Jan 19, 2024 - Unlocking AI Magic: How to Create Engaging GPTs for Fun and Learning

Description

Chapters

Transcript

0:01 you ready you
0:05 [Music]
0:11 [Music]
0:21 ready you and I get
0:25 all Sunday morning here at home
0:31 sky blue and C strong
0:35 [Music]
0:37 true then I open my eyes to dream
0:42 realized in front of
0:46 me and I haven't got a clue what in the
0:49 world is
0:50 happen
0:52 [Music]
0:53 me thank you Robert Rossy think I think
0:57 I'm
0:59 happy like first day summer vacation
1:02 happy got to get a little rest and
1:05 relaxation
1:08 happy on Sunday morning s
1:11 [Music]
1:14 TR Oh Sister when you cook that
1:19 way sweet distraction rules my
1:23 day kitchen table just my brain from the
1:28 sound of you and I rocking up by is yeah
1:32 I don't understand why wasn't your of a
1:35 man hundred years
1:39 ago if they was going to have to
1:42 understand tip their hat to the man who
1:46 [Music]
1:47 scored think I think I'm
1:51 happy yeah CH Shannon champ Shannon on
1:55 lead vocals
1:59 fantastic
2:01 [Music]
2:03 a minor a minor
2:05 [Music]
2:16 dog dog loves an A
2:20 Minor no rapping we don't rap we're
2:24 singing melodious
2:29 tones
2:31 [Music]
2:34 [Applause]
2:34 [Music]
2:40 I got to tune my guitar at some point
2:43 maybe
2:44 [Music]
2:47 never you
2:49 [Music]
2:59 done
3:02 [Music]
3:09 well I heard there was a secret
3:13 dog David played and he set by log you
3:18 don't really care
3:20 for poopy do
3:26 you well it goes like this the fourth to
3:29 fifth M fall
3:32 m b king composing
3:39 hallelujah
3:42 hallelujah
3:45 hallelujah you done
3:51 hallelujah
3:54 hallelujah oh yeah Happy Date night
3:57 everybody cam katkin from Cleveland
3:59 what's happening
4:00 cam is a new mod here
4:03 tonight so troll away she'll have to
4:06 deal with
4:08 it thank you for modding cam sorry so so
4:12 here's the deal cam asked to mod like I
4:15 don't know a week or two ago and I said
4:17 sure and I so I let her in the The
4:19 Mighty networks Channel where the mods
4:20 hang out and then she couldn't ever mod
4:25 like she like Tik Tok was broken she
4:28 could not pin comments it was horrible
4:31 not quite as horrible as my hair but it
4:32 was
4:34 bad turns out I never set her up as a
4:38 mod with most computer problems it is it
4:42 is usually a user error so all right oh
4:46 what a fantastic way to spend a Friday
4:48 night champing Kyle Tunes well that's in
4:52 in doses right you know champ you could
4:54 listen to all night but
4:57 uh you know
5:01 it's like it's like an open mic night
5:03 that no one asked to
5:07 have all right so Cam's in we're all
5:10 good uh oh I don't see the mod icon on
5:12 comm's Cam's name I think she I think
5:14 it's there I think she just pinned
5:16 something I think she's good to
5:21 go all right I just got back back in we
5:25 were out doing a little Gallery
5:27 hopping and uh I eat like a quick little
5:31 meal here so if I get hiccups that's
5:33 what's going on and my glasses are dirty
5:36 looks like champ took a poop on my
5:39 lenses happy Friday night everybody
5:41 Welcome to the AI learning lab um the uh
5:47 I I'm also giving um uh quaffing advice
5:51 for any of you uh looking to do
5:54 something really nice with your hair and
5:56 I am the I am the before picture
6:01 so I'm here for your for your
6:03 inspiration for
6:05 your for for your oh my God what the
6:09 [ __ ] is going on up
6:11 there what are you GNA do and I don't is
6:14 this shirt this shirt nope can't pin huh
6:17 hang on hang
6:18 on hold
6:22 please where are settings
6:27 um
6:28 settings
6:31 moderators I know I set you
6:35 up and you're not in there
6:38 weird
6:40 Plus cam add
6:44 confirm I've already done this Tik Tock
6:47 try it now cam 1504 here for me Saturday
6:51 the 20th so 1504 is I used to be a
6:54 military brat
6:56 3:4 3:04 in the afternoon Saturday at 3
7:00 4 in the afternoon right is that right
7:02 steo don't use military
7:05 time we're in America we can't handle
7:08 the [ __ ] metric system you think we
7:10 can handle military time okay let's try
7:13 it you you pinned that right cam yeah
7:16 all right so cam is now pinning
7:19 away woohoo okay she's happy she's good
7:22 to go everyone's
7:25 excited except whoever did my
7:28 hair they need a
7:33 demerit always military time Sherry D
7:36 yeah I know if you're in the military I
7:38 was just an Air Force brat I was when I
7:41 was a kid I was I was like I'm either
7:43 want to be a pilot or a stunt man a
7:46 Hollywood stunt man I went the acting
7:47 route my dad went the pilot route I went
7:50 the acting route so you know and it was
7:52 funny he always talked about oh I wish
7:54 i' had have gone into
7:56 acting I'm like I wish I could learned
7:59 how to fly [ __ ] super cool Jets but
8:05 whatever I've got Sim racing now well I
8:08 used to have SIM racing now I've got a
8:09 Tik Tok Channel this this time slot in
8:12 my life used to be reserved for about 10
8:15 ft that way for my racing
8:19 rig all right so here's what we're going
8:21 to do I I don't know what we're going to
8:23 do we're we're going to talk about AI
8:24 [ __ ] that's what we do here it's your
8:26 gallery hopping hair yeah well it was
8:29 not on purpose purpose I just I don't
8:31 know I don't know what happened I don't
8:33 know what
8:36 happened you know whenever my hair looks
8:38 this shitty it my fallback position is
8:41 well at least you got
8:45 hair you'd think if I had hair I would
8:48 take some [ __ ] pride in it you know
