aiLL 9/1/23 - The Present and Future Impact of AI

Description

Chapters

Transcript

0:01 foreign
0:21 [Music]
0:28 flab joined hey flab what's shakin love
0:32 the name Tobias what's happening
0:35 Tobias I gave you a
0:39 uh irregular role on Discord on the AI
0:43 Salon Discord
0:45 and created a uh there's a Irregulars
0:48 channel in there now yay
0:51 what's happening
0:54 I didn't know that Mrs mrs's Discord
0:57 handle so I need to figure out
0:59 everyone's Discord handles
1:03 well welcome Happy
1:06 Friday night man so much has happened
1:07 today
1:08 I thought I was gonna have to have my
1:10 car towed and
1:12 replace the electrical system and none
1:15 of that happened turned out to be a
1:17 shorted out battery so got a new battery
1:20 and the car runs like a champ
1:23 it runs back a champ I felt like a
1:25 regular old
1:28 testosterone filled man working on a car
1:32 popping out the old battery tossing in a
1:35 new one
1:37 it's fantastic
1:39 yeah I know are you okay farmer and
1:41 Discord as well okay farmer
1:44 digging that Turtle Shield yeah Togos
1:46 did that took us uh who's in here he's
1:49 an irregular
1:51 his his English language is irregular
1:53 too he's French
1:57 what's cool though is
1:59 he and I have been starting to send
2:01 messages back and forth I'll send it in
2:03 front I have chat GPT right to French
2:05 for me so he can write he can write in
2:08 his native tongue and I can write in
2:10 mine it's kind of cool
2:15 but it's hard to break the Habit though
2:17 awesome about the car yeah what a relief
2:18 yeah exactly yeah that was that was a
2:21 bit of a nightmare it was freaky man it
2:22 was just like it was like it was
2:23 possessed relays going off wait it's hot
2:26 in here hang on
2:34 all right happy Friday night everybody
2:37 let's let some people get in here let me
2:40 dab off the moisture
2:44 here's the there's the jet engine
2:47 oh yeah New York City
2:51 I miss my New York I'll miss my New York
2:55 what do I oh I have food all over me
2:58 hang on a second
3:22 good Lord dress them up but you can't
3:25 take them out
3:28 all right okay
3:30 I have to contact you about an AI plot
3:33 and scheme let's do it cabrino
3:35 I don't know your Discord handle but I
3:38 just created for the Irregulars in here
3:40 tonight let the other regulars know I
3:42 created a private Discord channel for us
3:45 and I'm giving roles to people that I
3:48 know are
3:50 odd
3:51 and in here so we can chat in there
3:54 that'll make it easier Emilio's wife I
3:56 also don't know your
3:58 um
3:59 whatever you call it handle and Discord
4:01 same with you togas togas look what you
4:04 made
4:07 it's cool isn't it
4:09 for those of you that don't know where
4:11 you are
4:13 at the yeah learning lab
4:15 welcome everybody
4:19 A.I learning lab oh yeah
4:23 tell him what he's won and there's
4:26 courses online people can take for
4:27 certification purposes I don't know my
4:30 handle
4:31 well Dr Jay
4:33 you gotta figure that [ __ ] out
4:35 [Laughter]
4:37 click on Discord this is part of the
4:39 problem with Discord is it kind of
4:41 assumes anyone Everyone's an nft bro
4:43 this is like you'll figure it out man
4:47 but it's completely non-intuitive and
4:49 convoluted and
4:51 I can't follow discussions you can't
4:53 find anything exactly man we don't want
4:56 to let the plebes in
5:01 ah
5:03 tell him what he's won
5:07 you've won a new battery for your car
5:09 because your electrical system is just
5:12 fine
5:14 all right welcome everyone this is the
5:16 AI learning lab it is Friday night let's
5:19 get this thing going
5:22 I don't know who you people think you
5:24 are
5:27 but I like you
5:28 [Laughter]
5:30 um I really do appreciate you showing up
5:32 welcome
5:33 my name is Kyle Shannon uh I ramble here
5:38 I like talking about
5:41 generative Ai and what's coming and
5:43 what's happening and how the hell we
5:45 make sense of all this AI stuff
5:48 um if you're one of those people you're
5:50 like but Kyle
5:52 how do we make money with chat gbt well
5:59 [Music]
5:59 [Applause]
6:01 we make fun of those people
6:03 well not that people
6:05 not people not that people want to make
6:07 money but the people that prey on the
6:09 people that want to make money don't
6:11 like those people
6:14 but we can talk about that we can talk
6:16 about business I'm an entrepreneur I'm
6:19 uh my current gig is I'm a co-founder
6:21 and CEO of a company called storyvine
6:22 which is an automated video storytelling
6:24 platform
6:25 I uh
6:27 I've started a lot of companies in in
6:30 the early days of the world wide web in
6:31 the mid 90s I started one of the first
6:33 digital agencies company called
6:35 agency.com why do I say that because hey
6:37 it's kind of cool but more importantly
6:41 um the early days of the World Wide Web
6:43 are very very it's feel this AI stuff's
6:47 feeling very very familiar uh to the
6:50 early days of the web and that was
6:51 pretty transformative and I think this
6:53 stuff's going to be even more
6:54 transformative the other thing that I'll
6:56 do here is if you're like yeah but Kyle
6:58 I don't know where to start
7:00 everyone's been talking about Ai and I
7:03 would like to get into AI but I don't
7:05 know where to start Kyle what do I do
7:06 what do I do well we're going to take
7:09 care of that we're going to tell you
7:10 what to do
7:11 uh if you have questions comments
7:14 thoughts
7:15 clever trolls you did just don't put in
7:18 a dumb troll because
7:20 we'll make fun of you
7:22 the Irregulars the Irregulars will take
7:25 you down I'll Take You Down
7:28 what's a clever troll you might ask well
7:30 a clever troll is one that seems at
7:32 first like a compliment
7:35 and then the more you read it you're
7:36 like huh
7:37 huh that kind of stings
7:40 thank you for the hearty heart
7:43 um yeah so
7:45 if you're in a regular have at it start
7:48 tapping the screen get your automated
7:50 screen tappers out let the games begin
7:53 all right let's let's jump in here and
7:55 talk about some of this stuff so
7:58 if you have not played with chat gbt it
8:01 is my contention
8:04 that everyone on the planet right now
8:08 that's you know
8:10 that uses computers in any way
8:13 deserves
8:16 to play with these tools long enough to
8:18 have what I call your Kevin McAllister
8:20 moment remember Home Alone
8:24 because
8:25 somewhere between five minutes in and I
8:28 don't know an hour or two in you will
8:31 ask this thing to do something for you
8:34 and it will blow your [ __ ] mind oh I
8:36 also swear and I also make stupid voices
8:38 and act like an idiot and if you want to
8:41 know what my qualifications are none of
8:43 them make sense
8:44 so oh there's the champion chippy
8:49 um
8:52 so this thing this thing will do
8:54 something at some point that will make
8:56 you realize holy crap this is something
8:58 different
8:59 and and that's what these things are so
9:01 this is the official chat GPT website
9:03 Claude and Bing are also things you can
9:05 play with but if you're brand new just
9:07 start here and then if you really don't
9:09 know what to do there and you know
9:11 you've heard people talk about prompting
9:13 and prompt engineering and all this crap
9:15 just go to prompts.chat that thing down
9:17 there it's just a document I'm not
9:19 affiliated with any of these so if you
9:21 don't do it it's no skin off my back but
9:23 I think it is worth your while and just
9:26 go to prompts.chat that'll teach you how
9:28 to talk to these things it'll give you a
9:30 lot of ideas for what you can do with
9:31 them uh and uh yeah and then we I got
9:35 lots of other stuff we can talk about
9:36 and why don't you pop some questions in
9:39 I'll jump in there and do what I can to
9:42 answer everything anything and
9:45 everything oh and for the Irregulars I
9:47 know there's still some Irregulars
9:49 trickling on in
9:51 um on the AI salon so
9:54 I've got this thing called the AI salon
9:56 and this is the website the salon.ai
10:00 we have a meet up every two weeks
10:03 stop Siri please thank you thank you
10:06 thank you very very much Siri I don't
10:08 want to talk tonight okay okay okay okay
10:11 okay
10:12 Siri okay
10:15 all right
10:17 um
10:19 so if you go to the salon.ai that's a
10:22 link tree and that's got a link to what
10:24 the salon's all about what our values
10:26 are
10:28 um
10:29 it's got a link to the Meetup so you can
10:32 sign up for those every other week
10:33 meetings and then it's got a link to the
10:35 Discord
10:36 and for the Irregulars I created a
10:39 private channel on Discord I added as
10:41 many people as I could find that I
10:42 recognized their names but if I don't
10:45 know you
10:46 um
10:47 if I don't if I can't connect your name
10:49 here to the name there I I don't know
10:51 how to do it so just ping me on Discord
10:53 and say I want the I want the irregular
10:56 role and I'll give you the role and then
10:57 you'll have access to that private
10:58 Channel Isn't that cool Isn't that cool
11:00 I said I was gonna do it and I did it
11:02 I'm a man of integrity except for all
11:04 the stuff that I
11:06 um life's tough stuffer if you're stupid
11:08 listen
11:10 I have deep Integrity except for those
11:12 things that I didn't do that I said I
11:14 was going to do but didn't get to them
11:15 yet like the workshop and the t-shirts
11:17 and the everything I've said the many
11:19 many many things I've said but other
11:22 than that deep Integrity deep deep
11:25 levels of Integrity for all the stuff I
11:27 did do 100 integrity
11:30 discounting the things I didn't do which
11:33 is many and deep
11:35 all right move along
11:37 let's get to some questions let's answer
11:40 some questions sometimes people have
11:43 questions
11:45 thank you very much
11:47 it's very nice I like the heart
11:50 like the heart I like a lot
11:54 um
11:55 let's see sometimes people ask questions
11:58 sometimes they comment on my appearance
12:00 because there are things to comment
12:03 about we don't need to talk about it
12:06 unless you comment about it and we're
12:08 going to talk about it
12:13 happy Friday everyone Stevo's here
12:15 welcome Stevo wesn radio got the number
12:19 one gifter badge
12:20 wsn radio send enjoy music rock and roll
12:24 and Hip-Hop
12:26 light easy listening hip-hop
12:29 all right
12:32 thank you we'll go check it out Dr J hey
12:34 would Shake him was shaking
12:37 uh tobiasis plus one is okay farmer I
12:39 know that in here awesome about your car
12:42 what a relief thank you togas
12:44 cinnamon gummies yeah man you know so I
12:48 took some gummies
12:50 I think I took too many man
12:54 what oh they're just Haribo
12:57 they're just regular gummies man I feel
13:00 so high
13:04 uh unsure if I had had Discord but we'll
13:09 fix her up all right
13:12 um I have to contact you okay the AI
13:14 plot and scheme I like it was take over
13:17 the world we can do this
13:19 oh wait let me put on I want to put on
13:21 tennis Djokovic is playing right now in
13:24 the U.S open
13:26 we might as well multitask we have ADHD
13:29 why would you do one thing at once who
13:31 are you weird people with your
13:33 monotasking
13:36 monotasking
13:41 all right let's see
13:44 all right
13:46 there we go Coco golf just won
13:49 Coco cuff yeah no is it golf yeah I
13:52 don't I don't forget I think
13:55 uh
13:56 I don't want to miss the end perfect are
14:00 there courses online one can take for
14:01 certification purposes there are
14:04 starting to be some here's the here's
14:06 the deal about uh
14:08 workshops and learning right now and I
14:10 know I know folks in here want me to do
14:12 a workshop I'm going to do a workshop at
14:13 some point I'm probably going to work on
14:15 it this weekend
14:16 um
14:18 any sort of any certification you get
14:21 right now any course that you take right
14:23 now anything that you teach yourself
14:25 right now is likely going to be
14:28 Irrelevant in
14:31 four months
14:33 this stuff is moving so fast it's moving
14:36 so fast that all of the stuff that is
14:39 critical skills today critical skills
14:42 for your AI career all of that [ __ ]
14:45 it's all changing all the time so what I
14:48 so I I don't know that any certification
14:51 has any particular gravity these days if
14:55 if you're if you're on the the
14:57 engineering and machine learning side of
15:00 the house then yeah then then you know
15:03 go to Stanford go to Harvard go go get
15:05 those kind of certificates or degrees
15:07 that's super valuable but with this
15:09 generative AI stuff with chat gbt and
15:12 the image generation tools and all this
15:13 new stuff that is is getting merged into
15:16 our tools uh Google yesterday I think uh
15:21 released uh or announced General
15:24 availability of duet
15:26 so all of their barred AI [ __ ] they've
15:29 announced General availability Within
15:33 their office suite and uh Meetup beats
15:37 me you know meet.google.com
15:40 um so I just applied for that for my
15:42 company today so I don't know I don't
15:44 know where I'm going to get it but
15:46 that's coming so this stuff's coming
15:47 it's changing fast so what I would say
15:50 is
15:51 if you find someone that you like
15:53 someone that you trust if if it's you
15:56 know not ridiculous amounts of money and
15:58 if they're not making all sorts of
15:59 promises like
16:01 I thought he was
16:03 legit avoid those people like the
16:05 [ __ ] plague
16:07 um
16:08 what I would say is
16:11 take one of those courses to to get into
16:14 a curious adventurous mindset that's the
16:18 biggest thing you can do right now is
16:20 just start to open yourself up to
16:22 these things force you to work in very
16:25 different ways it's just it's I it's
16:27 essentially impossible to explain
16:31 how different it is to work with these
16:34 tools than everything we've done in
16:36 history thank you share bear
16:38 um
16:40 so so if you find a course you like go
16:43 for it except that anything you learn
16:45 there's going to be irrelevant within
16:46 six months and just get into that
16:50 Curiosity mindset and what it what what
16:52 that does for me is I I don't I don't
16:55 think it's necessary to take a course
16:57 right now I do think it's necessary to
16:59 just start exploring start exploring
17:00 start playing figure out what you're
17:02 good at figure out what these things are
17:04 good at what they're not good at where
17:05 those things overlap and at some point
17:08 in there you'll go oh
17:09 it's really good at this or I'm really
17:11 good at this with that
17:13 that's the thing to do that's your thing
17:15 to do that's it who are you calling odd
17:18 oddly I'm Emilio's wife in Discord I
17:21 didn't see you in there
17:22 but okay I'll look for you I'll find you
17:25 um I just eavesdrop on Discord well
17:28 that's okay if let me know I'll I'll
17:30 give you the uh Irregulars role and you
17:33 can join the private Channel hey buddy
17:35 hey Irregulars what's happening Stevo
17:37 Stevo I think I gave the the role does
17:40 anybody know if the texts
17:42 be featured in ideolab needs to be in
17:46 quotes I think it works works better if
17:48 it's in quotes to bias so yeah so let's
17:50 let's go show let's go
17:54 play with what Tobias is asking we can
17:56 figure this out right now we can do it
17:59 eat meat Earl Audio Lab or ideogram so
18:04 if you have not seen ideogram
18:07 so this was an ex googler
18:11 that raised
18:16 16.5 million dollars
18:18 to create ideogram and ideogram does
18:21 words it does words so like all of the
18:25 images that
18:28 you see here
18:31 these were all done in ideogram
18:35 so all that text now
18:38 is not always good like these are these
18:41 are the edits
18:44 that was done in ideogram I'm not weird
18:47 I'm a regular
18:48 [Laughter]
18:52 wow
18:56 all right so let's copy
18:59 let's copy this dinosaur here
19:03 so what you can do in ideogram
19:05 is you can just click on any image you
19:08 can see the whole
19:09 prompt right there you can also remix it
19:12 so I can just use this as the base image
19:14 and then do iterations of that but I'm
19:17 going to go ahead and copy
19:20 copy The Prompt and paste in a new one
19:23 let's see an outline Vector black and
19:25 white cartoon this is going to be Turtle
19:26 wait zombie Turtle zombie
19:31 turtle
19:33 that's an inside joke but if you hang
19:35 out here long enough you'll see what the
19:37 hell we're talking about it's very
19:39 clever oh we're clever in here
19:42 um sticker adorable and cute Kawaii
19:46 outline vector graphics written and then
19:48 quote marks I love music so yeah in
19:51 quotes to bias uh so I'm gonna put I'm
19:55 your regular
20:02 painting vibrant poster conceptual art
20:04 so we're gonna do poster typography
20:10 um
20:11 uh that'll go illustration vibrant
20:16 that's good all right and then we're
20:18 gonna go do we want it Square yeah we'll
20:20 do Square generate
20:24 and then we hopefully will see some
20:26 words
20:28 that are what we wanted to say now if
20:30 you've ever tried to make words and
20:32 stable diffusion or mid-journey almost
20:35 impossible there's one exception to that
20:37 the Doodle Draw on clipdrop.co which is
20:41 also stable diffusion it does letters
20:43 pretty well but not as well as this
20:45 thing does
20:48 yeah
20:49 all right so so there's there's some
20:52 Irregulars
20:53 I'm irregular blob I am ERG relar
21:00 [Laughter]
21:06 that's good so you can come in so if you
21:08 get something this is pretty close so
21:09 we'll play with this one let me see
21:11 we'll just remix this one and see if we
21:12 can get it to
21:14 let's get rid of painting and poster
21:18 oh and conceptual art let's just do
21:20 those three all right let's see if this
21:21 gets us any better
21:24 I haven't played with this enough to
21:26 really know what what are the things
21:28 that are are better
21:30 at you know making making this stuff
21:32 work
21:35 uh that one looks close that one
21:38 actually no not quite they're all close
21:43 I'm a redular
21:48 I'm a repular I'm irregular look we got
21:52 one yay
21:54 yay we did it
21:57 you win
22:00 yeah I like turtles
22:02 I like turtles
22:10 I think that deserves a little Victory
22:13 don't you a little Victory lap back here
22:16 live at the waterfront village with my
22:18 friend the zombie Jonathan you're
22:20 looking good Jonathan just got an
22:22 awesome face paint job what do you think
22:24 I like turtles
22:26 all right you're a great zombie good
22:29 time here at the waterfront Village open
22:31 for the next 11 days
22:37 it's good every time I'm a regular
22:42 oh so good
22:45 all right all right people let's keep
22:47 this moving let's Cape this show on the
22:50 road shall we
22:52 what'd you do Friday night I don't want
22:54 to talk about it it's a little
22:56 embarrassing
22:58 oh no it's not you can tell me I'm your
23:00 best friend
23:01 I watched this old guy on Tick Tock try
23:03 to learn how to use it
23:05 oh that sounds awful
23:07 I know
23:09 that's what I like
23:11 I don't think we can be friends anymore
23:13 I know I figured that would happen
23:16 all right
23:19 uh it's a penny of The Irregular Emperor
23:21 with some
23:23 wait some of your
23:25 aliatori genetic
23:28 nice oh nice yes this image
23:33 now you go
23:36 it's an irregular Penny don't you see
23:39 I like it
23:41 I like it a lot I like it I like it I
23:44 like it like so many bit Corners coiners
23:47 who have coin saved on Old hard drives
23:49 and forget their safety info exactly uh
23:53 should you pay for power just gpt4 chat
23:56 gpt4
23:58 I used to pay for
24:00 po
24:02 I think I'm still paying for Poe but I I
24:04 need to cancel it
24:06 um
24:08 I would not pay for Poe right now I
24:10 don't see the advantage of it
24:14 you can
24:17 um if you pay for Poe you get
24:19 Claude Claude and Claude two
24:23 hundred thousand token things but you
24:25 can get them for free on cloud.ai
24:28 um the bot maker you can use for free
24:30 even if you don't pay
24:32 so
24:34 I wouldn't pay for Poe
24:38 um this thing's gonna start charging
24:40 soon I would imagine they're gonna be
24:42 free for a while because they want to
24:43 get a bunch of users but um I pay for
24:46 mid-journey I pay for chat GPT gpt4
24:53 I paid for zapier I pay for
24:58 um
25:00 Google collab but I haven't paid for
25:02 that in a while
25:03 I'm probably gonna pay for teams
25:08 at work because most of my Pharma
25:10 clients use teams and it's going to get
25:12 some AI stuff pretty soon
25:14 we pay for Google
25:16 work stuff like the office suite
25:22 I think that's about it oh I paid for I
25:24 pay for uh
25:26 11 labs and D- ID for The Voice the
25:31 synthetic voice and the animated people
25:33 sort of thing so those are the things I
25:35 pay for I wouldn't pay for Poe
25:39 um
25:42 what do you think will be after AI
25:46 what makes you think we're going to be
25:49 here after AI young man
25:55 um
25:56 I so
25:58 I don't know that there's an after AI
26:01 I think AI is
26:04 is a profound shift in how we use
26:07 computers how we communicate how we
26:10 create
26:12 um
26:17 you know like like I almost feel like
26:19 the question is kind of the equivalent
26:21 of
26:23 um
26:25 you know what do I think will be after
26:26 the World Wide Web like ask it asking me
26:28 that in 1994.
