1:51:25 an open source NLP framework worth
1:51:27 learning
1:51:29 um
1:51:30 I I don't know when you say NLP
1:51:33 framework you're talking large language
1:51:34 model
1:51:36 um there are so many moving so fast I I
1:51:39 can't recommend one I'd say just start
1:51:41 hanging out on Reddit would probably be
1:51:45 the place to hang out and and hang out
1:51:47 with geeks that are geeking out on large
1:51:49 language models they are moving so fast
1:51:52 that it's it's hard
1:51:54 I can't get Dolly to make an image that
1:51:57 remotely looks like my prompt is
1:51:58 mid-journey way better mid journey is
1:52:01 way better Dolly sucks
1:52:05 it would be nice if it didn't because
1:52:07 you can do easy API icons to
1:52:11 um Dolly but you can't do them to Mid
1:52:13 Journey
1:52:14 but mid journey is just a whole other
1:52:17 level and you have to be a bit geekier
1:52:20 to use stable diffusion so I I mid
1:52:24 journey is the only one I pay for of the
1:52:26 image generation things I think Dolly
1:52:29 sucks I don't use it at all
1:52:31 for that very reason it just don't work
1:52:35 if you've got Photoshop go play with the
1:52:38 go go get the Photoshop beta in your
1:52:41 Creative Cloud little app there go to
1:52:44 the beta apps and get the Photoshop app
1:52:46 and play with generative fill
1:52:48 here I'll show you some to work so this
1:52:50 image so the image I have behind all of
1:52:53 that um
1:52:55 all of that uh
1:52:57 whatever you call it the uh like the the
1:52:59 chat CPT thing so the core image here
1:53:03 let me see
1:53:09 right where's my history hey hold please
1:53:13 I flattened this image so let me see if
1:53:15 I can use my history to go back
1:53:21 yes okay
1:53:23 so
1:53:24 let me move this thing out of the way so
1:53:27 the original image here
1:53:30 was that
1:53:32 and so then I selected the left and
1:53:33 right sides and I just said so so when
1:53:36 you select something
1:53:38 this little window pops up that says
1:53:41 generative fill and you just click that
1:53:42 button and then you can either type in a
1:53:44 prompt or or not
1:53:47 and so what I did was I selected the
1:53:50 areas to the left and right of that
1:53:51 image and I said generative fill I
1:53:53 didn't type anything in and it it filled
1:53:56 in that now this is like really subtle
1:53:58 gradients like to photoshop all of this
1:54:01 stuff in would take you forever to get
1:54:03 the subtlety of the gradients right like
1:54:05 gradients are really hard to to clone
1:54:08 and then I grabbed the bottom part of
1:54:11 the image and you know where the sort of
1:54:14 shoulders were and it gave me this cool
1:54:16 smoke option was one of them but there
1:54:18 were different options it gave me
1:54:20 um let's see
1:54:22 there was it gave me that option with
1:54:25 kind of like a hipster hoodie
1:54:27 and that option with like a dark dark
1:54:30 zip up front but it gave me that and I
1:54:31 like that so then I grabbed the little
1:54:33 two white areas to the left and right
1:54:35 and I said generative fill and it filled
1:54:37 in the rest of the smoke and then I just
1:54:39 grabbed the top and I said fill that in
1:54:40 and again there are a couple of options
1:54:42 here oh sorry sorry that's the wrong
1:54:44 layer there was that option which is
1:54:48 kind of cool
1:54:49 there was that option which is bad like
1:54:52 it screwed up the top of that thing and
1:54:54 then there was this one which I like
1:54:55 pretty good so generative fills really
1:54:58 really really powerful so if you've got
1:55:00 access to photoshop get the beta and
1:55:02 start playing with the generative fill
1:55:03 because it is
1:55:05 awesome it's awesome awesome awesome
1:55:09 awesome
1:55:10 can I explain the Transformer that
1:55:13 um Google built in 2017 please
1:55:17 I can I cannot explain what it does
1:55:19 technically I can tell you what it made
1:55:22 possible
1:55:23 um if you want to read about it yourself
1:55:26 you can go to go search Google for for
1:55:30 um
1:55:31 the phrase attention is all you need so
1:55:35 that I have it up here somewhere maybe I
1:55:37 don't anymore I'll just go look
1:55:39 my attention is all you need so go
1:55:42 Google attention is all you need
1:55:46 hang on
1:55:51 and you'll see this paper that archive
1:55:55 that arxiv is said archive
1:55:59 and you go to attention is all you need
1:56:01 and this is the paper and then and then