8:51 but no no we've got I like this isn't
8:54 even like '90s so I I sort of oscillate
8:57 between 70s Carpet Sales man and and
9:01 like you know '90s
9:04 agency hipster and then this is like
9:07 some
9:08 1980s Flock of Seagulls meets Mad Men
9:13 it's just
9:16 bad co-pilot Pro came out oh yeah
9:19 co-pilot Pro came out that's interesting
9:23 so co-pilot Pros 20 bucks a month um
9:27 this is um being able to have
9:30 essentially gp4 inside Microsoft Word
9:33 Excel all the Office 365 suite for
9:36 individuals Enterprises had access to it
9:40 previously but now they've done General
9:42 release I think Enterprises it was 30
9:44 bucks per seat per person I think
9:47 individuals is 20 bucks per seat per
9:49 person so it seems like all these
9:51 companies are kind of landed on 20 bucks
9:53 a month as the uh as the price so that's
9:56 what I'm guessing um Gemini Ultra is
9:59 going to be I think they're going to
10:00 call it
10:01 Bard Bard Advanced is the is is what I
10:05 remember reading 30 minute Hells indeed
10:08 this is the time this is the time we do
10:10 the 30 minute hellos we'll talk a little
10:12 bit of uh we'll talk a little bit of of
10:15 AI stuff in the hellos here um I also
10:19 saw an interesting thing
10:22 that I didn't believe it was real there
10:24 was a sort of selfie video of Mark
10:27 Zuckerberg talking about um llama
10:31 3 and how many h100s they have and how
10:34 many h100 equivalents they have it's
10:36 pretty insane what meta is doing if this
10:38 is
10:39 true so we can go look because it may
10:42 not be true but
10:45 um they apparently have 300,000 h100
10:49 gpus those things are like 20 grand a
10:51 piece or 22 Grand a piece or 30 grand I
10:53 don't know they're a lot of money so
10:55 they got 300,000 of them and then if you
10:58 combine all their other gpus that are
11:01 not h100s they have sort of the
11:03 equivalent of 300,000 more h100s just in
11:07 raw compute power so they have
11:11 600,000 h100 GPU equivalents that they
11:16 are training llama 3
11:20 on that it's [ __ ]
11:23 insane like perplexity that just raised
11:26 a billion
11:28 dollars they
11:30 um they bought 22,000
11:34 gpus for their server farm and meta
11:38 throw in 600,000 at this thing and then
11:43 what they what Zuckerberg said is
11:45 they're going to open source llama 3 one
11:47 of my thoughts was that they
11:49 wouldn't who's using going to use a
11:51 galaxy galaxy AI on the s24 that's
11:54 another thing we could talk about is
11:55 that is the uh the Samsung keyote note
11:59 where they they announced that new
12:01 Galaxy um I think people will use
12:04 it I mean anyone who buys it the the uh
12:07 they did this cool thing called Circle
12:09 to search um with AI so at anything on
12:14 your screen you can scribble over it or
12:16 draw a circle around it and it'll do an
12:18 immediate search on
12:20 it what's the best way to get into AI as
12:23 a
12:24 Dev
12:28 h well I I I would say it depend
12:33 like a Dev doing what like the the it's
12:37 that's such a it's such a broad question
12:39 it's a good it's a good question do you
12:42 want to work for a research lab doing
12:44 sort of the hardcore math do you want to
12:47 develop
12:49 applications um that are Standalone do
12:52 you want to do kind of front-end Dev and
12:57 um you know just just just be doing
13:00 basic calls
13:03 like I would think a a really good place
13:06 to cut your teeth is is start you know
13:09 setting up your own open-source servers
13:12 right set up a Google um
13:17 collab instance with gpus on
13:21 it and just go nuts on the open source
13:23 world just start learning all the
13:24 different open source models I think I
13:27 think as a Dev
13:29 the more educated you can be like you
13:31 should understand what Lang chain is you
13:35 should understand um some of the agent
13:39 softwares uh like like agent GPT and
13:42 auto
13:43 GPT
13:46 um you should understand you know
13:48 semantic search versus embedding and how
13:51 you and and Rags retrieval augmented
13:55 generation if you're not ragging you're
13:57 not developing these days
13:59 you should follow David Shapiro on
14:03 YouTube backend stuff for Google Etc
14:06 yeah I so if you want to do backend
14:07 stuff I would you should you should be
14:09 you should be go going deep into
14:12 autonomous agents Lang chain Rag and um
14:17 get into some of the open source
14:19 projects um David Shapiro's got two that
14:23 I think are are really interesting one
14:26 is a cognitive architecture for AGI
14:28 total open source you can join that
14:30 project and contribute to it um and then
14:33 the other one is an agent
14:37 spawning some sort of agent [ __ ]
14:39 thing when when chat GPT came out with
14:42 the assistant API um building a whole
14:44 bunch of cool [ __ ] on that so I I just
14:47 go start building go get educated as
14:50 [ __ ] on the open source community and
14:52 build build build understand the open AI
14:56 apis understand Google's apis because
14:59 Google's apis are really interesting
15:01 right now and they're going to get
15:02 really interesting when when Ultra comes
15:04 out Google has a free course MIT has a
15:08 free course Stanford has a free course
15:11 meta has a free course I think meta has
15:14 one
15:16 um but I I
15:20 wouldn't I wouldn't worry too much about
15:23 courses if you're already a developer
15:25 like I just just start building [ __ ]
15:28 just go
15:29 just go to hugging face and GitHub and