26:31 I can't foresee us ever not having the
26:33 web and the internet all this
26:35 interconnected stuff I can't ever see us
26:38 not having
26:39 you know some flavor of Chachi PT and
26:42 it's not even chat gbt that's like I
26:45 actually feel chechi BT is kind of a it
26:47 is by definition a limiting
26:52 um name
26:55 right because it kind of is it it kind
26:58 of implies this right you type in words
27:00 like a Google Search and you get back
27:02 text
27:04 um these things are very very quickly
27:06 gonna get non they're gonna get
27:09 multimodal you'll be able to talk into
27:11 them put images into them shoot live
27:14 video into them pull all that stuff out
27:17 music out it's
27:20 it's gonna be as ubiquitous as IP into
27:24 it Internet Protocol
27:26 it's gonna be everywhere it's going to
27:28 be all software is going to have this
27:31 layer this veneer of magic
27:35 it's like literal [ __ ] pixie dust
27:39 sitting on top of all of these complex
27:42 systems that we've spent the past 70 or
27:44 80 years building
27:47 and then this magic pixie dust is gonna
27:49 make it easy fun it's gonna automate the
27:52 boring [ __ ] it's gonna make you know
27:55 everything easier
27:57 and so I don't I don't think there is
27:59 life after AI I think there's just
28:01 different levels of it
28:03 you know when the World Wide Web first
28:05 came out it was very very simple you
28:07 couldn't even Center images
28:10 you want to see something cool
28:12 this was
28:21 um
28:23 [Music]
28:30 [Music]
28:34 oh that was it
28:41 so oh
28:43 where
28:50 why can I not
28:53 oh I see
28:56 there we go
29:00 this was the
29:05 um
29:06 first home page
29:09 of an online Zine I did in in 1994
29:13 called Urban desires so it's just white
29:15 background black text with blue
29:18 hyperlinks and the cover of the magazine
29:22 which was that but you had to click on
29:25 it to get to it the cover of the
29:27 magazine
29:28 was left Justified because the center
29:30 tag didn't work on images
29:35 so
29:36 we're right now with AI
29:40 like this
29:43 so if you think about the internet today
29:45 and like how sophisticated it is on your
29:48 phone and how sophisticated it is and
29:50 we've got streaming services and we've
29:52 got YouTube and we've got social
29:54 networks and we've got streaming
29:56 services and we've got YouTube and we've
29:58 got social networks and we've got all
29:59 that [ __ ] none of that existed it was
30:02 this
30:03 it was blue hyperlinks on a white page
30:06 with black text and left Justified
30:07 images that's what the World Wide Web
30:09 was
30:12 that's where we are with generative AI
30:14 right now
30:15 as sophisticated as it is it's the
30:18 equivalent of this in terms of its in
30:21 terms of what it's going to be so what's
30:23 coming is going to blow our [ __ ]
30:25 minds and I think I think we've got a
30:27 big one coming I think so Google is
30:31 um
30:32 announcing Gemini hopefully launching
30:35 Gemini this fall so they've got their
30:37 barred thing right the new model is
30:40 called Gemini it's multimodal from what
30:43 I understand and they're claiming it's
30:46 five times better than chat GPT now
30:51 um Sam Altman goes that's just the
30:53 marketing team and Sam Altman's right
30:54 but
30:57 what does Google have
30:59 well Google's got everything we [ __ ]
31:02 searched for and done and written for
31:05 the past 20 years
31:07 they've got YouTube
31:09 they've got they've got all the data
31:12 they've they've got the entire [ __ ]
31:14 web index so the depth of the data that
31:16 they've got is
31:19 I Can Only Imagine
31:22 it is beyond our our ability to
31:25 comprehend how far beyond our
31:26 comprehension it is
31:28 so now
31:31 what's the challenge with Google
31:34 they're Google they're they're they're
31:35 engineering heavy and Humanities life
31:38 right
31:40 um open AI has more of a feeling kind of
31:43 like apple did in the early days where
31:44 it's this balance of arts and human or
31:46 you know Arts and Sciences right Google
31:49 all science very little art right
31:52 these are large language models so
31:54 that's a Humanity
31:56 um so we'll see so so they can screw it
31:58 up and they've you know they they've had
32:00 some pretty embarrassingly bad
32:02 technology for the beginning part of
32:04 this year but it's gotten better
32:06 so big stuff is coming but this is this
32:09 is where we are so if you weren't around
32:12 to see the internet when it looked like
32:14 this that's what it looked like and
32:16 that's that's what we're here's here's
32:18 what it sounded like
32:20 when when I made the first issue of
32:23 urban desires so I was living in an
32:24 apartment in Brooklyn and you know I
32:27 came up with this I was super happy
32:28 about it I'm like how do I get it on the
32:29 internet and someone someone sent me
32:31 back a note and they said
32:33 um there's this computer in California
32:35 here's the phone number for it you gotta
32:37 stick your files there it was it was an
32:40 internet server and so this is what it
32:42 sounded like when I called that computer
32:45 foreign
32:47 [Music]
32:54 that was the internet you had to call it
32:58 and they charged you by the minute it
33:01 was [ __ ] awful you kids these days
33:03 you don't appreciate what we went
33:05 through you have no idea
33:11 none
33:13 you little whippersnappers in greats get
33:18 off my lawn
33:19 so anyway that was a long answer your
33:21 question fourteen thousand
33:23 uh but if I talk French Canadian it'll
33:27 be worse oh until Gus getting in the
33:29 getting in the digs to Canada I like it
33:32 it's solid
33:33 troll the northern border
33:36 Kyle Gets behind on questions so
33:39 Irregulars will help up answer questions
33:40 too that is true thank you Bruno for
33:42 doing that oh I'm 99 questions behind
33:45 already uh let's see well and I'm just
33:48 gonna keep going how I'm going I'm just
33:50 gonna keep answering questions so I'm
33:51 more than 99 questions behind so if
33:54 you're bitter if you're a bitter type
33:56 and listen I relate bitter you betcha
33:58 that's gen xers we're a little
34:01 just just right right there it's always
34:04 right there
34:06 um so if you're like that too and you're
34:07 like well I won't answer my questions
34:09 I'm special
34:12 um I will
34:14 but you got to be patient
34:17 and if if
34:19 it is mystifying you why all of these
34:22 hearts are flying up the screen it
34:24 mystifies me too but here's what we've
34:26 learned the Irregulars the ones who show
34:28 up to these lives regularly
34:31 um they get bored they have add this is
34:35 a whole this is a whole crowd there's
34:37 some neuro spiciness going on with
34:39 everything about this
34:42 and so they get bored and so they try to
34:45 top themselves in terms of how many
34:46 likes we get so that's what that is so
34:50 just I'll get to your question just just
34:53 pour yourself a whiskey
34:55 sit back watch Djokovic he's behind but
34:59 he's Djokovic you know he's gonna
35:00 [ __ ] turn it around
35:03 all right champ yeah champs back well
35:05 Champs fat and
35:08 taking a well-deserved rest after all of
35:11 that eating
35:12 damn it I just I just snarfed
35:17 Seltzer in my eye
35:22 you can't make fat dog jokes and not
35:24 expect to laugh you can't
35:28 Claude is really good but sometimes
35:30 doesn't work I know well so all of these
35:32 things so here the reason that I have
35:35 um I haven't talked about all the
35:37 different things let me let me do this I
35:38 think it's good a good level setting for
35:41 the folks that are here
35:43 okay
35:45 if you're brand new to uh generative AI
35:47 if you've not played with chat gbt
35:50 there's no shame here but damn it get
35:52 thee to a nunnery right now get the to
35:56 chat.openai.com that's the official chat
35:58 GPT website and start playing and you're
36:00 like but Kyle what do I do I don't know
36:02 what to do just tell it to write you a
36:04 country song about a fat dog
36:06 and just watch what it does and then try
36:09 to stump it and just watch what it does
36:11 and then download prompts.chat or don't
36:15 download it go to prompts.chat it's a
36:17 document and it'll teach you how to talk
36:19 to these things how they work what they
36:20 are how to talk to them it's really
36:22 powerful document and then it's it's
36:24 just a good starting off point
36:26 so chat apt is really good gpt4 the one
36:30 you pay for is is the best of them out
36:33 there it's possible Gemini comes along
36:35 and knocks them off the hill but right
36:36 now it's the best it's also got this
36:39 thing called Advanced data analytics
36:41 used to be called code interpreter that
36:43 is [ __ ] mind-blowing so that's that
36:46 Bing is Microsoft's offering of chat
36:50 gpt4 it's connected to the internet it
36:53 can make images it can read images uh
36:56 and it's got too many safety guard rails
36:58 on it so it annoys the [ __ ] out of me
37:00 but it is connected to the internet
37:02 Claude is anthropics
37:05 um competitor to chat GPT
37:08 the big selling point with Claude is
37:10 it's got a hundred thousand token
37:12 context window and you're like but Kyle
37:14 what's that mean
37:16 doesn't really [ __ ] matter what it
37:18 means is Geeks
37:20 count tokens and humans count Words so
37:23 it's about 75 000 words it's got a
37:26 window of about 75 000 words it can
37:29 process in any given conversation so
37:32 like instantly so you could paste in a
37:35 50 000 word novel into Claude
37:39 and tell it to write you an epilogue for
37:41 that novel and it will
37:44 and you can say write an epilogue for
37:46 this and keep it in the tone of the
37:47 novel or you know write it in the first
37:50 person of the you know the protagonist
37:53 that's in the third person in the novel
37:54 or whatever it might be all right oh so
37:58 so that's that and yet Claude sometimes
38:01 does sometimes doesn't work so these
38:03 three I consider kind of the three my
38:05 three go-to's
38:06 I'm I'm kind of on the edge of flipping
38:09 out
38:10 um Bing for bard
38:14 because I think Bard is a little more
38:16 useful than Bing Bing the safety guard
38:18 rails on Bing just just annoy the [ __ ]
38:20 out of me but what I in what I find
38:22 myself ending up doing is I'll bounce
38:24 between these three when I'm when I'm
38:26 when I get stuck on one and it's not
38:28 giving me what what I want I'll just pop
38:30 over to one of the other ones and just
38:32 try recreating the whole prompt chain
38:34 that I did in the one that was failing
38:36 because they're all kind of different so
38:38 that's a good way to play with these
38:40 um these four are four other options so
38:43 Bard that's the one I mentioned that's
38:44 Google's labs.perplexity is good for
38:47 research and citations uh well no
38:50 perplexity.ai is good for research and
38:52 citations Labs dot perplexity is they're
38:56 hosting meta's llama model for all of us
38:59 just to play with so that's the open
39:01 source model that meta put out
39:04 um there's code llama and then three
39:06 flavors of llama two
39:09 um there and you can play with them for
39:11 free and if you like them you can go
39:13 make your own servers and do all sorts
39:15 of happy horse [ __ ] and then Pi dot AI
39:17 is chatty GPT it's a it's a
39:21 conversational chat bot so if you put it
39:23 in voice mode and talk to it it's kind
39:25 of mind-blowing and then Poe is uh you
39:30 can play with different models there you
39:31 can create Bots and things like that all
39:33 right so that's that and then on the
39:35 image front these are the biggies that I
39:38 think are worth your attention on the
39:40 image front mid journey is the best
39:42 but it doesn't do text
39:45 ideogram
39:47 is not great with images it's kind of
39:49 like stable diffusion was a year ago or
39:52 dolly is today it's not great
39:55 but it can do letters it can do words
39:56 all of
39:58 this stuff all of these words were done
40:01 in ideogram
40:02 okay so that's
40:05 and it's like seven days old seven days
40:07 old like why this channel like why I
40:11 have this up here is as new [ __ ] comes
40:13 up I I swap [ __ ] out
40:15 um I used to have Adobe here
40:18 a seven day old company
40:21 booted Adobe off my list
40:24 um Leonardo and clip drop are from
40:26 stable diffusion well clip drop is from
40:28 stable diffusion and then Leonardo is a
40:31 web front end in front of stable
40:32 diffusion and so if you're geeky go play
40:36 with stable diffusion on automatic 11 11
40:38 these are the non-geeky options and then
40:41 the prompt book is a thing that'll teach
40:43 you how to talk to these image tools and
40:45 future PD is a big old list of list of
40:47 stuff okay all right
40:53 I don't know what that was about there
40:56 was some sort of message there
40:59 [Music]
41:03 let's see let's get a pretty picture
41:07 let's do that one I like that one
41:12 it looks like a rocket remember the
41:14 first of the Starship the little first
41:17 section that Elon Musk tested it looked
41:19 like this
41:20 this looked like a silo flying up in the
41:23 air
41:27 look at that look out look how shiny and
41:29 nice that is oh yeah that's really nice
41:31 Kyle you made a nice little Silo there
41:34 I'm sure it's full of corn Corn's tasty
41:37 no love for Bard yeah I just love bard
41:41 yeah Pete Bard's on the list now they're
41:45 a second on the list you know what Pate
41:47 in honor of you what we're gonna do
41:49 right now
41:54 um
42:06 [Music]
42:09 Bing just got knocked to page two
42:14 in honor of Mr pate
42:17 Mr Pete Mata PhD hello hate model PhD
42:23 I work for the Google
42:26 I don't know if you know the Google
42:29 we're rather large we're rather
42:30 impressive people don't like us we said
42:33 we would do no evil and then we became
42:36 evil
42:39 anyway uh and we don't like to talk
42:41 about that it's it's not in the the
42:43 corporate handbook says we should we
42:45 should not acknowledge that reality but
42:49 I think
42:52 the reality of my
42:55 experience with this [ __ ]
43:00 oh because there's is there something
43:01 over here
43:03 all right oh yeah uh
43:07 what is going on
43:12 huh weird whatever
43:15 so we got Claude
43:18 so I still think Bart's third I think I
43:20 think Claude out impresses
43:25 out impresses Bard but Bard makes the
43:28 top three
43:30 how's that does that work I think that
43:33 works
43:34 I think that works
43:40 here we come to save the day
43:43 all right Pete I hope you're still here
43:45 and you got to see the the real-time
43:47 interaction where where uh where yield
43:51 yield Bard made the top three
43:55 I'm kind of doing that in anticipation
43:57 of Gemini so Gemini better not suck
44:00 laughs
44:03 because I've heard I've heard Google Now
44:05 on multiple occasions ago there's big
44:07 stuff coming out and then they put out
44:09 something embarrassing so let's not do
44:11 that but I think that's right I think
44:14 this is right
44:15 and Bing is relegated to the uh
44:18 to the Kids Table
44:25 all right there you go
44:28 um
44:34 look there
44:36 love for bard
44:38 I did it see
44:40 I'm responding in real time to the needs
44:43 of all of you
44:45 well not the needs but it was just time
44:47 wasn't it
44:49 I think
44:50 Bart took a licking from this channel
44:52 for a good long time so I think it's
44:54 okay
44:57 I think one of the things that's nice
44:59 about Bard is it's the way it integrates
45:03 um
45:04 hyperlinks from the web into its answers
45:06 it's actually quite sophisticated and
45:08 good and not spammy and advertising I'll
45:12 tell you what pate if you have if you
45:14 have any influence over there at the
45:16 GOOG
45:19 if Google
45:22 [ __ ] up GPT answers with too much paid
45:26 injection
45:28 You're Gonna Lose everyone
45:30 because search has gotten
45:33 essentially unusable because it's just
45:35 it's just too much paid crap
45:37 so if that happens to your GPT answers I
45:40 you're gonna you're gonna
45:42 you're gonna blow an opportunity
45:46 um so hello Irregulars from the hot tub
45:49 AI Channel trying not to drop the phone
45:51 in the water you gotta go you gotta get
45:53 an iPhone there you can I take this one
45:55 in the shower with me
45:57 although it is disconcerting because
45:59 with the with the touch screen when
46:01 water droplets hits hit the touch screen
46:03 it sometimes you know does [ __ ] on the
46:05 interface so sometimes it'll turn on
46:08 it'll turn on the camera It'll like flip
46:10 the camera and I'm watching myself there