1:56:03 over on the right hand side
1:56:05 you go to the PDF
1:56:09 and you can read it you can also just
1:56:12 copy and paste the text from this and
1:56:14 paste it into chat gbt and ask it to
1:56:17 explain it
1:56:18 but what this paper did one of the the
1:56:20 primary things that this paper did that
1:56:22 that has had all the impact on
1:56:25 everything is
1:56:27 um
1:56:28 it introduced this thing called the
1:56:30 Transformer and what the Transformer
1:56:32 allowed was for these large language or
1:56:35 the these machine learning models to be
1:56:37 trained on massive amounts of data and
1:56:40 so open AI
1:56:42 um realized this and and they kept
1:56:44 throwing more and more compute power and
1:56:46 more and more data at it and it just
1:56:48 kept getting better and so that's when
1:56:50 they flipped from a non-profit company
1:56:52 to a for-profit company and they got
1:56:55 Microsoft to invest and they used all of
1:56:57 Microsoft Azure servers to be able to do
1:57:00 all this compute and things like that so
1:57:02 it's it's basically just the thing it's
1:57:05 the T in GPT generative pre-trained
1:57:08 Transformer so that t and GPT is the
1:57:11 Transformer architecture that allowed
1:57:14 these things to be trained on massive
1:57:15 data sets like all of the internet
1:57:18 so that's the result of what it is how
1:57:20 it actually does it I don't know well
1:57:22 enough to talk about
1:57:25 workforce management kind of hurt us
1:57:29 on the find a way to solve a problem
1:57:31 yeah I know I listen corporations are a
1:57:34 pain in the ass and
1:57:36 they do not always serve the best
1:57:37 interest of the people so
1:57:41 um I talked about Microsoft pricing
1:57:44 tiers I'm gonna probably I'm getting
1:57:47 tired I'm gonna sleepy
1:57:51 so I'm going to answer a few more
1:57:52 questions and then I'm going to take off
1:57:55 all right can I explain what people mean
1:57:58 when they say AI will make millionaires
1:58:00 yeah so
1:58:07 so if if you if you were there for the
1:58:12 early days of the web and you started
1:58:14 you know a little site
1:58:16 called you know Yahoo or pets.com or
1:58:20 whatever it might have been
1:58:23 um
1:58:26 when you first started it no one
1:58:29 everyone assumed that no one would ever
1:58:32 put their credit card in the internet
1:58:34 right they assume no one would ever buy
1:58:37 anything on the internet
1:58:39 if you were someone like Jeff Bezos and
1:58:41 you're like you know I think we could
1:58:43 probably sell books on this thing and
1:58:45 and a lot of the early days are like no
1:58:47 one's gonna put their credit card in the
1:58:48 internet and so Jeff Bezos became Amazon
1:58:51 right became Jeff Bezos
1:58:54 um because they did
1:58:55 we're in a very similar place with this
1:58:57 generative AI stuff right now where ai's
1:59:00 been around for decades but chat GPT has
1:59:03 made GPT accessible to business people
1:59:06 you don't have you don't just have to be
1:59:07 a machine scientist data scientists
1:59:11 um machine learning people or data
1:59:12 scientists or researchers you can be
1:59:15 just like a guy like me that's got a
1:59:16 business that's like holy [ __ ] this
1:59:18 could change everything about my
1:59:19 business
1:59:21 the next five to ten years
1:59:24 they're gonna be they're going to be
1:59:26 businesses and concepts of business and
1:59:29 concept of how we work and how we make
1:59:32 money that we can't even imagine right
1:59:34 now
1:59:35 so
1:59:36 the opportunity is start learning about
1:59:39 this stuff because as janky and weird
1:59:41 and imperfect as it is right now it's
1:59:44 remarkable
1:59:45 and what will happen is you might be the
1:59:48 person or I might be the person to go oh
1:59:51 you know what this could do if you if
1:59:54 you took this tool you stuck it together
1:59:56 with that tool and you solved this group
1:59:58 of people's problems that could be kind
2:00:00 of cool and then you put that out there
2:00:02 as a business which you can do really
2:00:04 quickly now because you have all these
2:00:05 really powerful tools to help you do
2:00:06 that
2:00:07 you put that out in the world and 5 10
2:00:10 20 years from now that might be worth
2:00:13 billions of dollars right now whether or
2:00:15 not you survived all that and
2:00:17 capitalized on it that's up to you right
2:00:19 that's what entrepreneurship is