15:32 look at the trending projects
15:34 everything's AI every project on hugging
15:36 face every project on GitHub that are
15:39 trending it's all AI projects just go
15:42 [ __ ] pick three and just go dive into
15:45 them that's what I do anyway all right
15:48 we haven't even started yet what are we
15:50 going to talk about
15:52 people what are we going to talk about
15:54 lovely image here from Danielle very
15:56 nice little little Japanese zenar
16:00 Garden or you can buy an AI
16:03 certification for
16:05 $997 you can make
16:07 money you can make money if you sell
16:10 $997 certifications right now although I
16:14 got to say if you're on the technical
16:15 side of things some of those
16:16 certifications you know if you're if
16:17 you're getting into machine learning and
16:19 Ma mathematics and things like that
16:21 that's that's enough enough established
16:22 that there are valid courses and
16:25 certifications and workshops on that
16:27 side if you're on the business end of
16:29 chat gptt where you're not so much a
16:32 developer that's where it's a lot
16:35 squishier um but but even still even
16:37 even that backend stuff is moving so
16:40 fast because what's happening what's
16:42 happening in the open source Community
16:44 is you
16:46 know um uh open Ai and um anthropic and
16:53 Google they're all kind of leapfrogging
16:55 each other these big tech companies are
16:56 doing what they can to you know one up
16:59 one another and then the open source
17:03 Community is sitting back there taking
17:05 you know the Llama 2 model and they're
17:06 they're sort of mixing and matching all
17:08 these different approaches that they're
17:10 seeing all these different companies do
17:12 like mixture of experts right that was
17:14 the whole thing came out on on chat gp4
17:17 using mixture of experts model and so
17:21 now mixl came out with that and so the
17:23 open source Community is just going nuts
17:25 on all these different models and so I
17:28 think that's if you're developer I think
17:30 that's the place to be
17:31 because open- Source projects are going
17:34 to start leapfrogging commercial
17:36 entities and it's going to get
17:38 wild wait there's not a plan for
17:40 tonight's class there's never a plan I
17:43 never
17:45 plan I do not read manuals my son Zach
17:48 does it drives me [ __ ]
17:51 baddy Dad we got to read the manual it's
17:54 a shovel do you know what that little
17:57 Notch is for no and I don't
18:00 care it's usually something cool the
18:03 shitty discovers by reading the manual
18:05 I'm not saying it's not a it's not a
18:08 positive impact on my life it's just
18:11 come
18:12 on you you just you just you rip open
18:16 the box you do you do not open it
18:19 carefully so that you can put it back
18:20 together and return it no if you bought
18:24 the thing you're into it you rip the Box
18:26 open you take the thing out you you
18:29 break it because you didn't read the
18:31 manual and then you take it back to the
18:33 store you get in an argument with the
18:35 people that they sold you a defective
18:37 thing and then the guy behind the
18:38 counter goes you didn't read the manual
18:40 did you and you're like no no I read the
18:42 manual this was defect it was broken
18:44 like that when I got it and you get in a
18:46 fight with them and then you take your
18:48 broken thing back and try to fix it this
18:50 is how this is how you live life people
18:53 if you read the manual you miss that
18:55 human
18:57 interaction
19:01 you're welcome the these are the kind of
19:03 life lessons that other channels just
19:05 don't give
19:10 you we keep our pinkies out cuz we're
19:15 classy can you touch on what hugging
19:17 face is
19:19 barely so hugging face okay yes I
19:26 can there are kind of two main
19:28 repositories right now of Codi kind of
19:31 stuff there's GitHub which is a big code
19:34 repository so developers if they're
19:36 developing stuff put that there hugging
19:39 face was set up as a model repository so
19:44 um with stable diffusion for example
19:46 which is an image generation tool
19:47 there's all sorts of different
19:49 fine-tuned models that people are
19:50 building on top of the open source model
19:54 that stability AI put out and so they
19:56 save them as these new checkpoints and
19:58 models and they upload them to hugging
20:00 face so it started out as just a
20:01 directory of AI models that you could
20:05 use for different tools it's since sort
20:08 of grown lots of tentacles they've
20:10 raised a lot of money there's now
20:12 runtime environments there where you can
20:14 play with models so it's not just about
20:16 downloading them now you can play with
20:18 them
20:19 there's I think there's code
20:21 repositories there it it both GitHub and
20:24 hugging face if you are a
20:27 non-developer are the shittiest
20:30 interfaces on the on in in in the
20:33 history of the internet they're just
20:35 there's just links everywhere it's just
20:38 like a sea of hyperlinks and if you
20:39 don't know which one to click on well
20:41 that's just cuz you're a clueless Noob
20:44 um and that's their attitude so um the
20:48 one cool thing about hugging face though
20:50 which is kind of fun so if you go to
20:52 hugging face as a non-developer if
20:54 you're a developer you're on your own go
20:56 go figure that [ __ ] out um let me lower
21:00 my brightness there let's go to hugging
21:04 face so it's hugging face.