46:13 in the camera like ah
46:15 it's not pretty
46:17 just trying to watch tick tock
46:18 [Laughter]
46:20 uh that's funny recovering from back
46:24 surgery but God can't miss Kyle off
46:26 Tobias I'm so sorry back surgery is no
46:29 [ __ ] joke
46:30 I get the bad back every now and again
46:32 oh miserable
46:38 hey Kyle I'm continuing with your
46:41 training here
46:42 here comes the cowboy hat thank you I
46:46 know it's good I'm getting better I
46:47 don't I don't fumble as much when the
46:49 cowboy hat comes on yeah life's tough
46:52 it's tougher if you're stupid see it's
46:54 just too short
46:56 I'll get there I'll get a full line in
46:59 Life's Too life's tough it's tougher if
47:01 you're stupid is that the quote I think
47:03 that's the quote let's go find out add
47:06 moment hold
47:08 uh John all right damn it dang it I
47:11 wasn't looking I mean
47:13 life's
47:18 tough
47:20 yeah
47:21 life's tough but it's tougher when
47:23 you're stupid
47:25 you're stupid all right all right we got
47:28 it
47:29 okay all right fine fine fine oh it just
47:32 jumped me to the bottom there we go Tick
47:35 Tock well I did not spell oh life's
47:37 tough stuff when you're stupid oh yes
47:39 victory
47:41 [Laughter]
47:45 mm-hmm
47:50 I don't have to brag but oh thank you
47:52 very much for the butterflies Ted looks
47:54 Ted looks tell me about her Flats I feel
47:56 so light and fluffy
47:58 it feels like marshmallow
48:00 oh Djokovic lost set number one
48:04 all right Joker
48:06 get your head on
48:08 get your head screwed on
48:11 he's playing another Serbian
48:14 I think I got a splinter dang it
48:18 duet uh brother 52 wait I want to find a
48:22 decent way to get my students to cite
48:23 GPT in their submitted work duet could
48:26 be a way
48:28 um
48:31 bard
48:33 is good for it's connected to the
48:36 internet Bing is connected to the
48:38 internet
48:39 both of them do citations
48:41 um be careful with them they still
48:43 hallucinate even though they're
48:45 connected to the internet you're like
48:47 give me some citations
48:48 you gotta click on this [ __ ]
48:51 if it gives you a citation make sure it
48:53 exists
48:55 um the one that I think is is
48:57 potentially interesting for your
48:59 students is perplexity.ai but I don't
49:01 know that it's free
49:03 um not
49:04 labs.perplexity.ai but just go to
49:06 perplexity.ai because you can limit it
49:09 you can say I just want you to search
49:10 Wikipedia or this or that you can kind
49:12 of limit your searches you can focus it
49:14 on certain parts of the internet
49:17 it's not really a chat GPT like thing
49:20 it's more it's like it's somewhere
49:21 between search and GPT I don't quite get
49:23 it yet I just haven't spent enough time
49:25 with it
49:26 um but that that might be an option so I
49:28 would say Bing Bard and perplexity are
49:32 are three to explore for students doing
49:35 citations
49:36 um
49:37 there's a woman on Tick Tock her name's
49:39 Gibson like I am Gibson she's a she's an
49:41 English teacher she's got glasses she's
49:44 super sweet
49:46 um and she kind of dove into the chat
49:48 EBT thing way back in December January
49:51 of last year
49:52 and
49:54 um she just did a video and talked about
49:58 um for her senior English class she's
50:01 bringing GPT Chachi PT into the
50:03 classroom like full on into the
50:05 classroom and and there's two projects
50:08 that seniors have to do
50:11 um for the you know for the semester for
50:12 the year
50:14 um one of them is write a paper with
50:16 chat GPT and one of them is write a
50:18 paper without it
50:20 and then you know the the whole thing is
50:22 going to be them using critical thinking
50:25 back to critical thinking to determine
50:28 what was chat GPT good at what wasn't it
50:30 good at how did you prompt it in such a
50:32 way to get better results just right
50:34 using the critical thinking and then how
50:37 did that change how you wrote the
50:38 regular paper so if you don't if you
50:41 haven't found her search for Gibson I
50:43 think I think her name is like I am
50:44 Gibson or Gibson is here or it's some
50:47 weird thing like that
50:49 um but just look up English teacher chat
50:51 GPT and you'll probably find her she's
50:54 great
50:55 she's great
50:58 um duet could be another option if if
51:01 your school has the Google office suite
51:03 and Google grants grants it to you like
51:06 I don't know how long their backlog is
51:07 to grant that you actually have to ask
51:09 them to give you free access
51:11 um but that could that could be
51:12 something because then it's built into
51:15 um
51:16 you know sheets and docs and the slides
51:20 whatever they're called they call their
51:22 presentation software so I found Claude
51:25 to be easy to work with but gives a lot
51:26 of inaccurate information yep it does
51:28 all of them do Luna stick there's
51:30 there's no
51:32 the the only thing that that gets
51:38 that gets close is when you take a
51:40 source document and analyze it so if you
51:43 go into
51:44 code interpreter or Advanced data
51:46 analytics or if you use a plug-in on a
51:49 PDF or if you're in Claude and you copy
51:52 and paste a big long document
51:54 and then ask it questions about the
51:56 document it tends to be really good at
51:58 that it doesn't seem to hallucinate too
52:00 much
52:01 when it when it just has a clear source
52:03 file
52:04 but anything you're asking kind of kind
52:07 of of the general internet it will get
52:09 pretty hallucinating
52:11 um
52:13 the thing that helps that
52:15 is if you haven't gone to prompts.chat
52:18 just go there and you know the the
52:20 Persona things where you tell chat GPT
52:22 to act like a certain thing and respond
52:24 in a certain way
52:26 and have back and forth conversations
52:28 with it the more sort of context you can
52:30 round out to give it to narrow its data
52:33 set the better it seems to get so that
52:36 may be a a tip but yeah all of these
52:38 things right now again back to my my
52:40 idea of you know
52:42 generative AI today is like web pages
52:45 were in 1994
52:47 um they're just they're just janky
52:50 pieces of [ __ ] right now so so you have
52:52 to act accordingly and you know humans
52:55 are still very very necessary with these
52:58 tools you know it's funny though like a
53:00 lot of times I'll hear I'll hear people
53:03 talk about
53:04 um
53:05 you know oh they hallucinating you know
53:07 we're gonna always need humans it'll
53:09 never be able to write as good as we do
53:11 um
53:13 all right
53:17 yeah yeah I don't know maybe
53:21 [Laughter]
53:24 maybe it's possible possible Bill never
53:28 write as good as we do it's also
53:30 possible they will and a lot better
53:33 so they just are where they are right
53:35 now which AI app is best for voice to
53:38 text voice to text inputting
53:42 um
53:45 whisper so open a eyes Whispers really
53:50 good so I've got an 11 year old video
53:53 company we've never had automated
53:54 transcriptions until about a year ago
53:56 because of whisper
53:59 um
53:59 it's finally good enough that it doesn't
54:01 suck and and that my criteria was uh it
54:05 should not suck
54:08 it's still not perfect like you still
54:10 have to fix it but it's at least [ __ ]
54:12 in the neighborhood
54:14 um so there's a couple of ways you can
54:15 play with whisper you can download the
54:17 Open AI app on iOS and I'm pretty sure
54:20 it's available for Android now and then
54:23 you can do their kind of voice
54:24 transcription it's really good it's
54:27 better than Apple's Voice to Text
54:28 transcription the other thing you can do
54:32 is you can go to the open AI playground
54:38 so the the URL for this is platform
54:43 dot openai.com playground
54:54 oh that's funny it doesn't exist in in
54:57 the chat interface
55:07 [Music]
55:17 all right so
55:21 this is this is open ai's playground so
55:24 this is where developers come to play
55:27 um
55:29 hang on sorry
55:34 so
55:35 platform.openai.com playground and then
55:38 this is this is where you come and then
55:40 over on the right hand side these are
55:42 the different modes you can put it in it
55:44 defaults now to chat mode where you've
55:46 got kind of a system message like I want
55:48 you to act like a you know screenwriter
55:51 and then over here is where you do your
55:54 your prompt engineering
55:56 um if you flip this to complete instead
56:00 of chat it's just it's the old way chat
56:02 GPT used to be in the upper right hand
56:05 corner there's a little microphone and
56:08 if you click that microphone
56:10 you can upload audio Yeah audio files
56:13 and it will generate the text of them
56:15 now you can't they can't be super long
56:17 so you might need to chop it up into
56:19 bits but this is this is whisper
56:22 and you can you can access whisper I'm
56:25 pretty sure from zapier things like that
56:27 so
56:28 here or in the open AI app there are
56:31 probably some others but this one's just
56:33 really really good this is the first one
56:35 that I've seen that it doesn't suck
56:38 because they're all they're all a little
56:40 special they're all a little special
56:42 you know they're what I would call neuro
56:44 atypical
56:47 they all purport to be Fantastical and
56:51 then you use them and you're like I'm
56:53 sorry is this a piece of poo poo or is
56:55 it me
56:59 and most of them are a piece of poo poo
57:03 that's a technical term here at the AI
57:05 learning lab
57:11 all right it was a need
57:13 Firefly just a disclaimer I don't work
57:16 on Bard I work on performance for all
57:18 models oh that's cool Pete no yeah I
57:21 it's cool that you're actually working
57:23 on performance for all models so you
57:25 know something about Gemini I want to
57:27 know what's happening with Gemini
57:30 is it trained on YouTube is it trained
57:32 on everything we've put into Google for
57:34 the past 80 years come on hey give it up
57:37 tell us
57:39 Thomas
57:41 anyway it's cool but you can tell The
57:44 Bard team that Bard moved from the from
57:46 the kids table with the AI learning lab
57:49 thanks to your input the goog
57:52 the search interface is trash trash
57:57 Pew research only 18 of use chat gbt we
58:00 are way ahead of the curve yep Ted looks
58:02 we are way ahead of the curve way way
58:05 way ahead this is early early
58:10 we have
58:12 amongst us little chickens here in my
58:15 opinion
58:17 12 to 18 months
58:19 where you can kind of dick around
58:21 experiment learn and you have to because
58:23 the tools are so raw right now that's
58:25 just the deal
58:27 there's so much opportunity for anyone
58:30 that just all you have to do right now
58:32 is be curious and adventurous and start
58:34 playing and start talking about what
58:36 you're doing tell people what you're
58:38 doing
58:39 um Luna stick from in here if you're
58:41 still in here Luna stick
58:43 um I was I was flipping through Tick
58:45 Tock today
58:46 and there was Luna stick and she's like
58:49 hey like have you heard of chat GPT if
58:53 you haven't and she's got a comedy
58:55 audience right or I don't know quite
58:56 what her audience is but it's like it
58:58 ain't a chat GPT audience and she's like
59:01 if you don't know about this I'm going
59:03 to tell it to you you know straight from
59:05 a writer's perspective and and she's
59:07 she's putting it out there like it's a
59:09 big deal it's it's a big deal to do that
59:12 and so that that was really cool uh
59:14 seeing that and it was the first one and
59:17 she's like I'm gonna do more like it's
59:18 gonna be a whole series I'm gonna tell
59:20 you what this shit's all about
59:22 um
59:23 the opportunities
59:26 like both professionally and personally
59:28 for all of us by playing early are just
59:30 they're going to be massive because
59:32 you're going to be aware I'm gonna be
59:34 aware
59:35 of
59:37 when new stuff comes
59:39 like here's the deal if you know how
59:42 limited these things are if you know
59:44 what the limitations are you know what
59:46 you want it to do like for example
59:48 mid-journey was [ __ ] amazing and you
59:50 know what everyone always asked me oh
59:52 could you make it say my business here
59:55 and I was always like well no because
59:57 it's just that's kind of where we are
59:59 right now like there was this apology so
1:00:00 like one of the things mid-journey
1:00:02 people all wanted was text and then
1:00:06 ideogram comes along and bang there's
1:00:08 text so we're going to be in that place
1:00:11 where we know the stuff that when it
1:00:14 comes how big a deal it's going to be
1:00:17 and you'll be able to pounce on it thank
1:00:19 you AI videos I appreciate I like pizza
1:00:23 I love me some good pizza
1:00:25 but I live in [ __ ] Denver
1:00:27 ask me how the pizza is in Denver
1:00:35 three places I've found three
1:00:39 New York City you you trip and you find
1:00:41 three pizza places that are [ __ ]
1:00:43 world class
1:00:45 Denver Pizza Denver Pizza
1:00:48 wire hangers wire hangers
1:00:52 Google understands the subsidy model yes
1:00:55 FaceTime accidents are priceless
1:00:59 exactly
1:01:05 this channel is addictive sorry Firefly
1:01:08 I know I well like you've been in here
1:01:12 before we're like I'm like okay another
1:01:14 10 minutes and then like an hour later
1:01:16 I'm still like [ __ ] going strong my
1:01:18 eye looks blood running out of my tear
1:01:20 ducts
1:01:22 um hello wanted to say thank you for
1:01:24 your hard work I appreciate that Roger
1:01:26 Joel Ramirez that's very nice of you to
1:01:28 say
1:01:29 careful with that camera in the shower
1:01:31 exactly
1:01:34 it's rough enough clothes
1:01:39 how's all Joker doing
1:01:44 did you just get broken
1:01:47 he just got broken in the second set
1:01:50 Djokovic come on boy we'll turn it
1:01:53 around
1:01:54 uh
1:01:58 all right I tried call it out today it's
1:02:01 so incredibly sensitive in terms of what
1:02:03 it seems is inappropriate so so Tech
1:02:05 audit TV
1:02:07 um
1:02:08 these
1:02:11 all of these things have safety guard
1:02:14 rails on them
1:02:16 and the safety guard rails are as
1:02:19 annoying as [ __ ]
1:02:21 if you really want to you can install a
1:02:25 llama 2 or code llama The Meta
1:02:29 um large language models you can install
1:02:33 them on your own servers with no guard
1:02:35 rails
1:02:36 a buddy of mine
1:02:39 um installed llama 2 70 billion
1:02:42 parameter as a slack bot on his on his
1:02:46 company's slack Channel
1:02:48 with no guard rails and his his
1:02:50 experiment is
1:02:52 he's taking all of the conversations in
1:02:54 slack and embedding them and putting
1:02:57 them into a vector database and
1:02:58 fine-tuning fine-tuning the bot on all
1:03:01 the conversations he wants it basically
1:03:03 to become one of them
1:03:04 and he said it's absolute Insanity he
1:03:07 said it cusses all the time like
1:03:09 uncomfortably so like really bad really
1:03:12 awkward
1:03:13 um he says about every three days it
1:03:16 just changes personalities the other day
1:03:18 he said it um it started speaking
1:03:22 Spanish for three hours he said no
1:03:24 matter what they did they couldn't get
1:03:26 it to not speak Spanish
1:03:30 so if you you know don't ask a question
1:03:33 you don't want to know the answer to but
1:03:35 yeah like being especially
1:03:38 is is just so oversensitive to like what
1:03:41 you're allowed to say and what you're
1:03:42 not allowed to say and if you if you get
1:03:44 some army with it it gets even worse
1:03:46 it's like wow I guess we're a little too
1:03:48 upset to talk then aren't we we're just
1:03:51 gonna go ahead and start this
1:03:52 conversation over it's like Bing is like
1:03:55 the church lady of big is like the
1:03:58 church lady of gpts
1:04:00 [Laughter]
1:04:06 uh
1:04:08 um they're nice very nice all right
1:04:11 Tick Tock just jump into the bottom
1:04:13 again hang on let me see where I was
1:04:17 when the T-shirt will be on selling uh
1:04:19 because winter is coming um
1:04:23 hang on I think I have to sneeze
1:04:29 maybe not
1:04:32 I told you I knew it
1:04:35 I knew it was coming
1:04:37 let petite marked
1:04:40 that's for Utah gas
1:04:43 um
1:04:44 I'm gonna work on
1:04:47 yeah
1:04:49 I'm gonna work on the uh
1:04:54 the workshop and the T-shirt t-shirt
1:04:57 thing tomorrow
1:04:58 so hopefully I'll have something up
1:05:04 all right
1:05:05 sneeze about took me out
1:05:13 which one is the
1:05:15 best to take a list of locations from an
1:05:17 Excel sheet and plot it on a map oh
1:05:19 Brandon You're Gonna Love death