2:00:21 but but we're at such a small teeny tiny
2:00:26 um
2:00:26 part of the beginning of of this
2:00:29 industry that we don't even have a clue
2:00:31 what it looks like yet so there's just
2:00:33 massive amounts of opportunity so that's
2:00:35 what they're talking about
2:00:36 I was talking about aliens with them and
2:00:38 I see I cannot get
2:00:40 to the love level of intelligence that
2:00:43 they will help us further develop our
2:00:45 brain but not surpass it
2:00:47 uh I would say gpt4 has significantly
2:00:50 surpassed my brain at this point
2:00:53 probably not some of the nuanced stuff
2:00:55 and the critical thinking stuff and the
2:00:57 reasoning and the empathy but the
2:00:59 knowledge holy [ __ ] [ __ ] it knows
2:01:01 everything
2:01:03 we are guiding the teens and young
2:01:04 adults into the web 3 Industry they
2:01:07 create exonomy yeah interesting
2:01:10 should go back by Act
2:01:13 they pick one word that they say and
2:01:15 build it into it okay I think you're
2:01:18 talking amongst yourselves here that's
2:01:20 cool MJ is learning more every day
2:01:22 mid-journey is amazing
2:01:24 if you haven't played with mid-journey
2:01:26 it's worth it you can play with
2:01:27 mid-journey I think it's up to 25 images
2:01:29 just to go see what it what it is and
2:01:31 what it's like you have to have a
2:01:33 Discord account and you have to learn
2:01:34 how to do it in Discord but it's not
2:01:36 terribly just go to midjourney.com and
2:01:39 you can learn it but it's remarkable
2:01:41 screen's freezing again I know because I
2:01:43 have Xfinity and they suck
2:01:45 I will be needing to illustrate a book
2:01:47 for kids I use can I use mid Journey yes
2:01:50 brilliant for that brilliant for that
2:01:52 and you can say like watercolor you you
2:01:54 tell it whatever style you want and
2:01:57 whatever you want the animals to be and
2:01:58 I wanted to you can tell it children's
2:02:00 book illustration in watercolor style
2:02:03 that's got an elephant walking near a
2:02:05 river
2:02:07 it's incredible you could probably do it
2:02:10 over the weekend
2:02:13 um do I have anything on stem program do
2:02:16 I do I have anything on stem program
2:02:19 how to learn
2:02:21 I go to gpt4 and have it create you a
2:02:25 custom stem curriculum for your kid with
2:02:28 whatever their challenges might be at
2:02:30 whatever age they are and have it
2:02:34 generate a stem program tell it what
2:02:38 tell chat gbt what your kids hobbies and
2:02:41 interests are
2:02:42 and tell it to generate like a summer
2:02:45 stem program curriculum for them to
2:02:48 learn
2:02:49 using only examples and activities that
2:02:53 they are interested in
2:02:55 that that would be a good one is Google
2:02:56 bar better than Bing and chat EBT not my
2:02:59 experience certainly not for writing if
2:03:01 you if you want to do writing chat GPT
2:03:03 or gpt4 is the way to go Bard sucks at
2:03:06 writing
2:03:08 um I've heard Bard is better at coding
2:03:11 but gpt4 is pretty [ __ ] capable of
2:03:15 coding so I I have every time I've tried
2:03:19 Bard even with the new stuff that just
2:03:21 came out last week
2:03:23 is still a little stinky it's stinkier
2:03:26 than it should be for the company that
2:03:28 invented the core technology which one
2:03:30 is best I don't know which which which
2:03:32 thing you're asking about but
2:03:35 um I mean the reason I have these on
2:03:38 screen is these are the things that I
2:03:39 think are best
2:03:41 I think
2:03:42 probably probably
2:03:44 well that is definitely number one and
2:03:47 then you could probably flop
2:03:49 Bing and po they just they're kind of
2:03:53 um the kind of different advantages
2:03:56 but chat gpt4 is
2:03:59 pretty remarkable that's the one I would
2:04:02 use all right I'm at the bottom of the
2:04:04 questions if I missed your question I'm
2:04:05 sorry I'm just an old dude um oh it
2:04:07 looks like we got some new people in
2:04:09 here uh I'm taking off this is the AI
2:04:11 learning lab I talk about AI stuff feel
2:04:13 free to give me a follow
2:04:15 um I try to go live once a day this
2:04:17 weekend I went live four times so I'm
2:04:20 Crispy I can't talk I need some water
2:04:22 water I need water or Doug and Wendy
2:04:26 wine or we've got diverticulitis
2:04:30 that's for the Gen xers out there
2:04:35 all right everybody thank you so much
2:04:38 peace out have a good night