21:08 Co and then when you first get
21:12 here um here's models and that's like
21:15 super geeky just you know models and
21:20 like the h94 I adapter dface ID you know
21:24 I I'll tell you what when I was playing
21:26 with the h94 D P adapter dface ID it was
21:31 the funniest
21:33 thing no one's saying that okay so
21:36 there's models and then there's data
21:37 sets so chat GPT prompts telat right
21:41 then in the middle this thing with color
21:44 these These are called spaces and so
21:46 what the spaces
21:48 are are places where you can go play
21:51 with models that that developers have
21:53 set up where you don't actually have to
21:54 install anything a lot of times they're
21:57 kind of they're over run with people
21:59 using them and they're slow and you
22:00 can't do it but let's see segment
22:03 anything
22:05 web and basically the way they're
22:07 ordered is whatever are the the ones
22:10 that show up at the top are kind of the
22:11 most popular ones that people are
22:13 playing with right now so this is
22:16 segment anything so we can upload an
22:18 image
22:22 here if you haven't seen segment
22:24 anything it's pretty
22:27 slick yeah this this will be a good
22:30 one and I I haven't played with this
22:32 before by the
22:33 way so reset image extracting image
22:38 embedding left click positive points
22:41 right click negative
22:43 points oh yeah okay so
22:46 see
22:49 wait so what segment anything does this
22:52 is a this is a free model that meta put
22:54 out is it automatically goes into this
22:57 image and it identif
22:59 ifies different objects within the
23:02 image so like there's the floor there's
23:05 the main subject there's this
23:07 dude there's that dude's shirt there's
23:11 that dude's
23:13 pants let's see there's her dress
23:16 there's her shirt that's that lady
23:19 that's that dude there's his face
23:22 there's his
23:25 hat look each of these letters is
23:27 segmented now so so you're like yeah but
23:30 Kyle what is this so this is just
23:32 basically a thing that meta put
23:34 out that every every one of these things
23:37 that was highlighted a developer would
23:39 be able to identify things within this
23:41 so if you if you see those crazy um you
23:45 know parking lot things that identify
23:47 you know here's cars and here's parking
23:49 spaces that have come available it's
23:51 it's using this kind of Technology but
23:52 this is just a super
23:54 easy a super easy thing that meta put
23:57 out for free so anyway um so that's what
24:00 hugging face is you can go find models
24:02 there if you want to geek out it's a
24:03 place where you can go you can run deep
24:06 wi with all of this stuff a good place
24:08 to start is is with Reddit so go to
24:11 Reddit and then go to subreddits for
24:13 things like stable diffusion or open-
24:16 Source AI or hugging face or right
24:19 within subreddit there's all these
24:20 different communities that you can play
24:21 with all right Kyle which is the best
24:24 image generator you recommend good
24:26 question um what's your name Z D um Let
24:30 me let me go show you something else
24:32 first and then and then I'm going to get
24:36 you there so so hang with me so the
24:39 first place I want you to go is there
24:41 the salon. a okay the salon. a don't use
24:46 https it doesn't work because I set up
24:49 the domain as a lazy
24:51 man and I didn't I didn't check some
24:54 stupid [ __ ] box that said use https
24:58 but where's the manual for AI the salon.
25:00 a yeah there you go so that's where I'm
25:02 taking you the salon. A and then what
25:05 that's going to take you
25:07 to
25:11 is the salong and how you pronounce it
25:14 is the
25:17 Fon
25:18 a it's not Salon Salon is where you go
25:22 get a a lovely hairo like mine but the
25:26 phon is where you go uh get your [ __ ]
25:29 together with AI okay so when you come
25:31 here it's a link tree um the first link
25:33 is to the community that's where we're
25:35 going to go that's we're going to talk
25:36 about image generation recommendations
25:39 um there's also a couple of links here
25:40 for the GPT for good event directory in
25:43 the YouTube playlist the GPT for good
25:45 event um was three weekends ago Ann
25:49 Murphy and I um and Vicky Baptist put
25:52 together this
25:54 event that we our goal was to have
25:57 people build 100 gpts in a 24-hour
26:00 period for nonprofits this community got
26:03 together and kicked ass and built more
26:05 than 150 so that's that's what these are
26:07 this are the directory that Vicky built
26:10 um these are the YouTube playlist of the
26:11 sessions where Anna and I taught and
26:13 then there's also a uh What's this
26:15 called merch store uh that supports the
26:17 salon so the salon doesn't cost you
26:19 anything the salon community so go to
26:21 the go to the salon
26:23 community and sign in join if you're if
26:25 you haven't joined and accept your
26:27 cookies accept your cookies people we've
26:30 got to track you with your cookies we're
26:32 going to grab your cookies we've dropped
26:34 cookie crumb crumbles on your just just
26:38 do your cookie just accept all just
26:40 don't read the terms of service we
26:43 everybody owns your data it's gone it's
26:46 gone you gave it up 15 years ago all
26:48 right so we're going to log in
26:51 here
26:54 and I am quite a fond I I'm quite fond
26:57 of this Comm community so this community
26:59 is growing quickly the first thing I'd
27:00 like you to do when you come here is go
27:02 to welcome to the salon you can ignore
27:04 that dude I'll show you what that's