1:05:22 Brandon's Gonna Love deaths he's gonna
1:05:23 love it
1:05:26 whiskey we're gonna go to a site called
1:05:29 kaggle
1:05:31 you know you're like Kyle isn't that on
1:05:34 the dark web no kaggle's just an awkward
1:05:38 name for a data company a data website
1:05:42 where you can go get data sets
1:05:46 so we are gonna go
1:05:50 we're gonna go get a data set we're
1:05:52 gonna go get something with location
1:05:53 data
1:05:55 let's see
1:05:58 um locations
1:06:01 Chipotle locations perfect
1:06:07 so we're gonna download this
1:06:12 and it's just called
1:06:13 archive9.zip because I've done this
1:06:15 before but I haven't done it with
1:06:16 locations so this is this is uh I used
1:06:19 to do this thing in storyline called we
1:06:21 called it danger selling where we had we
1:06:24 we had an iPhone app that was Tethered
1:06:27 to a single Mac laptop
1:06:29 and
1:06:31 I would film someone
1:06:33 in a sales pitch meeting in the in the
1:06:36 app
1:06:37 it would then send the files from the
1:06:40 app to that Mac computer into into a
1:06:43 specific directory and it would fire off
1:06:45 a set of macros for Final Cut Pro and my
1:06:49 business partner Monique would sit there
1:06:51 watching the computer edit this video
1:06:54 together using Final Cut Pro macros and
1:06:57 it would take like 10 or 15 minutes so
1:06:59 I'd be just pitching my butt off and
1:07:03 um and then when it was ready she'd give
1:07:04 me like a subtle thumbs up and I'd say
1:07:06 and now your video's ready and I would
1:07:08 show them their their fully edited video
1:07:10 that was automated in the cloud and it
1:07:12 was all just vaporware and uh I got a
1:07:15 standing ovation once and the other the
1:07:18 other time Final Cut Pro crapped out
1:07:20 about two-thirds of the way through the
1:07:22 video so she showed me where it it
1:07:24 [ __ ] up and so I started playing it
1:07:26 and right before it got to the [ __ ] up I
1:07:28 said well that's you don't need to see
1:07:30 all of the thing I stopped it so that
1:07:32 was danger selling so this is uh this is
1:07:34 danger chat gpting
1:07:38 okay so we're gonna go oh this is a cool
1:07:40 project too so this this is super cool
1:07:44 um I use Chachi PT to write a whole
1:07:47 series of prompts for
1:07:49 like fantasy images and then went to
1:07:52 mid-journey and literally just copy and
1:07:54 pasted prompts one after the other and
1:07:55 created like a hundred images that are
1:07:59 all super [ __ ] cool
1:08:01 um but what am I doing I'm going to
1:08:04 Chacha pate
1:08:05 you all can make money
1:08:09 okay so we are at
1:08:13 Chachi BT we're gonna go to gpt4 and I'm
1:08:16 gonna go to what used to be called code
1:08:18 interpreter but thanks to this channel
1:08:20 they saw the errors of their ways and
1:08:23 they renamed it Advanced Data analysis
1:08:25 which is good that's what it is
1:08:29 and so what you can do now is you got
1:08:32 this little plus button so you can
1:08:33 upload a file
1:08:35 and so I'm gonna upload this
1:08:37 archive9.zip which is Chipotle locations
1:08:40 so I'm going to say
1:08:41 please
1:08:44 identify the data here
1:08:50 and
1:08:53 let oh and by the way yes I do say
1:08:55 please and thank you to this thing not
1:08:57 because I think it's makes me a good
1:09:00 person or or it makes better answers but
1:09:02 because
1:09:03 it's so positive in how it responds to
1:09:07 me it's just kind of instinctive like
1:09:09 I'm talking to another person it's a
1:09:11 weird thing I think these tools are
1:09:13 actually going to make us better people
1:09:15 oddly enough ironically enough
1:09:18 um please identify the data here and let
1:09:19 me know
1:09:22 um if there is Geo
1:09:27 data
1:09:28 and let's see if it knows what Geo data
1:09:31 even means it should
1:09:34 so this is this is
1:09:37 um you have to pay for GPT plus to get
1:09:40 access to this Brandon and you have to
1:09:41 enable it in your settings under beta
1:09:44 features
1:09:51 so so what it did there so the first
1:09:54 thing it did it said all right
1:09:56 um first I'll unzip the file so so what
1:09:59 advanced
1:10:00 data analysis does
1:10:03 when you launch it when you launch a
1:10:05 chat it's actually spinning up a virtual
1:10:08 machine in open AI servers somewhere
1:10:12 and it's got access to a python library
1:10:14 and so it it's actually writing and
1:10:17 executing its own python code to analyze
1:10:19 this data for you so the first thing it
1:10:21 does is it is it Imports two libraries
1:10:24 to to unzip the file and then it unzips
1:10:27 it figures out what's in it so then it
1:10:30 said the presence of Json with the name
1:10:31 uh suggesting U.S states hence a
1:10:34 potential geodata so it knows what
1:10:36 geodata is I'll take a look at the first
1:10:39 few rows of the Chipotle store CSV file
1:10:42 and it's got State location address
1:10:45 latitude longitude the presence of lat
1:10:47 launch columns indicates the CSV file
1:10:50 contains geodata next I'll inspect the
1:10:54 structure of the U.S states file to see
1:10:56 how it goes
1:10:58 um
1:11:00 type features okay from the sample data
1:11:03 the geometry section contains
1:11:05 coordinates
1:11:11 wait from the sample data provided the
1:11:13 geometry section contains coordinates
1:11:16 that describe the shape of Alabama
1:11:19 so it's determined in this file it's got
1:11:22 the shape of states
1:11:24 indicating that this is geodata
1:11:27 representing the boundaries of U.S
1:11:28 states
1:11:29 holy [ __ ]
1:11:33 that's amazing
1:11:35 would you like any further analysis or
1:11:38 visualizations based on the data
1:11:42 yes please can you create a map of the
1:11:46 United States with the Chipotle
1:11:48 locations
1:11:50 represented as flashing Stars
1:11:55 on a dark blue background
1:12:01 let's see
1:12:03 certainly here's the plan use the base
1:12:05 map of the United States with a dark
1:12:07 blue theme
1:12:09 place the Chipotle stores locations on
1:12:12 the map represent the store locations as
1:12:14 flashing Stars
1:12:16 let's create the visualization so it's
1:12:18 working now sometimes this thing shifts
1:12:20 to bed what's amazing when it shits the
1:12:22 bed is here's a map
1:12:25 you can zoom in and out so it didn't
1:12:28 make it let's see
1:12:30 I don't see the map can you try again so
1:12:33 that I can see the map in the browser
1:12:40 I apologize for the inconvenience
1:12:42 instead I'll wait
1:12:45 displaying interactive maps directly in
1:12:47 this environment could sometimes be
1:12:48 problematic instead I'll generate an
1:12:50 HTML file of the map and you can
1:12:53 download and open with your web browser
1:12:55 so let's see if that worked download and
1:12:57 open the map
1:13:02 open file
1:13:07 there you go
1:13:09 now are they Stars no wait are they
1:13:15 they didn't quite get it right but
1:13:16 there's all your Chipotle locations map
1:13:18 you know put on a map
1:13:22 code written by
1:13:24 let's let's go to the let's go to the
1:13:26 videotape let's go to the um
1:13:30 let's go to the file
1:13:37 thank you
1:13:45 so here's all the code that it wrote for
1:13:48 that map
1:13:54 so this is taking all the location data
1:13:58 from the Json that knows the shape of
1:14:00 the states the Latin longitude of the
1:14:03 Chipotle stores from two different files
1:14:05 and it wrote all of this code to make
1:14:09 that map of the United States
1:14:15 now I don't know [ __ ] data analysis
1:14:18 but for me this is one of these
1:14:23 this is like a Kevin McAllister moment
1:14:28 look at
1:14:29 the amount of code and what did this
1:14:31 take 30 seconds
1:14:33 to generate the code
1:14:38 it's [ __ ] insane
1:14:41 it's [ __ ] insane it's nuts
1:14:47 it's nuts it's nuts I tell you it's nuts
1:14:49 I tell you it's this cuckoo it's cuckoo
1:14:53 a cuckoo
1:14:55 if you're like sitting on the sidelines
1:14:56 going yeah whatever man
1:15:00 this is some major [ __ ] [ __ ]
1:15:06 some major [ __ ] pay attention dig in get
1:15:10 curious
1:15:12 doesn't just do code it does pictures it
1:15:14 does audio it does animations it does
1:15:16 film it does words it does writing it
1:15:19 does research
1:15:21 does it do it all perfectly [ __ ] no it's
1:15:23 horrible these are janky pieces of [ __ ]
1:15:26 TM tmai learning lab
1:15:29 2023
1:15:31 state of generative AI janky pieces of
1:15:34 profoundly remarkable [ __ ]
1:15:38 [Laughter]
1:15:43 oh all right the hair's starting to go
1:15:46 this is good this is good we got see
1:15:48 what the problem is I wear three two
1:15:50 pays I I braid my toupees together
1:15:53 and sometimes the braid falls apart and
1:15:55 that's kind of what we got got going on
1:15:58 tonight
1:16:00 all right so I hope that helped there
1:16:02 Brandon G
1:16:05 [Music]
1:16:07 like take a list of locations and plot
1:16:09 it on a shareable map you mean like we
1:16:11 just did Brandon
1:16:16 um
1:16:16 [Music]
1:16:17 chat gbt will do that for you
1:16:21 voice to text I use vid to text oh okay
1:16:23 good Vicky I don't know that one
1:16:25 um cool listen to Vicky
1:16:29 um how is whisper compared to something
1:16:31 like rev I don't know rev sorry
1:16:34 um rev might be really good rev might be
1:16:36 using whisper on the on the back end
1:16:38 because Whispers just you know the the
1:16:41 why I showed it to you in the playground
1:16:43 is they're giving it they're giving you
1:16:44 access to it as a developer tool so it's
1:16:47 possible that rev is using
1:16:50 um whisper but but I don't know they
1:16:52 might have their own thing as well I
1:16:53 don't I don't I've heard of them but I
1:16:55 just haven't used them do you think you
1:16:56 can ask GPT acting as a communication
1:16:59 way
1:17:00 with someone you share an ID
1:17:04 um
1:17:06 not right now but but
1:17:09 um
1:17:10 it sounds like
1:17:12 chat GPT Enterprise you're going they're
1:17:16 going to have a prompt Library so you'll
1:17:17 be able to share
1:17:19 your prompts amongst people
1:17:22 and I would imagine
1:17:25 that once they put permissions in place
1:17:27 for different levels within an
1:17:29 organization that a manager might be
1:17:32 able to see what you know one of their
1:17:34 reports is doing I don't know
1:17:37 right now right now the chat CPT stuff's
1:17:39 very singular there may be if you go to
1:17:43 um if you go to futurepedia.io
1:17:46 there may be tools that are doing
1:17:48 collaborative workspaces where you can
1:17:51 be chat gpting in groups that feels like
1:17:54 an application layer thing that sits you
1:17:56 know on top of gpt4 so there might be
1:18:00 stuff that people have developed in the
1:18:02 past year
1:18:03 um that's worth checking out that that
1:18:05 may do that took us
1:18:07 basically I have locations from a list
1:18:09 hopefully we got that am I late to the
1:18:11 party yeah Winston you missed it was so
1:18:13 much fun like everyone was dancing I
1:18:16 twerked for like 15 minutes there was
1:18:18 Applause like I could hear it oh it was
1:18:20 so good
1:18:21 so anyway but now we're just talking
1:18:23 about AI [ __ ] it's really boring now
1:18:25 sorry my air conditioner is not working
1:18:27 it's getting a little moist in here
1:18:29 hold please
1:18:38 [Music]
1:18:42 come on
1:18:49 [Music]
1:18:52 all right
1:18:55 I wonder if it's cool out is it cool
1:18:57 here in Denver's like 100 today
1:19:00 yeah a little bit that might be a little
1:19:02 better we'll see
1:19:04 holy [ __ ] joke about Djokovic lost the
1:19:07 first two sets six four six four oh
1:19:12 see if we can come back from two down
1:19:16 this is round three
1:19:21 started out the third set with an ace
1:19:23 that's promising
1:19:26 all right
1:19:28 is it free is what free what was I
1:19:31 showing
1:19:32 it might be
1:19:34 I don't remember what I was showing
1:19:38 um the image thing
1:19:40 ideogram is free for now it won't be for
1:19:43 long I think
1:19:45 just the same as you can do with Alexa
1:19:46 like Alexa
1:19:48 call irregular Alexa uh can you make
1:19:51 money with chat EBT I don't know can you
1:20:05 make the money you can make the money
1:20:08 with the GPT
1:20:10 and then you and then you make the money
1:20:12 and then you take the money and then you
1:20:15 put in the bank
1:20:17 but not the bank you thanks not good uh
1:20:21 but you think you could make the money
1:20:22 with the church just just watch my
1:20:25 course it's it's five easy courses and
1:20:27 1995
1:20:28 okay
1:20:33 all right fantastic bro I'm telling what
1:20:36 is one
1:20:41 I've already said I'm ahead of my time
1:20:43 the problem is it's only about two
1:20:45 minutes yes I'm right there with you
1:20:48 I started an automated video company 11
1:20:51 years ago and they're just now starting
1:20:53 to figure out hey we need some of this
1:20:54 [ __ ] it's like
1:20:58 I could I could have saved myself some
1:21:00 time
1:21:02 all right let's see remember when Yahoo
1:21:05 and Google were sharing the search
1:21:06 engine Market yeah I remember Alta Vista
1:21:10 remember remember prodigy
1:21:13 AOL CompuServe
1:21:16 it used to be AOL Prodigy and CompuServe
1:21:18 were the big three then it was like
1:21:20 Netscape something and something
1:21:23 the browsers
1:21:25 and then it was like
1:21:27 Alta Vista and Yahoo
1:21:32 and then something else and then Google
1:21:34 kind of just snuck in there and took
1:21:36 over yeah it's the the sort of big three
1:21:39 tech companies that are the the big
1:21:41 three of the day that kind of cycle of
1:21:44 every five years or so one of them will
1:21:46 drop out or like a whole category will
1:21:48 drop out like you know AOL CompuServe
1:21:51 and Prodigy those are the big three and
1:21:53 then the the web browsers just went
1:21:56 and just [ __ ] ate them up they just
1:21:59 bang we're gone
1:22:01 that was pretty cool
1:22:04 I'm excited I know lunar stick it's so
1:22:06 cool it's it's so cool that you started
1:22:08 that
1:22:09 I haven't planned out daily and
1:22:10 everything very very cool that's awesome
1:22:14 can I get a job as a prompt engineer
1:22:16 probably
1:22:18 um
1:22:19 just be AI literate and talk about it
1:22:22 it's I I wouldn't necessarily here's the
1:22:25 thing so I was talking to one of my big
1:22:28 Pharma clients when we went to New York
1:22:30 um
1:22:31 two weeks ago
1:22:33 um my big Pharma client
1:22:37 three months ago
1:22:38 said I need to hire a prompt engineer
1:22:42 and then the more she learned about this
1:22:43 AI stuff the more she realized she
1:22:46 doesn't need a prompt engineer she needs
1:22:48 like an AI operations person she needs
1:22:50 someone kind of like me who's just like
1:22:53 looking at all the [ __ ] figuring out
1:22:55 what's possible sifting through that
1:22:57 maybe can do some prompting maybe can do
1:23:00 a little you know sort of whatever you
1:23:02 know system architecture kind of stuff
1:23:04 product stuff
1:23:06 um knows business
1:23:09 so someone who can be sort of a
1:23:12 translator Between Worlds right so you
1:23:14 got the business world you got the
1:23:16 content world
1:23:17 you've got the AI World you've got the
1:23:19 technology World you've got the
1:23:21 marketing world you got the sales world
1:23:23 you got all these different worlds
1:23:26 so she wants to find someone that can
1:23:27 sit at the middle of that and and be a
1:23:30 linchpin for her she initially thought
1:23:31 that was a prompt engineer and then she
1:23:32 realized it was way bigger than that so
1:23:35 while you probably can get a job as a
1:23:37 prompt engineer just start getting
1:23:39 curious start exploring start building
1:23:41 [ __ ] start building [ __ ] and zapier
1:23:43 because the difference between just
1:23:45 being good at prompting and being able
1:23:47 to create an automation that
1:23:49 automatically creates a machine that
1:23:51 does prompting big big difference and
1:23:54 might be a big difference in value
1:23:56 and then if you can understand business
1:23:58 that's another bonus point on top of
1:24:00 that and then if you understand writing
1:24:01 on top of prompting that's good because
1:24:04 not only can you prompt you can edit the
1:24:06 [ __ ] that chat EBT gives you when you
1:24:08 realize oh this is dog [ __ ] this is good
1:24:12 so start thinking horizontally in terms
1:24:17 of your skill set
1:24:18 because these tools allow you to do lots
1:24:21 of different things well
1:24:23 and you can be more like an
1:24:25 a conductor of an orchestra
1:24:27 rather than a virtuoso prompt Engineers
1:24:30 very much a virtuoso rule role