27:06 about in a second that's some lovingly
27:08 troll love the construction paper and
27:10 the cool cutout for the microphone
27:12 exactly that's my microphone cut out and
27:14 here's my trackpad so I can scale things
27:16 up and down look this is this is some
27:19 serious this is some serious
27:21 construction
27:27 paper
27:29 nothing I'm telling nothing but the best
27:31 production values for 2024 for the AI
27:34 salon and look not for
27:37 nothing for you people look how
27:41 much black construction paper I ordered
27:45 just so that I could do that one thing
27:47 look at that look at that what am I
27:49 going to do with
27:52 this it is this is this the burden this
27:56 this the burden of doing a worldclass
27:59 video production
28:01 Studio you got to deal with pounds and
28:03 pounds of black construction paper okay
28:07 so when you get to the salon go to
28:09 welcome to the
28:11 salon and read read up on who we are and
28:14 what we're about label it and put it on
28:16 the merch store this is a high class a
28:20 high class Salon if ever there was one
28:22 exactly exactly people um and then the
28:26 other thing that I'd like you to do is
28:27 please please please um check out our
28:30 values values what is this some kind of
28:33 successories poster you gen xer no
28:36 values are actually really important
28:39 because as this community grows like
28:41 right now this community is quite
28:43 remarkable and as it grows I want it to
28:45 stay remarkable how do we do that well
28:48 we actually understand who are the
28:49 people that are here and we talk about
28:51 it so like we want people here that are
28:54 curious and are actively seeking answers
28:56 to things not cynical people people that
28:58 are like it's okay it no we want people
29:00 that are curious we want people that are
29:02 generous generosity is one of the values
29:05 how do we how do we learn well people
29:08 within the community they learn [ __ ] and
29:11 then because they're generous they share
29:12 what they learned and this is happening
29:15 right now in this
29:16 community there's not information
29:18 hoarding going on maybe a little bit you
29:21 know people got people got to keep their
29:23 Competitive Edge but for the most
29:25 part they're out there going here's what
29:27 I learned
29:28 here's what I couldn't figure out here's
29:30 where this thing sucks and then someone
29:31 else goes oh I figured out this piece of
29:33 it and then someone else goes oh I
29:34 figured out this piece of
29:37 it and then they they you know explore
29:41 try new tools together so that's
29:44 exploration is a value then there's
29:46 collaboration hey what if we work
29:47 together hey you want to put something
29:49 together that's happening a lot in here
29:52 inclusion because everyone's coming from
29:55 everywhere no everyone's being affected
29:56 by this AI stuff so we don't know where
29:58 people are going to come from so we
29:59 better be open to that and then empathy
30:02 understanding that right now people are
30:05 scared of AI they're angry about AI
30:09 largely because they don't understand it
30:12 there's a tremendous amount of
30:14 misunderstanding and misinformation
30:16 about AI good morning from Greece
30:19 Angelique welcome welcome
30:22 welcome that's
30:25 fantastic so empathy is really important
30:28 understanding that that people are going
30:30 through [ __ ] and if we can if we can
30:34 understand that and have some empathy
30:35 for where they
30:39 are then we can say yeah yeah this stuff
30:42 is scary I can understand why you're
30:43 angry want to see something
30:46 cool and then you show them something
30:48 cool and if you do it right they have
30:51 their first Kevin mallister moment which
30:53 is this
30:55 moment if you remember Home Alone
31:01 generative AI when you get into it is a
31:04 series of rolling Kevin mallister
31:06 moments where you just get your mind
31:09 blown like one thing after another and
31:12 you think you've seen it all and then
31:13 you go look at one of these image
31:15 generation tools and then you think
31:16 you've seen that all you're like oh I've
31:17 got mid Journey mastered and then you go
31:19 over to Korea or Deco here and it starts
31:22 rendering images in real time as you're
31:24 typing your prompt you're like the hell
31:26 is going on here
31:35 so
31:36 anyway that's what you do when you first
31:38 get to the salon is read about who we
31:41 are and what we're up to and if that
31:44 resonates with you then this is likely
31:46 your people and you can kind of think of
31:48 the AI Salon as as a support group for
31:51 people that are curious about AI one of
31:53 the things I'm hearing a lot lately is
31:55 oh my God is this ser seriously a place
31:58 I can talk about AI cuz every time I
32:00 talk about AI with my friends they tell
32:01 me to shut up and I'm a weirdo and they
32:03 don't like me anymore is this can I talk
32:05 about it here is it going to be
32:09 okay and we welcome them into the circle
32:12 Point them to the coffee machine and
32:15 tell him to smoke a cigarette and start
32:17 making images of
32:20 do it's a support
32:22 group so so come in introduce yourself
32:25 so that's the next thing to do so first
32:27 thing is check out what we're about the
32:28 second thing to do is in the community
32:30 section the first little tab here is
32:33 introduce yourself so my request my
32:36 strong strong strong
32:38 request is to go introduce yourself wait
32:41 there's a coffee
32:43 machine um and be generous don't just
32:46 say hey I'm here I'm here eat it losers