1:24:34 um I think they're going to be a dime a
1:24:35 dozen in two years and quite frankly I
1:24:39 think most of these tools
1:24:42 the requirement
1:24:44 of learning how to prompt is going to go
1:24:46 away
1:24:47 on the on the engineering side you'll
1:24:50 still need to understand prompting
1:24:52 but the users of these tools won't
1:24:55 so it depends where you want to be and
1:24:58 what you want to do but I would go more
1:24:59 horizontal than than thinking of it that
1:25:02 that tactically
1:25:05 it's like browser Wars how can we take
1:25:07 advantage well so multi-dimensional man
1:25:11 what I just said right how we take
1:25:13 advantage right now all of us is
1:25:16 step one is just dick around with these
1:25:19 do the parlor tricks have it write poems
1:25:21 have it write songs have it write a meal
1:25:24 plan for the family a project plan for
1:25:27 for the community group
1:25:29 um a marketing plan for work uh a list
1:25:32 of names of the new product that you
1:25:34 give to the CEO just to kiss his ass a
1:25:36 little bit
1:25:38 um that kind of [ __ ]
1:25:41 and then
1:25:42 and then step two is
1:25:45 think about what are you good at and how
1:25:48 can this stuff augment what you do
1:25:53 and then step three is
1:25:55 really start to expand beyond that and
1:25:58 start to say okay how can I bring those
1:26:00 those worlds together of all the stuff
1:26:02 that it is possible
1:26:04 it can do all of the stuff I know how to
1:26:07 do and make it do and then all of the
1:26:10 stuff that I might have never even
1:26:12 considered doing but is now possible so
1:26:14 it's you know it's whatever it's nested
1:26:17 Russian dolls or it's nested it's
1:26:19 appealing The Onion
1:26:22 um just start with curiosity and
1:26:24 exploration and then expand out
1:26:26 and talk about it talk about it talk
1:26:29 about it this is the most important
1:26:31 thing why this channel exists get
1:26:34 yourself out there but Kyle people think
1:26:36 I'm weird [ __ ] them
1:26:40 you know what that is you're weird why
1:26:43 are you always talking about this AI
1:26:44 stuff you dumb dumb that's fear
1:26:48 that's fear of change you know they
1:26:51 don't they don't always tell the truth
1:26:52 you know they're not connected to the
1:26:54 internet you know they're only as good
1:26:55 up till September 2021. those people
1:26:58 [ __ ] them
1:27:02 because you know what those people are
1:27:04 going to do two years from now
1:27:06 hey uh could you teach me how to use
1:27:09 that AI stuff I think I might want to
1:27:11 learn it now oh really
1:27:13 sorry I'm two years down the road gotta
1:27:16 go
1:27:17 I got a new job bye
1:27:24 you can be nice to them you don't have
1:27:25 to be a jerk to them like I just was
1:27:27 they don't exist anyway so
1:27:29 [Laughter]
1:27:32 uh
1:27:36 by the way if you're new here welcome my
1:27:39 name is Kyle Shannon this is the AI
1:27:40 learning lab I'm more than 99 messages
1:27:43 behind which means if you've asked a
1:27:45 question I haven't seen it yet if you're
1:27:48 sitting there all bitter and turned off
1:27:50 and like hey Anthony answered my
1:27:51 question yet when is he gonna answer my
1:27:53 question I'll get to it or not
1:27:56 but that's not the point I do try to
1:27:58 answer all the questions you can ask the
1:28:00 Irregulars the Irregulars are the people
1:28:02 that are here regularly but they're
1:28:04 irregular
1:28:05 um
1:28:06 if you haven't played with chat GPT just
1:28:09 like I said the multi-dimensional man
1:28:10 the place to start is just to start
1:28:12 don't you don't have to know computers
1:28:15 this is like as simple as a Google
1:28:17 search
1:28:18 to get started go there
1:28:21 chat.openai.com that's the official chat
1:28:23 GPT website it's free you can pay 20
1:28:26 bucks if you want the fancy [ __ ] but the
1:28:28 free version is still [ __ ] remarkable
1:28:30 go there
1:28:33 and
1:28:35 try to stump it
1:28:37 ask it to write you
1:28:40 a poem in iambic pentameter
1:28:44 um
1:28:45 about
1:28:47 the myth of Bonnie and Clyde
1:28:51 and it will
1:28:53 and then ask it to write you a country
1:28:55 song and then ask get to write you a
1:28:56 recipe book and then ask it to write you
1:28:58 a table comparing the Mesopotamian War
1:29:01 with Cheetos and it will
1:29:06 and then go to that thing prompts.chat
1:29:11 and that will teach you what these
1:29:13 things are and how to talk to them
1:29:16 and these three are the three that that
1:29:18 are kind of the the three that I bounce
1:29:20 between when I'm trying to like when one
1:29:22 frustrates me I'll just bounce it
1:29:23 between the three of those
1:29:26 so go do that so welcome welcome
1:29:29 we like to have fun here I I I like to
1:29:32 to be a dum-dum
1:29:34 let's see Djokovic just broke the other
1:29:37 dude de jir
1:29:42 he's down two sets man
1:29:45 can he pull it back he's Djokovic he
1:29:48 probably can
1:29:50 the rich voice I was on the fence but
1:29:52 I'm going all in on AI I'm going in on
1:29:55 AI so that's huge like so here's the
1:29:59 deal I know I kind of make fun of like
1:30:01 people that poo poo it and fear this
1:30:03 shit's scary it's
1:30:07 and if your name is the rich voice I
1:30:09 assume you might be a voice actor you
1:30:12 might have lovely tones don't said tones
1:30:15 Lucid tones tones that people like and
1:30:18 appreciate and they want to hear your
1:30:20 voice they want you to read a book
1:30:23 um that industry
1:30:25 is in for a shake-up as our all
1:30:29 Industries so I don't know if you're in
1:30:31 that in that Biz
1:30:34 um
1:30:37 the only thing
1:30:39 that I know how to do
1:30:42 is get educated and get curious and and
1:30:45 start to figure out how do these tools
1:30:50 how well do these tools do the [ __ ] that
1:30:52 I care about
1:30:55 how well do they do them how well do
1:30:57 they not do them what are their
1:30:59 strengths what are their weaknesses what
1:31:01 are the things that scare the ever
1:31:02 living [ __ ] out of me
1:31:06 what are the things where I'm like oh
1:31:07 okay I've probably got three years or
1:31:09 maybe two well maybe one but if I have
1:31:12 one what am I going to do in that year
1:31:13 until that thing comes
1:31:16 like in the voice synthesis World these
1:31:19 things are still kind of robotic and
1:31:21 monotone
1:31:22 a year from now they're not gonna be
1:31:25 three years from now
1:31:28 it's going to be indistinguishable
1:31:32 so what do you do I don't know
1:31:35 you know Grimes
1:31:37 um elon's X
1:31:40 when the when the Drake and the weekend
1:31:42 song when when the the someone made that
1:31:45 song with the Drake Drake in the weekend
1:31:46 and it got all those Spotify views and
1:31:48 then
1:31:49 people freaked out or Sony Music freaked
1:31:52 out or someone freaked out and Spotify
1:31:54 pulled the song
1:31:56 and then like a week later Grimes said
1:31:58 synthesize me all you want just give me
1:32:01 50 percent
1:32:03 that's an approach right and there's
1:32:05 going to be agencies that specialize in
1:32:09 um creating the best models to
1:32:12 synthesize Talent so that other people
1:32:14 can use those
1:32:16 um and then people could potentially get
1:32:17 compensated for that that's a lot of
1:32:19 what sag's trying to work on right now
1:32:20 so
1:32:22 um
1:32:23 so yes so yay don't sit on the fence if
1:32:26 you can at all avoid it um
1:32:29 even though it's scarier than [ __ ]
1:32:30 sometimes and it's it's
1:32:33 some of this stuff's existential like
1:32:35 some of this stuff is like
1:32:37 you know my value as a human is
1:32:41 the the blank that I do right The Voice
1:32:44 work that I do the drawing that I do the
1:32:47 music that I do the talking that I do
1:32:48 the writing that I do the thinking and
1:32:52 articulating that I do
1:32:54 and then you use a tool that can do some
1:32:57 version of that and sometimes it can do
1:32:59 the version of that better than you can
1:33:02 and then you go [ __ ] what now
1:33:06 I don't know keep learning keep
1:33:09 expanding keep keep
1:33:12 figuring out where the edges of this
1:33:15 thing are because it's expanding really
1:33:17 rapidly
1:33:19 so [ __ ] Kudos Rich voice
1:33:22 Nancy I wish I could get your
1:33:24 notifications
1:33:26 um Nancy join the Discord
1:33:28 and
1:33:30 let me know that you joined and I'll
1:33:32 give you the role of an irregular
1:33:35 and I made a private channel in the
1:33:36 Discord what I'm thinking is I might I
1:33:38 might do announcements of when I'm going
1:33:40 live in the Discord and then you can get
1:33:44 notifications there because tick tock's
1:33:46 an [ __ ] about
1:33:47 I don't know they like I'm in some weird
1:33:50 sort of Shadow band state right now
1:33:51 where these lives used to get you know
1:33:54 two or three hundred people sometimes
1:33:55 then it was like 150 then it dropped to
1:33:58 like 100 now it's down to like between
1:34:00 50 and 80.
1:34:02 um so I'm in some sort of Shadow band
1:34:03 thing so I I don't think it notifies a
1:34:05 lot of people when I go live but that's
1:34:07 all right it'll it'll turn around again
1:34:08 they'll change some algorithm and go we
1:34:12 really want to prioritize old fact guys
1:34:19 right right now apparently it's still
1:34:21 the twerking the twerking hotties still
1:34:25 getting apparently they're still getting
1:34:27 the views they're still getting the
1:34:29 engagement that Tick Tock likes
1:34:32 laughs
1:34:35 oh good lord
1:34:37 note to self learn to twerk get the
1:34:41 sequins
1:34:43 bedazzle your cutoffs
1:34:45 [Laughter]
1:34:50 that's why everyone in Denver is skinny
1:34:52 not me
1:34:54 um
1:34:55 all right listen I am going to I'm more
1:34:57 than 99 messages behind it's we've been
1:34:59 doing this long enough I'm going to jump
1:35:00 all the way to the bottom and just go up
1:35:02 a little bit
1:35:04 because I the people that have just come
1:35:07 in recently are going to be like he
1:35:09 never gets to the questions please get
1:35:12 to the questions so I just skipped a
1:35:13 whole bunch of questions so if I skipped
1:35:15 your question my apologies all right you
1:35:18 do great job explaining and showing us
1:35:20 100 thank you micro I appreciate the
1:35:23 kind words I really do it means a lot
1:35:26 um by the way you know also tell me like
1:35:29 someone someone the other day said you
1:35:31 know hey could we could we stop with the
1:35:33 chit chat and get back to more Show and
1:35:34 Tell kind of stuff I'm happy to do
1:35:36 whatever I mean I really
1:35:38 I don't know if you noticed but the
1:35:40 whole
1:35:41 Jam of this channel is like I answer
1:35:45 I read as many of the what you say is
1:35:48 possible and talk to that
1:35:50 um so if you want to see [ __ ] let me
1:35:51 know I'll do that if you want me to talk
1:35:53 [ __ ] I'll talk [ __ ] and then yeah you
1:35:56 know we'll do funny stuff with funny
1:35:58 stuff you said Kyle okay good thank you
1:36:01 I I don't know what the funny stuff
1:36:03 you're gonna do with the stuff I said
1:36:05 but you know blackmail is funny that
1:36:07 could be funny
1:36:09 um all right good evening everybody I'm
1:36:11 not a cat very nice love his humor thank
1:36:15 you uh what about Wednesday what about
1:36:18 Wednesday
1:36:23 I do not know
1:36:26 Tick Tock won't let me post gpt4's
1:36:29 excellent poem in iambic pentameter
1:36:31 about Bonnie and Clyde
1:36:33 post it in the uh go to go to Sam are
1:36:36 you on the Discord for for the AI Salon
1:36:38 if you're not go here so the salon.ai
1:36:42 this is a group that I started back in
1:36:44 December
1:36:46 um that's a group of AI adventurers we
1:36:49 call ourselves so it's a community and
1:36:50 we have a Discord so the link to the
1:36:52 Meetup we meet every other week and then
1:36:54 we've got a Discord go post your poem in
1:36:57 the there's a show and tell Channel and
1:36:58 there's General discussion and then um
1:37:01 put it in there and I'll I'll add you
1:37:03 I'll add I'll give you the role of an
1:37:04 irregular and we've got a private uh
1:37:06 channel in that Discord for this group
1:37:09 as of today I just added it
1:37:13 um that's awesome
1:37:14 yep I was stuck with a bug in es 32
1:37:17 library with terrible documentation chat
1:37:19 GPT saved me a few hours
1:37:21 yes so these tools are [ __ ]
1:37:25 remarkable at reading other [ __ ]
1:37:28 it can read a novel you wrote it can
1:37:30 read your shitty screenplay and help you
1:37:32 fix it it can read
1:37:34 um
1:37:35 code and document it it can read
1:37:37 documentation and code it I gave so
1:37:41 my companies the way we assemble videos
1:37:44 at storyvine is an a kind of a
1:37:46 proprietary XML thing
1:37:48 and how how we think about how we
1:37:51 structure videos is different than the
1:37:53 XML language we have so it's a really
1:37:55 complicated translation where art what
1:37:59 we call story Engineers have to think
1:38:01 visually but then they have to code in a
1:38:04 very in a timeline kind of way it's this
1:38:06 weird thing
1:38:07 and they have to code it and so
1:38:10 I asked chat GPT to take a look at our
1:38:12 XML analyze it which it did [ __ ]
1:38:16 flawlessly
1:38:17 um and then I said here's how we think
1:38:20 about storytelling how would you
1:38:22 re-architect the XML to match our story
1:38:25 structure and it [ __ ] did it
1:38:28 perfectly it was [ __ ] that I've been
1:38:30 thinking about for five years like man
1:38:32 if I ever do this here's kind of how I
1:38:34 would approach the XML translation to to
1:38:36 a new version of XML and it [ __ ] just
1:38:39 did it like 100 it took an existing
1:38:44 chunk of XML with live data in it
1:38:47 rewrote it with new tags new syntax new
1:38:50 structure and maintained all of the
1:38:53 variables all of the links to the assets
1:38:56 it was it was [ __ ] insane
1:38:59 this stuff this stuff when what's it
1:39:02 what's interesting about this stuff
1:39:03 there's kind of a new version of
1:39:05 writer's block that I get with with
1:39:07 generative AI
1:39:09 which is writer's block is completely
1:39:11 gone right because any sort of little
1:39:14 hint of an idea that you have you can
1:39:16 stick it in one of these tools and like
1:39:17 bring it Springs to life that's gone
1:39:21 but what's what's what's replacing it
1:39:23 the sort of there's a new paralysis and
1:39:26 the new paralysis is you look at that
1:39:29 prompt box
1:39:31 and you're like well whoa
1:39:33 what what do I what what do I do with it
1:39:36 what do I put in there you're so cute
1:39:39 your little box so I could I could put
1:39:41 anything in here I don't know what to do
1:39:43 and I just sit there and I'm like well I
1:39:46 could make a poem again for about the
1:39:48 40th time that's [ __ ] no okay that's
1:39:51 dumb well I could have it write an email
1:39:53 I don't have any emails to write and I
1:39:56 can write emails I'm fine with that uh
1:39:59 and so
1:40:00 where they where how you break through
1:40:02 that is don't think about what do I do
1:40:05 in chat GPT
1:40:07 flip the script and say what's [ __ ] that
1:40:10 I have in my life
1:40:11 that is a headache for me or is it is
1:40:14 hard or like I do repetitively and it's
1:40:18 Soul crushing to me
1:40:20 start with that [ __ ] and then go how can
1:40:22 this stuff
1:40:23 fix that
1:40:25 and it's you'll solve it like that it's
1:40:28 amazing
1:40:29 it's amazing when you flip flip that
1:40:31 that script even the first verse is too
1:40:34 long oh
1:40:35 oh no
1:40:39 um
1:40:42 always learning always be learning
1:40:48 abl
1:40:49 remember Glen Gary Glenn Ross ABC
1:40:53 always be closing
1:40:57 Djokovic up five love in the third set
1:41:00 after going down two sets
1:41:02 he's pith
1:41:04 he's pissed he's spanking that other guy
1:41:07 spanking him
1:41:11 um I've been trying to set up VPC
1:41:12 networking for Google's vertex AI
1:41:15 matching engine Vector database very
1:41:17 cool although you're a level I don't I
1:41:20 am not enough on that side of the I'm
1:41:23 not that deep in the AI [ __ ]
1:41:26 but sounds cool
1:41:28 uh I don't I don't know it I don't know
1:41:31 it I ain't been in it yet
1:41:34 I when when uh for like the last five
1:41:37 years every time Google would put out
1:41:39 something about tensorflow they're like
1:41:41 there's a new version of tensorflow and
1:41:43 you can do machine learning I'm like yay
1:41:44 we can do machine learning if you went
1:41:47 to Stanford and have a math degree [ __ ]
1:41:52 and then when was it it was I don't know
1:41:55 early in 2022
1:41:57 when I I was doing generative image
1:42:00 stuff and I discovered the GPT
1:42:02 playground and I'm like oh okay I can
1:42:04 get my head around this because this is
1:42:06 like English language as a programming
1:42:09 language this I can this I can have some
1:42:10 fun with
1:42:12 um yeah GPT saved my absolute ass
1:42:15 because networking has never been my
1:42:16 strong suit oh that's awesome very very
1:42:18 cool thank you for the car reminder the
1:42:21 dryer just stopped
1:42:24 took us you're the best it's got to be
1:42:27 late where you are isn't it it's got to
1:42:30 be like [ __ ] one in the morning or
1:42:31 later
1:42:33 no
1:42:35 AI is going to change the world in ways
1:42:38 we can't even imagine probably a potato
1:42:40 okay first of all probably a potato 10
1:42:43 out of 10 on the name solid name now
1:42:47 zero out of ten on the on the Avatar
1:42:49 because you don't you didn't put a
1:42:51 picture in there probably a potato go to
1:42:53 Mid Journey go to go to anywhere and
1:42:56 make a potato and put it in your picture
1:42:58 there's a massive lost opportunity here
1:43:01 probably a potato but anyway
1:43:04 anyway tater listen Tater yeah AI is
1:43:08 going to change the world in ways we
1:43:10 can't even imagine I this is why I come
1:43:14 here every night
1:43:15 because I do not want to lose touch with
1:43:26 understanding everything I possibly can
1:43:30 about what's coming and how it's
1:43:32 impacting people and how they're
1:43:34 figuring it out
1:43:36 because honest to God every time someone
1:43:38 in here says oh I figured out something
1:43:40 cool and they give me a little hint I'll
1:43:42 go off these lives and I go try that
1:43:44 [ __ ] but I'm like oh that's a different
1:43:46 way to think about it
1:43:49 this is we this is
1:43:52 having been through this with with
1:43:55 um with the World Wide Web
1:43:57 you know starting early and like riding
1:43:59 that [ __ ] rocket ship up
1:44:01 and then back down and then across
1:44:05 um
1:44:09 the the only thing that I know for sure
1:44:11 and I know this for sure like there's no
1:44:14 any any question about it the only thing
1:44:17 I know for sure is
1:44:18 this is a singular moment in history
1:44:22 we are
1:44:24 in rarified air as humans
1:44:28 to be lucky enough to be alive right now
1:44:31 to be lucky enough to understand what
1:44:34 life was like before AI that generative
1:44:36 AI this stuff November 30th 2022 chat
1:44:39 gpt's launched that for me is the is is
1:44:43 the demarcation
1:44:44 of there was the time before that and
1:44:47 there was the time after that the fact
1:44:49 that we get to be at that cusp
1:44:53 is [ __ ] amazing
1:44:58 and so
1:44:59 all we can do all we can do is surmise
1:45:04 but the more you learn these tools and
1:45:06 the more you learn what they do and the
1:45:08 more you hear someone was like hey I
1:45:10 didn't know anything about networking
1:45:11 but I used chat GPT to help me figure
1:45:13 all this networking [ __ ] out and I
1:45:15 didn't look like a schlub to my boss
1:45:18 [ __ ] awesome and then you go oh [ __ ]
1:45:20 that's amazing okay wait that's kind of
1:45:24 like what I'm doing over here but it's
1:45:25 not networking but it's this but I don't
1:45:27 understand it maybe maybe I could use
1:45:29 Chachi BT huh yeah oh [ __ ] it worked for
1:45:33 me too hey boss here's a cool thing oh
1:45:36 my God
1:45:37 Jolly Rancher how did you do that well
1:45:40 I'm just special and I Had a Little Help
1:45:42 From A Friend
1:45:46 it's a remarkable time
1:45:49 and we
1:45:53 this is the only time
1:45:55 when it's this Raw
1:46:04 we can now say
1:46:06 remember last month
1:46:09 when you made images you couldn't put
1:46:11 words in them that was weird
1:46:15 remember six months ago when hands
1:46:17 looked like this no matter what you did
1:46:20 wonder how they fixed that
1:46:22 like this is
1:46:25 like this time is [ __ ] Magic
1:46:31 because these tools are magic right
1:46:33 Arthur C Clarke any sufficiently
1:46:35 advanced technology is indistinguishable
1:46:38 from Magic
1:46:42 if you look at the math behind this it's
1:46:44 all science these are actually very
1:46:46 simple probability engines well they're
1:46:48 not very simple they're trading on
1:46:51 massive data sets but conceptual they're
1:46:52 conceptually very simple
1:46:54 you type in prompts it turns it into a
1:46:57 mathematical probability and matches
1:46:59 that probability to some word in
1:47:01 thousand-dimensional Vector space not
1:47:04 even a word a point in Vector space
1:47:06 which is then tied to a word and it
1:47:09 brings that word back to your document
1:47:12 it's just technology it's just
1:47:14 remarkable [ __ ] mathematician geek
1:47:18 [ __ ] insane technology
1:47:21 that occurs as magic
1:47:24 because it's [ __ ] magic I don't care
1:47:27 and you know people are like well when's
1:47:29 it going to be sentient and when's it
1:47:30 going to be this and when's it going to
1:47:32 be AGI you know what it doesn't [ __ ]
1:47:34 matter that's all semantics
1:47:38 it right now is the most profound Human
1:47:42 Assistance tool in the history of
1:47:45 humanity it just is
1:47:48 you don't have to like it you don't have
1:47:50 to
1:47:55 think it's perfect or brilliant or even
1:47:57 good
1:47:59 but it's [ __ ] remarkable
1:48:02 and it's going to change how we think
1:48:04 it's going to change how we interact
1:48:05 with one another it's going to
1:48:07 fundamentally change jobs
1:48:10 and even what work is
1:48:14 and you're like but Kyle it's just a
1:48:16 stupid [ __ ] poem writer I know
1:48:19 you're right
1:48:21 let's talk in five years
1:48:29 you know
1:48:30 you know you know what I'm saying you
1:48:32 know what I'm saying man all right what
1:48:34 time is it it's going on 10 o'clock here
1:48:36 in Denver Colorado You Know Where Your
1:48:38 Children Are Mine are right upstairs
1:48:39 right upstairs
1:48:42 yeah the hair's gotten a little a little
1:48:45 kind of
1:48:46 it's gone kind of 80s hasn't it
1:48:49 it's like it's just my hair doesn't know
1:48:52 a decade it's in on an hourly basis it's
1:48:55 like [ __ ]
1:48:57 [ __ ] we're gonna go 70s carpet salesman
1:49:00 80s
1:49:02 corporate cool dude
1:49:08 uh uh I know my Discord handle now cool
1:49:10 so Dr J the other thing you can do is
1:49:12 just ping me I'm Kyle Shannon in the
1:49:14 Discord just ping me in the Discord and
1:49:16 I'll give you the uh I'll give you the
1:49:18 the roll
1:49:20 um you get to this you get to the salon
1:49:22 Discord go to the salon.ai
1:49:25 and it's a link tree and then there's a
1:49:27 link there's an invite to the Discord
1:49:29 there and just go join it and then I'll
1:49:31 see I'm going to see a bunch of people
1:49:32 have joined tonight if indeed anyone
1:49:34 joined I'll just add you as as
1:49:36 Irregulars if you want to join that
1:49:39 channel
1:49:40 and here's the trick if I put you in the
1:49:42 irregular Channel and you're a poop head
1:49:44 you're going to be right out
1:49:46 no poop heads in the Irregulars Channel
1:49:48 unless you're actually a regular and
1:49:50 then you know get yourself to the
1:49:53 drugstore
1:49:59 Discord just for Subs no
1:50:02 I don't you know I I thought about doing
1:50:04 a thing just for Subs like I I should
1:50:07 probably just do something for Subs I I
1:50:08 haven't here's the deal I'm an old
1:50:11 [ __ ] guy on tick tock I'm I'm doing
1:50:13 my best to figure it out listen this is
1:50:15 better than it was when I started this
1:50:17 channel
1:50:18 um
1:50:21 I'm starting to get my head around it
1:50:22 and so
1:50:25 I'll I'll I'll figure some [ __ ] out I'll
1:50:28 figure some [ __ ] out but it's just that
1:50:30 it's just a place so what ended up
1:50:31 happening was
1:50:34 a lot of people from here came to the AI
1:50:36 salon and they're like oh this is great
1:50:38 and then and then it was like Hey you
1:50:40 know we want to play in the Discord and
1:50:41 it would be cool if we had an area where
1:50:42 we could just hang out basically and so
1:50:44 that's that's so I made that today so
1:50:47 we'll figure this stuff out but right
1:50:49 now it's it's just if you go to the
1:50:51 Discord and
1:50:52 you you want to regularly participate
1:50:55 here then great go do that and we'll
1:50:57 hang out
1:50:59 um
1:51:11 Kyle what's your schedule this holiday
1:51:14 weekend well Tobias
1:51:17 today I'm going to be doing the AI
1:51:19 learning Lab live
1:51:21 I'll start around 8 p.m and I'll go
1:51:24 that's Mountain Standard Time thank you
1:51:26 shabbat I like gummies gummies tasty oh
1:51:31 Share Bear reached level 18
1:51:34 is closer to the promised land
1:51:37 the AI learning lab is not dissimilar to
1:51:40 Scientology okay we have levels
1:51:44 I'm gonna hire you achieve the closer
1:51:47 you are to
1:51:49 the thing
1:51:51 that shall not be spoken we shall not
1:51:53 spake
1:51:55 the name micro reach level 10. it's a
1:51:59 competitive Bunch the the Irregulars
1:52:03 that's awesome
1:52:07 I have no idea what this channel is
1:52:09 about well probably a potato first of
1:52:12 all it's about
1:52:13 um
1:52:14 it's about judging people's names and
1:52:17 you have a good one but again your
1:52:18 avatar sucks uh
1:52:21 the channel is just really about
1:52:23 learning about this [ __ ]
1:52:26 um I'll answer questions if people have
1:52:27 questions about AI I'll answer those
1:52:30 um if you want to see something demoed
1:52:32 I'll demo it
1:52:34 um
1:52:34 I'm just gonna sort of talk [ __ ]
1:52:37 about [ __ ] because I do
1:52:40 and
1:52:41 if you like it cool and if you don't
1:52:44 that's cool too
1:52:45 one of the things that I've learned
1:52:47 about this over the many years of my
1:52:49 life I'm an older fella I don't know if
1:52:51 you know that I look 25 I know that I'm
1:52:55 a little older than that anyway one of
1:52:57 the things I've learned in my life
1:53:00 is that
1:53:03 one of the most powerful things when it
1:53:05 comes to this kind of Channel
1:53:09 and storytelling in general is to have a
1:53:12 point of view
1:53:14 because if you don't have a point of
1:53:16 view and you try to please everyone then
1:53:18 you have Mambi Pam be boring [ __ ]
1:53:22 the problem with having a point of view
1:53:24 is that some people don't like it
1:53:29 they're like he's stupid
1:53:32 the other night someone said you're
1:53:33 swearing makes you look tacky
1:53:37 well your lack of swearing makes you
1:53:39 look like a [ __ ] snore
1:53:41 [Laughter]
1:53:43 to me
1:53:45 so I don't know so it's just me like
1:53:48 shooting my [ __ ] mouth off and you
1:53:50 know it's like this Channel's evolved
1:53:53 right it it started out very much just
1:53:55 kind of pure educational stuff and then
1:53:57 we started having fun and then some
1:53:59 people started showing up regularly and
1:54:01 then we started having fun and building
1:54:03 a community so I don't know what it is
1:54:05 like it's I don't know I've been doing
1:54:07 these lives for about three months the
1:54:08 Channel's been around since
1:54:11 December or January
1:54:13 and I put up educational content so the
1:54:15 whole channel is just like [ __ ] that I
1:54:17 found on you know when I'm learning
1:54:19 about AI it was really just a way to get
1:54:21 the [ __ ] out of my head
1:54:24 and and kind of into the world because I
1:54:27 think that's actually really important
1:54:28 not not [ __ ] out of my head but like
1:54:29 stuff that I'm learning put it put it in
1:54:31 the world in a way that people can learn
1:54:34 and then I started doing these lives and
1:54:35 this channel has kind of shifted it's
1:54:37 it's now more about the lives than it is
1:54:39 about the recorded video I'm still
1:54:42 wow fantastic Winston that was beautiful
1:54:45 thank you
1:54:47 good Lord that was very nice
1:54:50 um
1:54:52 yeah so so that's what this is and so
1:54:55 what is it I don't know it's an
1:54:57 experiment
1:54:58 it's an experiment
1:55:02 what's the future of this look like I
1:55:04 don't know these are experiments if you
1:55:06 play with this software it's quite clear
1:55:07 their experiments they're janky pieces
1:55:09 of [ __ ] because we're new we're new here
1:55:11 Djokovic won the third set
1:55:16 um
1:55:17 that's what it's about it's about
1:55:18 exploration
1:55:20 so anyway Tobias I was talking I was
1:55:22 talking before
1:55:26 um I'll probably do one of these over
1:55:29 the holiday weekend why would I not
1:55:32 it's Labor Day Stevo thank you for the
1:55:35 trumpet
1:55:37 um
1:55:40 Labor Day is interesting I've always
1:55:42 labor day's always been really
1:55:44 interesting
1:55:45 to me I never quite understood it
1:55:50 but
1:55:51 we're moving into a world where these
1:55:54 tools are going to start doing a lot of
1:55:56 the labor
1:55:59 you know I think I think the
1:56:02 the manual labor world for years has
1:56:05 been confronting automation right
1:56:09 and
1:56:12 robotization
1:56:14 and the knowledge workers of the world
1:56:17 thought they were immune
1:56:20 and the creatives of the world thought
1:56:21 they were immune
1:56:23 and they're not we're not
1:56:26 we're not
1:56:28 so it's it's interesting Labor Day
1:56:30 weekend maybe it's probably more
1:56:32 important to talk about this [ __ ] this
1:56:34 weekend
1:56:35 Stevo remember level 14 with the gummy
1:56:39 thank you very much congrats
1:56:42 on your achieving a new level ascending
1:56:45 to the uh to the tip of the pyramid
1:56:52 um
1:56:53 yeah I think this stuff's really
1:56:55 important to talk about because
1:57:00 what work is is not going to be the same
1:57:03 five years from now as it is today
1:57:06 where our value is in the workplace and
1:57:10 in society is is about to make a massive
1:57:20 I don't know reinvention
1:57:26 and part of that scary as [ __ ] but part
1:57:29 of that's really exciting here's here's
1:57:31 the part I'm excited about
1:57:33 think about all the people whose work is
1:57:36 that they hate
1:57:38 but they do it because they have to do
1:57:39 it but as a society people have to do
1:57:42 that repetitive
1:57:44 soul-crushing work
1:57:47 well you know what these things are
1:57:49 really good at
1:57:50 repetitive soul-crushing work
1:57:53 go read through all of these data files
1:57:55 and and see what's in them
1:57:58 okay
1:58:00 eyes bleeding
1:58:02 um it appears to be Json data of the
1:58:06 shapes of States uh over here and then
1:58:09 this one appears to be Chipotle
1:58:11 locations and that took that person I
1:58:14 don't know two hours three hours five
1:58:17 hours ten hours depending on how shitty
1:58:19 the data was
1:58:21 and now you pop it into Chachi PT's
1:58:24 Advanced data analytics and say what's
1:58:26 in these files Bang
1:58:30 could you make that into a map and plot
1:58:32 the Chipotle locations accurately for
1:58:35 each state sure Bang
1:58:40 the scary part of that is what the [ __ ]
1:58:42 does the data analyst do
1:58:44 well
1:58:45 the data analyst could probably use a
1:58:48 tool like Advanced data analytics in
1:58:50 chat gbt in remarkable [ __ ] ways if
1:58:55 if we can figure out how to make it plot
1:58:58 you know restaurant locations on a map
1:59:01 of the states imagine someone that knew
1:59:03 what they were doing could do with it
1:59:07 but what does that mean for that
1:59:08 profession what does that mean for that
1:59:10 job what does that mean for how you get
1:59:12 hired what is the skill
1:59:14 if the skill isn't the ability to read
1:59:18 thousands of rows of data or write you
1:59:22 know Python scripts to do the same
1:59:24 if that just happens now
1:59:26 what's the skill
1:59:28 I think the skill pops up a level to
1:59:31 critical thinking
1:59:33 how are we going to use this tool why do
1:59:35 we need to use this tool what are we
1:59:37 trying to accomplish so yeah I think
1:59:39 I'll do this all weekend
1:59:41 I'm not going anywhere if that's what
1:59:43 you're asking
1:59:47 what is back propagation it's a thing
1:59:52 I don't have my head around
1:59:55 but I know it makes these tools better
1:59:57 um it's it's it's something to do with
1:59:59 the training and how the how the tools
2:00:01 work how they how they fine tune them I
2:00:03 don't know enough about it to be of any
2:00:05 use I can tell you how they basically
2:00:07 work but I don't have back propagation
2:00:09 in that conversation
2:00:11 fun our Jam is fun mixed with artificial
2:00:13 knowledge
2:00:14 see what I did there because we're
2:00:16 clueless exactly jrc see what's the