32:50 come in and tell us what you do tell us
32:53 what your background is tell us um where
32:57 you are on the AI Adventure I'm just
33:00 getting started I've been really into it
33:02 for a year but nobody knows it can I
33:04 talk about it here without being
33:07 humiliated what wherever you are I've
33:09 been in it for 30 years but this
33:11 generative AI stuff's different so I'm
33:13 trying to figure out people that are
33:14 figuring this out better than I
33:20 am and then another fun area and I am
33:24 going to get to the person that asked
33:27 what's an image tool recommendation
33:29 here's one of the things about this
33:33 channel it's really [ __ ] annoying if
33:36 you just want the answer the nickname of
33:39 this channel well there's oh Teresa just
33:40 subscribed for the first time thank you
33:42 very much I really appreciate the
33:43 support makes a big difference
33:46 um one of the nicknames for the channel
33:49 one of the many nicknames for the
33:50 channel is chat add so so we're neuro
33:54 neuros spicy friendly in here and uh I'm
33:57 I'm a laminated card carrying member of
33:59 the ADHD club and so I'm all over the
34:02 [ __ ] place but I will eventually get
34:05 there and if
34:06 if if it if I was heading in an
34:09 important Direction and I get lost the
34:11 mods actually put me back on track
34:13 they're like hey Kyle don't forget to
34:14 talk about that thing you promised so
34:16 I'll eventually get there
34:19 but um the show inel channel in the AI
34:22 Salon is an absolute [ __ ] blast
34:24 because what what's what what the show
34:26 and tell Chanel has turned
34:28 into while these lives are on is like a
34:31 total [ __ ] jam session where if I say
34:34 something stupid people will make images
34:37 about it or if I say something that
34:39 deserves being trolled they'll make
34:41 stupid images about it or or people are
34:44 sharing gpts that they've made or
34:47 they're they're sharing songs that they
34:48 wrote onso or or they're sharing tips
34:51 and tricks and things like that um so
34:53 it's just a really fun place so it's oh
34:57 look here here's the AI Salon welcome to
34:59 the AI salon
35:01 right this is the salon not the
35:05 f
35:07 um this is bad hair day week thank you
35:11 Mr it it's it's uh it is it is
35:19 flattering you got to have thick skin if
35:21 you're going to if you're going to have
35:23 a Tik Tok Channel at
35:26 58
35:28 because this is the [ __ ] you're going to
35:30 deal with the mission starts at 1505 the
35:33 hell time is
35:37 that that's how to stump the
35:44 Yankees backstory to
35:49 follow oh Joker Joker J Joker steo
35:54 please no
35:56 backstory
35:57 oh that's a solid one look it's sort of
36:00 a what that's sort of what my hair is
36:02 looking like blandon's
36:04 debut champy singing Me guitaring and
36:08 the and the Fantastic ble of
36:11 BL that's beautiful a salon Late
36:16 Show oh man these are cool g so anyway
36:20 so go go put some some wacky ass stuff
36:23 in
36:24 there beautiful oh that's nice actually
36:27 like that
36:30 logo Cindy [ __ ] welcome Cindy [ __ ]
36:33 what's
36:36 happening I'm excited about our project
36:39 we have a Sydney [ __ ] and I are working
36:40 on a super cool project that she
36:42 initiated that is it's going to be so
36:45 cool going to be so cool people I'm
36:48 blind this channel has helped me a lot
36:50 in my AI Journey Jeffrey Lombardo thank
36:52 you so much that's awesome that's
36:54 awesome I'm I'm listen one of the things
36:55 I'm so excited about
36:57 um with AI is accessibility like on all
37:00 levels it's we were talking about this
37:03 it's funny on the in our conversation
37:05 today talking about the project Cindy
37:08 and I are working on accessibility was a
37:11 huge piece of the
37:12 conversation and it was everything
37:15 from you know foreign languages to to
37:18 you know if if someone's blind or if
37:20 they're deaf or if they're cognitively
37:22 impaired um there's all sorts of
37:24 research going on where people are now
37:27 you know kind of
37:29 thinking you know thinking things into
37:32 existence because researchers are
37:34 finding ways to take brain patterns and
37:36 turn those into computer actions um
37:38 using generative AI like like just the
37:41 accessibility stuff with the these tools
37:43 are just insane and like we we've barely
37:45 gotten started so so thank you for that
37:48 I'm I'm really excited I'm glad this
37:50 this helps and you know it's like it is
37:54 it is still kind of random so just hang
37:56 in there there I'll eventually get to
37:58 the
38:00 point okay so that's show and tell so
38:02 there's also a water cooler discussion
38:04 Channel there's a news channel I'm going
38:06 to come back to tools and resources um
38:08 there's a Shameless self-promotion
38:10 channel so if you want to if you've got
38:11 some event you want to do you can put it
38:13 there and then there's also guilds and
38:15 the guilds are really cool what the
38:16 guilds are within the salon are are sub
38:19 communities so Stacy and Matt Mr K um
38:22 from here on Tik Tok just started um an
38:25 education Guild there's a video Guild
38:27 that's they've got their first meeting
38:28 coming up um Healthcare Pate who works
38:31 at Google has got the AI mechanics Guild
38:34 um if you want to go geek out there's um
38:37 a business Guild writer Guild Art Guild
38:39 and then uh Peter Kaminsky has an AI 101
38:42 Guild um and today Peter from the AI 101
38:46 Guild and Leah from the Art Guild got
38:48 together and they put on a mid Journey
38:50 class and Peter's like I don't know I