2:00:19 difference between fine tuning and
2:00:20 embedding with open AI it sounds like I
2:00:22 saw this before it looks like you're
2:00:24 helping answer one another
2:00:26 um
2:00:27 so so fine-tuning is so when you have
2:00:32 GPT now
2:00:33 they've added a thing called custom
2:00:35 instructions
2:00:38 and so
2:00:40 um
2:00:42 this is like what would you like chat
2:00:44 EBT to know about you you're a master
2:00:46 gardener who specializes in giving
2:00:48 gardening advice how would you like chat
2:00:51 GPT to respond when anyone asks you a
2:00:53 question about gardening before you
2:00:54 answer them ask two clarifying questions
2:00:56 after they answer provide them with a
2:00:59 short answer and a bulleted list of next
2:01:01 steps so this is fine tuning fine tuning
2:01:05 is where you're taking the core model
2:01:07 and you're saying I want you to act in a
2:01:09 certain way I'm going to give you
2:01:11 examples of well-written emails and
2:01:13 well-written business Communications and
2:01:15 whenever I ask you to do anything follow
2:01:18 the tone and structure of the examples
2:01:20 that I gave you that's fine tuning
2:01:22 embedding is where you take a whole pile
2:01:25 of your company data like let's say
2:01:27 we're going to take Salesforce data for
2:01:30 the past 10 years of my company and
2:01:32 we're going to embed that and that's
2:01:34 going to go into a vector database it
2:01:36 now becomes effectively gptable along
2:01:39 with all of the other chat GPT data so
2:01:42 you're kind of augmenting the larger
2:01:44 model with a bigger data set with your
2:01:46 own personal data set and this is one of
2:01:49 the things that
2:01:51 um Sam Alton just talked about they they
2:01:53 launched
2:01:55 um chat GPT Enterprise right now it's
2:01:58 just about privacy of data but the next
2:02:02 iteration of it is going to be you're
2:02:04 going to be able to put your company's
2:02:06 data within chat GPT
2:02:08 to be able to search on it complete with
2:02:11 permissions where certain people have
2:02:13 access to certain pieces of information
2:02:14 so embedding is think of it like a
2:02:17 larger data set where you're actually
2:02:19 able to pull
2:02:22 um pull relevant data out just as if it
2:02:25 was in the chat GPT model fine tuning is
2:02:27 more about steering it in in a direction
2:02:30 or two you know as it answers you
2:02:34 so hopefully that helps chat EBT does
2:02:36 great for springboarding code projects
2:02:39 yes
2:02:41 it's very good at that
2:02:44 wait let's get up let's get up one of
2:02:46 our pretty pictures we haven't had any
2:02:48 pretty pictures up here for a while
2:02:49 let's see let's see which one do we want
2:02:53 um that one's cool
2:02:55 we're gonna go
2:03:02 the eye the all-knowing eye sees all if
2:03:07 you don't think we're watching
2:03:13 we're watching all the time all right
2:03:16 maybe do questions only on certain days
2:03:19 show tell Tuesday Thursday for example
2:03:21 oh that's not a bad idea Ted looks at
2:03:24 life
2:03:27 yeah maybe that's a good idea maybe
2:03:29 that's
2:03:32 because I could make little Graphics now
2:03:34 now that we have a tool that makes words
2:03:36 I could make a little graphic that's
2:03:38 like q a day
2:03:40 Show and Tell day
2:03:42 so it's got a little Focus
2:03:45 that's not bad
2:03:47 it's a little structured for my taste
2:03:49 because you know I like structure but at
2:03:51 the same time my ADD says talk about
2:03:54 whatever the [ __ ] you want to talk about
2:03:56 don't listen to Ted Ted's just trying to
2:03:58 put you in a box
2:04:00 I don't want to be the Box test putting
2:04:03 you in a box
2:04:04 tell Ted to [ __ ] up but Ted's got a
2:04:06 point shut up
2:04:07 shut up
2:04:10 nobody put baby in a corner
2:04:13 wire hangers
2:04:16 [Laughter]
2:04:22 someone's scrolling through Tick Tock
2:04:24 they just saw that they're like the [ __ ]
2:04:26 is this
2:04:26 [Laughter]
2:04:29 ah
2:04:31 that's impressive I don't know that's
2:04:33 impressive love the XML storyline
2:04:35 example that yeah I'm telling you that
2:04:38 blew my mind not only could it read our
2:04:40 XML but it basically
2:04:43 completely rewrote the syntax the like
2:04:46 the like the whole schema it created a
2:04:48 whole new schema and it was perfect
2:04:51 it wasn't close it was perfect it was
2:04:53 exactly how I was thinking about it
2:04:57 because I gave it enough context and I
2:04:59 said here's what this is here's what
2:05:00 that is here's how we do [ __ ] now how
2:05:02 would you take that and
2:05:04 make it match the structure of this
2:05:09 perfect
2:05:12 it was just magical it was just magical
2:05:14 magical magical magical prompt
2:05:18 what amazing thing do I want to start
2:05:21 exactly probably a potato so I've talked
2:05:25 about this before where where I think
2:05:27 that
2:05:28 I think that these tools are um
2:05:34 they're kind of the kind of short
2:05:36 circuit the 10 000 hour argument right
2:05:39 the Malcolm Gladwell thing that if you
2:05:41 want to get any level of masteries at
2:05:43 something you need to put in 10 000
2:05:44 hours not anymore like
2:05:47 if you put in 500 hours with some
2:05:50 focused work on these you're probably at
2:05:53 a pretty masterful [ __ ] level
2:05:58 and so what is that like the
2:06:00 implications of that are [ __ ] insane
2:06:02 aren't they
2:06:04 like if I said if you know if I said
2:06:08 you know I want to get a I want to get a
2:06:10 good at technical writing
2:06:12 like it would take me years I might go
2:06:13 well let me go back to school let me
2:06:15 learn about it let me do that
2:06:19 I could put in 100 hours with this and
2:06:22 be doing
2:06:24 not world class but like
2:06:28 solid C plus level professional
2:06:30 technical writing and if I'm a little
2:06:32 clever maybe get that to kind of B minus
2:06:35 to be technical writing in a hundred
2:06:37 hours of training
2:06:39 huh
2:06:41 well that's interesting
2:06:43 so then I could take another 100 hours
2:06:45 and I could get to sort of B level work
2:06:48 and then I could take 200 hours here and
2:06:50 500. so in a thousand hours I could be
2:06:54 decently masterful at four or five
2:06:56 things that are not in my wheelhouse
2:06:59 right now
2:07:04 what do you do with that
2:07:06 well
2:07:08 an option is [ __ ] reinvent yourself
2:07:11 because you don't have to commit three
2:07:13 or four or five years now to know if you
2:07:17 enjoy
2:07:18 being masterful at that thing
2:07:21 I could figure out in a hundred hours
2:07:24 now that I'm good at technical writing
2:07:26 oh I [ __ ] hate this
2:07:29 but if I've put in three or five years
2:07:33 at technical writing and maybe even I've
2:07:35 paid for that and now I'm in debt and I
2:07:37 have to pay it off
2:07:39 I've got an emotional overhang of I
2:07:42 spent five years of my life getting to
2:07:44 this level with technical writing and
2:07:47 now I realize I hate it but [ __ ] I gotta
2:07:49 pay off this loan so let me just go work
2:07:52 for 10 years doing that and the whole
2:07:54 time I'm going oh my god I've made a
2:07:56 terrible mistake well what if that what
2:07:59 if that equation went away
2:08:02 and you could just experiment 100 hours
2:08:04 here 100 hours there you do 10 things
2:08:07 and you find out two of them you [ __ ]
2:08:09 love
2:08:11 oh
2:08:12 then you go put in 500 hours on each of
2:08:14 those all these tools at the center the
2:08:17 whole time
2:08:20 you're you'll be a completely different
2:08:22 human being
2:08:25 everyone has that opportunity now
2:08:27 everyone everyone if you're in this
2:08:30 channel go to [ __ ] chat GPT
2:08:34 go here now if you haven't go there now
2:08:37 check out openai.com
2:08:40 start talking to it go to prompts.chat
2:08:42 that'll teach you how to talk to it
2:08:46 experiment with Claude and bard
2:08:49 those three just start with the top one
2:08:52 learn how to talk to it go start talking
2:08:54 to it
2:08:58 like
2:08:59 is it me
2:09:01 I mean that's profound right
2:09:03 I don't have to invest three to five
2:09:05 years to try something out anymore
2:09:09 I can be executing at a professional
2:09:11 level
2:09:13 in
2:09:14 a month
2:09:16 if I just focus
2:09:18 for a month
2:09:20 not three years not five years
2:09:26 so yeah that's a good product that's a
2:09:28 good prompt probably a potato is see I
2:09:32 think having the self-awareness of
2:09:33 knowing that you're probably a potato
2:09:34 means that you're probably not a potato
2:09:37 which is why I love the name it's a 10
2:09:39 out of 10 name it's a solid name
2:09:41 makes General language model do specific
2:09:43 tasks oh great so yeah you're talking
2:09:45 about uh embedding and fine-tuning like
2:09:48 responding in a certain way as a company
2:09:49 chat bot I read what is prompt
2:09:52 engineering data camp today cool very
2:09:54 nice embedding usually refers to the
2:09:57 process of representing data as a vector
2:09:59 a sequence of numbers yep getting large
2:10:01 data sets into gptable form is how I
2:10:04 like to think about it similar words
2:10:06 have similar embeddings yep like dog and
2:10:09 good boy would have similar vectors
2:10:11 right and dog can also like dog that you
2:10:15 pet and love could be in a different
2:10:17 Vector space semantic cluster then he's
2:10:21 a dog and treated his lady wrong right
2:10:25 same word two very different semantic
2:10:27 clusters so that's embedding takes all
2:10:30 the words and blows them into these
2:10:34 um this Vector space which is
2:10:35 fascinating like when you what's what's
2:10:38 more what's remarkable about these tools
2:10:40 is that the more you learn about them
2:10:42 the more [ __ ] magical it is that they
2:10:45 do what they do
2:10:47 it just doesn't make sense that that it
2:10:49 works
2:10:51 but it does
2:10:53 and so like kudos to the [ __ ] super
2:10:56 uber nerds who figured this [ __ ] out
2:11:00 Jeffrey Hinton and the the
2:11:02 cats at Google that wrote the
2:11:04 Transformer paper
2:11:07 you know
2:11:09 the open AI team for going [ __ ] it let's
2:11:11 just go get all the data we can Let's
2:11:14 Jam as much [ __ ] down this pipe as we
2:11:16 can until it explodes and it didn't
2:11:18 explode it got remarkable
2:11:22 and then they put out chat GPT so usmere
2:11:25 Mortals could [ __ ] play with it
2:11:27 and that changed everything
2:11:32 boom
2:11:37 laughing my ass off I really just didn't
2:11:39 change the Avatar probably a potato you
2:11:41 have a world-class name and a generic
2:11:43 Avatar come on come on dude
2:11:46 you probably are a potato if you got
2:11:48 that Avatar
2:11:50 oh that's funny Sam corn given a lesson
2:11:53 on embeddings this is super cool it's
2:11:56 ever accelerating and it will leave you
2:11:57 behind this stuff will leave you behind
2:11:59 absolutely you can feed it your
2:12:01 company's data and embed it and use it
2:12:02 to answer questions about the company
2:12:04 exactly I had Internet Explorer before
2:12:07 images moved yep I remember those days
2:12:09 at agency.com we like I remember us
2:12:12 writing code
2:12:15 to animate um images
2:12:18 we would we would basically have you
2:12:20 know 10 different frames of an animation
2:12:22 and we wrote the code to call them one
2:12:25 after another inside a web page it was
2:12:27 all it was it was all server calls it
2:12:29 was amazing
2:12:31 I'm so sick of tick tock's character
2:12:33 limits there's so much about Tick Tock
2:12:35 to be sick of
2:12:38 um but it's hey we're here we get to
2:12:40 hang out so that's cool
2:12:44 that date was the event horizon it
2:12:46 exactly was
2:12:48 November 30th 2022 no more BCAA d c e or
2:12:52 vce now that's bgvt and agbt exactly
2:12:58 what did someone called this channel
2:13:00 chat add exactly
2:13:06 um
2:13:08 when it helped me to write an Lua for an
2:13:11 obscure game server I run I realized it
2:13:14 will be massive yeah like like that's
2:13:17 the thing when you take specific
2:13:19 projects
2:13:20 and you're like oh this is a thing I've
2:13:23 been ignoring for you know 10 years
2:13:26 maybe I could see if Chet gbd could do
2:13:29 something with it oh my God it did it
2:13:32 all
2:13:33 it's amazing
2:13:34 the regular here walk me through the
2:13:36 Discord Channel Sign up all right
2:13:42 let's see I've got
2:13:45 um the salon.ai
2:13:49 so you go here
2:13:52 to the salon.ai
2:13:56 and then you click on Discord community
2:13:59 and then that'll take you to this thing
2:14:03 and you say accept invite and then it'll
2:14:06 bring you into
2:14:09 I forget what channel it brings you into
2:14:13 is it browse channels no
2:14:17 it might bring you in to introduce
2:14:19 yourself
2:14:21 um but it brings you into some channels
2:14:23 so the way Discord Works you've got down
2:14:25 this side you've got at the very top if
2:14:27 anyone direct messages you they'll show
2:14:29 up up here then each one of these is a
2:14:32 different server right so think
2:14:34 different community so this one is the
2:14:36 is the AI Salon Community that's Adobe
2:14:38 that's Runway ml that's a thing called
2:14:40 suno somewhere down here is mid Journey
2:14:43 there's Neo Tokyo is an nft project I'm
2:14:45 a part of there's there's mid-journey so
2:14:48 these are all different communities and
2:14:50 then when you're in a community all
2:14:52 these things down the left are different
2:14:53 channels so basically
2:14:56 um you have to answer some [ __ ] like
2:14:58 when you first get into Discord I think
2:15:00 we make you click on something and then
2:15:03 you might need to say follow these rules
2:15:05 don't be an [ __ ] you know whatever
2:15:07 contribute nicely I don't whatever it
2:15:10 says and then you say yes I agree to
2:15:12 that once you agree to that then you're
2:15:14 able to see all of the channels down the
2:15:16 side
2:15:18 and then you know there's General
2:15:20 discussion channels there's news
2:15:22 channels there's a mid-journey channel
2:15:24 where you can play with mid Journey like
2:15:26 these are some of the images I was
2:15:27 making earlier for that spreadsheet I
2:15:30 showed where you know making all sorts
2:15:31 of cool
2:15:33 fantasy kind of kind of stuff
2:15:37 like all of these prompts were written
2:15:39 by it was actually Claude
2:15:43 that was a cool project I'll show you
2:15:45 how I did that that was pretty cool so
2:15:47 there's that and then somewhere in here
2:15:48 I've got
2:15:50 um
2:15:53 where is it here's an AI learning lab
2:15:56 Channel
2:15:57 and this is a private channel so if
2:15:59 you're an irregular I'll give you a roll
2:16:01 and then you'll be able to hang out in
2:16:03 there so I I hope that answers it
2:16:07 the the exact mechanism for how you sign
2:16:10 up once you're in there I'm not as clear
2:16:11 on
2:16:12 because I just I just don't remember
2:16:14 remember what's in there
2:16:17 [Music]
2:16:25 AI today on our hands is like giving a
2:16:28 screwdriver to caveman totally we don't
2:16:30 know what to do with it that's exactly
2:16:32 right multi-dimensional man you are
2:16:34 multi-dimensional
2:16:35 um
2:16:38 it is it's really hard
2:16:41 I like
2:16:44 the hardest thing about it right now is
2:16:46 understanding
2:16:49 what do you actually do with this thing
2:16:51 that can do all of this stuff
2:16:55 what do you do with that what do you do
2:16:56 with that knowledge what do you do with
2:16:58 that capability
2:17:03 what are the death throws of companies
2:17:06 going to be like that have their
2:17:08 business models ripped out from under
2:17:10 them
2:17:12 like I know the advertising world really
2:17:14 well
2:17:16 they make their money
2:17:18 for the most part on time and materials
2:17:21 it used to be they made their money on
2:17:23 media buys and they could give away the
2:17:25 creative right and they would make their
2:17:28 10 on the amount of TV ads they bought
2:17:32 because there were only three networks
2:17:33 and it was really [ __ ] easy and
2:17:34 that's why they had the mark three
2:17:36 Martini lunch where you take
2:17:39 someone out that had a lot of money and
2:17:41 say do you want to be on ABC CBS or NBC
2:17:44 I want to be on all three I want to blow
2:17:47 this out fantastic Bob bring us another
2:17:50 round
2:17:51 that was all it was you'd sell all that
2:17:53 [ __ ] you'd give them the creative for
2:17:55 free you'd make ten percent of their
2:17:57 billion dollar buy you got a hundred
2:17:59 million dollar agency that was simple
2:18:01 then cable TV came along and [ __ ] that
2:18:03 up
2:18:04 and then the agency model changed to
2:18:07 well
2:18:09 we gotta at least cover our costs on the
2:18:12 creative because
2:18:14 we're not making as much on the media
2:18:16 buy because the media buys are all
2:18:17 totally [ __ ] fragmented
2:18:19 and then the internet came along and
2:18:22 atomized it right so it was fragmented
2:18:25 because of cable television and then the
2:18:27 internet comes along and blows it
2:18:28 [ __ ] out of the water and then now
2:18:31 there's all sorts of AD fraud and
2:18:33 internet fraud and nobody knows what the
2:18:35 the ultimate measurable
2:18:39 um
2:18:40 institution of the internet where you
2:18:42 can measure everything becomes an
2:18:44 absolute [ __ ] [ __ ] show because of
2:18:45 advertising fraud and fake audiences and
2:18:48 click bait and
2:18:50 flake fake engagement and all that sort
2:18:52 of [ __ ]
2:18:54 but agencies made their money on time
2:18:56 and materials so they said to big
2:18:59 companies well we'll take just a little
2:19:00 bit on the media buys but we're not
2:19:02 really making money on that so we're
2:19:03 just gonna charge you a 120 hour 120
2:19:07 bucks an hour for the junior people and
2:19:08 200 bucks for the middle people and 300
2:19:10 or 400 bucks an hour for the important
2:19:13 people
2:19:14 and based on the work that you want to
2:19:16 do we think that's going to take 50 man
2:19:18 hours per but and that's how we make our
2:19:22 money
2:19:23 okay and then Along Comes A Tool
2:19:27 where the agency can do the same amount
2:19:30 of work
2:19:31 with some percent less of the people
2:19:35 they're like but that's how we make our
2:19:37 money we want those people doing that
2:19:39 work
2:19:40 which is Gonna Fly for some time
2:19:44 until the clients go uh let me get this
2:19:46 straight
2:19:48 with these tools you get a 50 to
2:19:53 let's be let's be let's be conservative
2:19:55 you get a 25 to 50 percent productivity
2:19:58 increase
2:19:59 but you're still charging us the same
2:20:01 amount of money is that am I
2:20:02 understanding that correct
2:20:04 okay
2:20:06 and then some snot-nosed little punk
2:20:09 agency comes along that's AI Centric and
2:20:12 says we'll do what they do
2:20:15 twice as fast just as good for half the
2:20:19 price
2:20:21 what's that agency do
2:20:23 I don't [ __ ] know
2:20:27 are they gonna flush
2:20:29 their own business model down the toilet
2:20:32 are they gonna have to have it flushed
2:20:33 down the toilet for them
2:20:36 it's gonna be painful either way and
2:20:38 like that's that's like one industry
2:20:40 that I know there's like many many
2:20:42 Industries I don't know [ __ ] about but
2:20:44 they're going to have equally as
2:20:45 profound ripples run through them
2:20:50 nuts right
2:20:53 Kyle you're my best friend hey thank you
2:20:57 Matthew so I'm sorry about that
2:21:01 I'm so happy to be here for this AI
2:21:04 inflection point I honestly wait I
2:21:07 honestly believe if human if Humanity
2:21:09 can
2:21:10 wait what is it where Winston if
2:21:13 Humanity can where's the rest of your
2:21:14 thing
2:21:16 oh [ __ ] it just jumped me to the bottom
2:21:18 yeah I lost your thing Winston that was
2:21:20 cool
2:21:21 but yeah this is where where this is
2:21:23 we're at a remarkable time
2:21:26 um
2:21:27 ah Winston back propagation short for
2:21:29 backward propagation of errors after
2:21:31 optimization algorithm for neural Nets
2:21:34 cool
2:21:35 nice it's based on the Chain rule in
2:21:38 calculus which is very basic
2:21:42 iOS just told me I need to go to bed
2:21:44 which means all of you need to go to bed
2:21:47 but it does get more confusing in
2:21:49 algorithm because you need to know about
2:21:51 the loss function see this is C Winston
2:21:54 this is what this is where whenever I
2:21:57 would say
2:21:59 no I read you know Google's like
2:22:01 tensorflow just got easier I'd go there
2:22:03 and it like tensorflows really easy and
2:22:05 I'd understand the first part of the
2:22:07 concept and then they would do [ __ ] like
2:22:09 and now you need to understand about the
2:22:11 loss function and I'm like all right
2:22:12 [ __ ] it you lost me
2:22:15 I must have tried that 10 times and then
2:22:18 you know I finally discovered gpt3 and I
2:22:21 was like okay now now it's starting to
2:22:23 come up to like human level
2:22:26 it's amazing and Limitless how these
2:22:29 tools can be used in all aspects of
2:22:30 daily life true true true I'm going to
2:22:33 need an AI anonymous group soon that's
2:22:35 what the salon is we actually joke about
2:22:38 that that that um
2:22:39 that that this group is like you know hi
2:22:42 I'm Kyle I'm into AI hi Kyle
2:22:45 I've been a uh I've been into AI for
2:22:47 like a year and a half now it's still
2:22:49 remarkable to me my jaw still hits the
2:22:51 floor
2:22:52 I I need to come up with words other
2:22:54 than face melting and remarkable and
2:22:57 incredible and I don't know what to say
2:22:58 anymore my friends all think I'm weird
2:23:02 do you think future higher ed will exist
2:23:05 to teach how to harness AI specific to
2:23:07 the profession
2:23:11 yeah I think I think there's always
2:23:13 going to be a need for educational
2:23:14 institutions I do think they need to be
2:23:17 reinvented
2:23:20 um
2:23:23 see I don't think AI is going to be a
2:23:26 profession
2:23:27 like
2:23:28 like we go we go through this cycle with
2:23:31 with technology
2:23:32 pretty much every time where
2:23:37 when
2:23:38 the web first came along there were sort
2:23:41 of Chief
2:23:42 internet officers or chief web officers
2:23:45 or whatever the [ __ ] they called them
2:23:46 because
2:23:48 the web was this thing off to the side
2:23:50 and then
2:23:51 it was like oh but every company needs a
2:23:54 website and it's actually part of the
2:23:55 marketing function and then the the
2:23:57 intranet is part of the it function and
2:24:00 it sort of found its way back into the
2:24:02 organization and then and then digital
2:24:04 happened and we got mobile phones and we
2:24:06 got social media and we got all this
2:24:07 stuff and then it was like Chief digital
2:24:09 officer right and and digital was this
2:24:12 sort of ghetto off to the side where
2:24:14 that's all those young punks and we're
2:24:16 just running our business over here and
2:24:18 we're the professionals and those young
2:24:20 snot-nosed little kids with their
2:24:22 digital well oh oh it's actually getting
2:24:24 important now uh make Charlie uh Chief
2:24:27 digital officer yeah yeah he's gonna do
2:24:29 all of the digital stuff and then the
2:24:31 digital stuff like everything becomes
2:24:32 digital right and that that role becomes
2:24:35 unnecessary and now we're going to do
2:24:37 the same with AI where no there's AI
2:24:40 stuff it's uh yeah it's powerful
2:24:41 whatever but we're over here we do we
2:24:43 run our business and we're all digitally
2:24:45 integrated so this is going to be easy
2:24:47 for us so oh oh but make Jane the chief
2:24:50 AI officer and then
2:24:52 so it's it's the same [ __ ] over and over
2:24:54 again
2:24:56 um
2:25:00 short term I think you need someone to
2:25:02 focus on this [ __ ] but I think this
2:25:04 stuff is gonna assimilate into business
2:25:07 a lot faster than the web did and a lot
2:25:09 faster than digital did
2:25:11 um just just because there's too many
2:25:12 productivity games to be had
2:25:14 so businesses are going to take
2:25:16 advantage of that
2:25:18 um what they do with the extra
2:25:20 productivity
2:25:22 without needing as many people to do it
2:25:25 that's a fascinating question
2:25:28 that I don't have a [ __ ] answer for
2:25:30 it either looks like Universal basic
2:25:32 income as a bridge until we figure out
2:25:34 what the new jobs are or it looks like
2:25:37 all companies say hey let's keep all of
2:25:40 our employees and just increase our
2:25:42 productivity I don't think that's going
2:25:43 to happen either
2:25:45 so I don't know I do not know on that
2:25:48 front I just don't know I just don't
2:25:50 know
2:25:51 that's another t-shirt from my Kyle
2:25:53 printify store chat add okay all right
2:25:56 yeah so actually you know that's
2:25:58 something that we could do in the
2:26:00 Irregulars channel on Discord is figure
2:26:02 out you know like what are the what are
2:26:04 the what are those the the uh this the
2:26:08 the merch we want to do
2:26:11 um what you do the funny custom settings
2:26:15 oh can you do the funny custom settings
2:26:17 again for chat GPT to perfect style you
2:26:21 did a funny one what did I do I did um
2:26:23 [Music]
2:26:24 oh I know what I did all right yeah I'll
2:26:27 do that so if if you're new here it
2:26:29 looks like we had a bunch of new people
2:26:30 here so this is chat gbt one of the
2:26:32 things chat EBT has now is a thing
2:26:34 called custom instructions
2:26:36 so
2:26:38 you can tell it who you want it to be
2:26:39 and how you want it to answer so I'm
2:26:42 going to turn it on
2:26:44 and now I can edit these so the one I
2:26:46 did was something in the neighborhood of
2:26:47 this
2:26:48 um
2:26:54 you are a pompous
2:27:00 over intellectual
2:27:03 academic
2:27:05 who fancies themself
2:27:09 a theatrical period
2:27:14 you take great pleasure
2:27:17 in using large words
2:27:19 that neither you nor your audience
2:27:22 understands
2:27:24 in an attempt to make yourself seem way
2:27:27 smarter than they are period
2:27:31 okay so that's that's who it is so so
2:27:35 now we're going to say
2:27:37 um
2:27:39 no matter what anyone asks you
2:27:46 respond with an obscure reference
2:27:51 to a Broadway play
2:28:00 or or a piece of
2:28:03 overly flowery literature
2:28:13 and include a quote
2:28:16 from that piece period
2:28:21 when you answer them
2:28:22 answer is if
2:28:26 uh
2:28:30 your Tennessee Williams writing a
2:28:33 masterpiece
2:28:36 all right save okay so now no matter
2:28:39 what I ask chat GPT it should be an
2:28:43 insufferable ass
2:28:44 no
2:28:47 all right so let's let's do this
2:28:50 um hey I'm
2:28:53 wondering if you could help me
2:28:57 understand
2:29:00 um
2:29:05 what marketing is
2:29:11 ah the Grand Theater of Commerce where
2:29:14 illusion and reality dance
2:29:17 um
2:29:21 marketing my dear interlocutor is the
2:29:24 Blanche Dubois of the business world
2:29:27 forever reliant on the kindness of
2:29:29 strangers yet weaving The Narrative so
2:29:31 compelling you cannot look away in the
2:29:34 words of Tennessee Williams himself I
2:29:37 don't want realism I want magic
2:29:40 such is the essence of marketing a
2:29:42 tapestry of promises perceptions and
2:29:45 persuasion Tech persuasive techniques
2:29:47 to entice the audience to partake in a
2:29:50 particular act to be to be to to be it
2:29:54 the purchase of a product or the
2:29:56 embracing of a brand ah yes just as in
2:29:59 the Phantom of the Opera where the mass
2:30:01 figure lures Christine into a world of
2:30:03 Melody and shadows so does marketing
2:30:06 beckon consumers into its seductive
2:30:08 grasp it uses various channels be they
2:30:11 social digital or traditional to perform
2:30:14 its siren song a tune uniquely crafted
2:30:16 to Captivate them and it goes on and on
2:30:19 so
2:30:21 so even though this is funny like what
2:30:24 what becomes clear here is you can make
2:30:27 this thing act in very specific ways
2:30:30 this is the fine tuning
2:30:32 um and they just chat GPT just added
2:30:35 um like server level fine tuning
2:30:38 um or API API level fine-tuning where
2:30:41 you could start now now making Bots with
2:30:44 your own interface not just the chat GPT
2:30:46 interface you can make your own
2:30:48 applications that one might act like a
2:30:51 sales person one might act like a
2:30:53 business coach one might act like a like
2:30:55 a like a like a human resources
2:30:59 um you know uh you know counselor right
2:31:03 in in very very specific ways and and
2:31:06 you know like I did this kind of
2:31:08 jokingly but you could put in examples
2:31:10 of writing or examples of thing that
2:31:12 things that you wanted to draw from and
2:31:14 I mean you know the
2:31:17 the examples that it drew
2:31:19 are actually they're actually quite
2:31:22 accurate right you know the
2:31:25 marketing is the Blanche Dubois of the
2:31:27 business world forever relying on the
2:31:30 kindness of strangers yet weaving a
2:31:31 narrative so compelling you cannot look
2:31:34 away
2:31:36 in the words of Tennessee Williams
2:31:37 himself I don't want realism I want
2:31:39 magic this is the essence of marketing I
2:31:42 mean that's [ __ ] good
2:31:44 now is it pompous and ridiculous yeah
2:31:48 but it's good it's like
2:31:52 that's what I asked it to do and it did
2:31:54 it quite well so
2:31:56 yeah crazy right
2:31:59 [ __ ] nuts it's nuts Italian
2:32:02 I gotta go people
2:32:04 it's going on 11 o'clock at night here
2:32:07 it's Friday night what are you still
2:32:09 doing with me you people are weird did
2:32:11 you know that did you know your cuckoo
2:32:13 you're just cuckoo you're more cuckoo
2:32:16 than I am you think I'm weird you're
2:32:18 weird I'm not weird you're weird
2:32:20 we're all weird
2:32:24 AI learning lab oh yes yes yes I see you
2:32:29 now
2:32:30 rather addiction hmm yes I can say this
2:32:35 if you understand back propagation you
2:32:38 understand a lot about neural networks
2:32:40 Winston Ridge Hardy III my name is
2:32:43 Winston
2:32:45 yeah
2:32:47 I don't quite get back propagation yet I
2:32:50 know that neural networks are important
2:32:52 I know that back propagation is a big
2:32:54 part of that but
2:32:56 I don't know much more than that by the
2:32:59 way Djokovic up
2:33:02 3-1 in the fourth set
2:33:06 so
2:33:08 he's gonna [ __ ] do it because he's
2:33:11 Djokovic isn't it amazing that the top
2:33:13 10 always are the top 10 in tennis it is
2:33:16 such a [ __ ] head game that game
2:33:19 just joined Discord how do I get said
2:33:21 roll uh probably a potato uh just let me
2:33:24 know that you came in there from here
2:33:26 I'll probably notice a bunch of people
2:33:27 join tonight
2:33:29 um so I'll do that but just just let me
2:33:31 know or if your name is probably a
2:33:32 potato and Discord I will recognize you
2:33:34 because you have an awesome name
2:33:36 oh and you gotta you got you got a uh
2:33:39 you got an avatar now congratulations
2:33:42 see
2:33:46 jrc gives probably a potato a potato
2:33:49 Emoji that's awesome
2:33:51 uh was your logo made with ideogram I
2:33:54 don't have a logo these are all ideogram
2:33:57 just AI learning lab makes a mate I what
2:34:00 was cool about these ones I did
2:34:03 um
2:34:05 yeah I did
2:34:08 this was 70s
2:34:12 70s poster AI learning lab
2:34:15 these were 80s this was 90s this was
2:34:18 2000s
2:34:20 that was 2010s
2:34:22 and I think that was 2000
2:34:26 80 or something like that so I just I
2:34:29 was just doing decades just to see what
2:34:31 it came up with and these were what it
2:34:33 came up with
2:34:34 [ __ ] insane isn't it
2:34:36 and this was this was ideogram too
2:34:40 I'm not weird I'm a regular
2:34:44 all right everyone listen listen listen
2:34:46 listen listen listen
2:34:48 I think you're all fantastic I need to
2:34:51 go to bed I need to go to bedibon
2:34:54 um
2:35:01 I'm gonna take off I will be back
2:35:02 tomorrow I'm gonna be working on some
2:35:04 [ __ ] tomorrow I got
2:35:06 I got a lot to do
2:35:08 and if you're thinking like Kyle it's a
2:35:10 vacation weekend why are you not
2:35:12 vacationing why are you working because
2:35:13 so I like this [ __ ]
2:35:17 I like this stuff so I do whatever work
2:35:19 I need to do for like you know the
2:35:21 domestic [ __ ] and then uh and then I do
2:35:23 this so so yeah I'll be working on [ __ ]
2:35:26 tomorrow good night everybody let's see
2:35:28 we'll do we'll do a little Romper Room
2:35:29 thing here I see multi-dimensional man I
2:35:33 see Togos I see Robert I see Nancy hi
2:35:37 Nancy I see Tobias night Tobias
2:35:41 uh Sam corn thank you good night and
2:35:44 check out the Bonnie and Clyde poem If
2:35:45 you see this I will definitely do that
2:35:48 Ted looks at life peace out thank you
2:35:50 for hanging out time for tomorrow
2:35:52 probably eight o'clock
2:35:54 um bye thanks Kyle 3D classic rock Sam
2:35:56 corn Firefly Winston thank you Emilio's
2:36:00 wife peace jrc Stevo Stevo peace 3G
2:36:05 classic rock all right everybody have a
2:36:07 good night
2:36:08 peace
2:36:10 oh by the way
2:36:14 68 800 likes you all are machines
2:36:17 [Laughter]