38:54 don't know how many images he's made but
38:56 it's probably north of 100,000 using mid
38:59 Journey so this is a guy that's that's
39:00 in hey I need to verify that my monthly
39:03 subscription is going directly to champy
39:07 as originally agreed I'll get him in
39:09 here champy champy do you get all the
39:11 chicken treats you need buddy
39:14 huh and well and listen technically
39:18 guitar strings go to champy you because
39:21 I can't play a chord without hearing
39:23 that [ __ ] dog sing so so it's chicken
39:27 strips and it's guitar
39:29 strings and an occasional whiskey you
39:32 know you got to keep the dogs loose all
39:34 right I've gone from wiring boards and
39:37 key punch cards to learning AI what a
39:40 journey yeah I so so let's see
39:43 Michael both boor Borth is that right um
39:48 yeah so so it's funny my um is Peter's
39:52 Class A recorded one I assume they
39:54 recorded it Peter's pretty good about
39:55 that stuff so I would say go go into the
39:58 uh uh I think it was in the Art Guild
40:01 but go into art or or Peter's 101 Guild
40:04 and just
40:05 ask um talking about Punch Cards so when
40:09 when I when I first started college at
40:12 Penn State here I'll flip back over for
40:14 a second when I first started college at
40:17 Penn State I I knew I enjoyed acting but
40:21 I'm like I can't you know you can't go
40:23 into acting like so but I was good at
40:25 computers and I'm like I'll do computer
40:27 science and so what I knew of computers
40:30 back then was like the Apple 2 and and
40:34 uh the Mac was was you know rumored to
40:37 come out and I I learned to program on a
40:41 trash 80 you know personal computers so
40:44 I walk into my first computer compai 101
40:47 class at Penn State and I walk into this
40:50 room and it's full of like gray and
40:54 Military green IBM M 3780 mainframe
40:59 computers and this dweeby this dweeby
41:03 little nerd walks in with a stack of
41:06 Punch Cards like it was like up to his
41:08 chin he was like holding holding him
41:10 down stack of Punch Cards and he walks
41:12 in and he goes this is a computer
41:15 program and he goes he goes what what
41:18 you'll be doing is you'll be typing
41:20 lines of code onto these cards and you
41:22 run them into these machines and then
41:24 they generate reports over here and it
41:26 was the big wide green and white paper
41:29 and and you'll be running reports for
41:31 big corporations for the rest of your
41:33 life and I was like the hell I will and
41:36 I changed my major to theater the next
41:38 day I was like I am [ __ ] out if
41:42 that's your definition of a computer and
41:43 if that's my definition of my next 30 or
41:46 40 Years nope I'd rather be [ __ ]
41:49 broke so anyway Kyle likes to say I work
41:52 at Google but I also have a PhD in
41:54 compai exactly
41:59 um yeah P's got it going on so so Pat's
42:02 mechanics guilds really cool if you're
42:04 into the geeky stuff he can he can he
42:06 can get you up to speed and and help
42:09 explain some of that stuff especially
42:10 for you know if people are in batch
42:13 printing yeah exactly yeah yeah because
42:15 you had to you had to um you had to time
42:18 share the computers right and so if you
42:21 were a freshman you you had to come in
42:23 at 10:00 at night cuz the the Elder
42:27 students got the earlier slots yeah
42:30 exactly I'm really interested in seeing
42:32 these tools in the hands of the elderly
42:35 oh my God the elderly and and dementia I
42:38 lost my dad to Alzheimer's and it's like
42:40 I
42:42 just I I just the idea of you know
42:45 people that are sort of lonely don't
42:47 have someone to talk to that can talk to
42:48 a familiar voice that knows all the
42:51 stories that is infinitely patient that
42:54 is there
42:55 247 holy [ __ ] it's
42:59 amazing
43:00 um okay so back to this so check out the
43:04 guilds the guilds are super cool let me
43:06 go to AI tools and resources so I I am
43:09 going to get to image generation tools
43:12 and and what my recommendations are um
43:15 this list is a list that I personally
43:17 curate it's the the AIS Salon uh you
43:21 know top
43:22 recommendations where I start is so if
43:25 you haven't tried chat GPT if you're in
43:28 this Channel and you're just here cuz
43:30 you know my supermodel good looks and
43:32 and you know clearly I've got a
43:34 worldclass Italian hair
43:37 quaffer um if you're just here for the
43:40 for the eye candy that's fine but if you
43:43 actually want to learn something about
43:44 Ai and you have not played with chat GPT
43:48 if you're sitting on the
43:50 sidelines right now drop everything
43:55 seriously go to chat. open.com and start
43:58 playing while you're listening to me
44:00 ramble start playing with chat GPT even
44:02 chat GPT the free version of chat GPT is
44:05 still better than most of the other
44:07 large language models out there it's
44:08 incredibly capable it's very very fast
44:11 it's it's incredibly
44:13 capable um and then if you can afford
44:16 the 20 bucks a month I personally think
44:18 if you want to be in the AI game at
44:21 all 20 bucks a month for chat pt4 is the
44:25 bargain of the century
44:27 like the amount of stuff it does for 20
44:29 bucks a month is insane but you actually
44:31 have to learn what it is so this list
44:34 not only tells you you know that it's
44:36 there but it says here's all the
44:38 functions within it that you should be
44:40 learning and you know I go over those
44:42 you you know come back to this channel
44:45 regularly because I go over this stuff
44:47 all the time so start with chat
44:49 gbt then below that I've got eight
44:52 different large language models that I
44:54 think are worth playing with um you've
44:56 got
44:57 the big four that I think are the big
44:59 four right now Claude Bard perplexity
45:02 and Microsoft co-pilot which that's
45:04 basically chat GPT 4 for free um
45:09 perplexity is uh really interesting they
45:12 just raised $75 million from andron
45:15 Horowitz B is Google and that's about to
45:18 get a real kick in the ass quality-wise
45:21 when Gemini Ultra comes out and Claude
45:24 is another very well-funded
45:26 uh direct competitor to chat GPT uh a
45:29 bunch of X chat GPT people the people
45:31 that created
45:32 gpt3 went off and created anthropic
45:36 which makes the Claud and then the
45:38 specialty llms you got Pi grock llama 2
45:41 and po all worth exploring okay so just
45:45 chat GPT if you haven't played with this
45:48 stuff before you could spend the next
45:49 month or two just learning what all
45:52 these things
45:53 are and what I would recommend is just
45:55 have fun with with them don't take it
45:57 seriously just go have fun make it do
45:59 stupid [ __ ] do what what we call in here
46:01 parlor tricks make it write poems and
46:04 songs and tables comparing ridiculous
46:07 items and create a meal plan that you
46:10 know features Gummy Bears I like just
46:14 [ __ ] throw [ __ ] at it and see what it
46:16 does and you'll learn a lot survey is up
46:18 okay we're going to take a survey now
46:20 hang on here people all right we got
46:22 four
46:24 minutes so we're asking what gen
46:26 generation are you so so the the the
46:31 survey click the the little it looks
46:34 like a little blue and white checkbox
46:36 over in that
46:38 corner and you got four minutes and so
46:41 we want to know are you a boom wait we
46:44 got Gen X in here twice I assume you
46:46 wanted gen
46:48 Z so the D says Gen X but D should be
46:52 gen
46:53 Z I don't know if you can edit if you
46:56 can edit that
46:58 but Boomers gen xers gen yers gen zers
47:02 or I don't
47:04 knows all right so just understand like
47:07 if you're in this channel you should be
47:08 smart enough to understand that B and D
47:11 are not the same B is Gen X D is Gen Z
47:16 okay glad we cover covered that up
47:18 definitely
47:20 Boomer start over uh that's up to you
47:22 guys I don't care I could end it you
47:24 want me to end it and start over all
47:26 right fine I'll do that hang on I'm
47:28 going to end this ending
47:32 it all right there you go so whoever is
47:37 doing the uh doing the surveys do it
47:39 again just make sure gen Z is right I
47:41 think you want a gen Z for
47:43 D um okay so let's keep going when that
47:47 comes back up I'll pull it up so so
47:50 you've got chat gbt then you've got
47:52 these eight large language models below
47:54 it then then you've got image generat so
47:56 here's the deal with image
47:59 generation these things are different
48:01 enough right now that the the answer to
48:04 which one's best is what are you trying
48:06 to do and where do you work and and what
48:11 are you looking for because they're very
48:13 very different um dolly3 is the image
48:16 generation tool that lives inside chat
48:18 GPT and inside Microsoft co-pilot so the
48:21 nice thing about Microsoft co-pilot is
48:24 that you get Dolly 3 for free
48:27 um if you're in chat GPT you need to
48:30 have the $20 a month subscription to get
48:33 Dolly 3 but I would argue that dolly3
48:35 inside chat GPT is a significantly
48:38 better tool than what's inside Microsoft
48:41 co-pilot just because of a lot of
48:43 reasons um the nice thing about Dolly 3
48:48 it's good with text it's not perfect but
48:51 it can do text so you see a lot of those
48:53 images in show and tell that have like
48:55 hey Kyle what's what's going on or
48:56 champy this or whatever that's that's
48:59 dolly3 for the most part mid journey is
49:01 getting better V6 is better at text but
49:04 it's still not great but mid journey is
49:07 if you want high quality images mid mid
49:10 journey is still the gold gold standard
49:13 okay Korea and Deco here are both
49:17 leaning into there's there's a new thing
49:19 that came out from stable diffusion
49:20 called sdxl turbo which is instantly
49:24 rendering images while you type prompts
49:27 and I I can show
49:29 those Korea and Deco here are almost
49:33 more
49:35 like like
49:38 jazz you know that then like like let's
49:41 how do we do this
49:43 so all
49:45 right okay I'm trying to hang hang on
49:49 okay okay mid journey is like speed
49:53 metal you got to you got to be really
49:55 technically good like uh like uh pfia
49:59 kind of you know stuff that's mid
50:00 Journey
50:02 um um doll3 is like Americana like
50:05 simple you know good songs good just
50:08 singal songs good
50:10 storytelling um Korea and Deco here are
50:13 more like Fusion Jazz they're like jazz
50:17 they're they're you you know wacky time
50:20 signatures and and things like that
50:22 Leonardo actually so so Leonardo is
50:25 built on top of a bunch of stable
50:26 diffusion tools it can now do realtime
50:30 image generation it can do traditional
50:32 stable diffusion stuff you can train
50:35 models within
50:37 Leonardo um they give you 150 credits a
50:42 day which is it it works out to about 10
50:46 images a day it's not as much as it
50:47 seems but it's still you can do you can
50:50 use it for free um and it's also doing
50:53 video now so uh St stable video
50:56 diffusion so Leonardo is a really
50:58 interesting tool cuz there's a lot there
51:00 and then you've got Adobe Firefly and
51:03 Adobe Firefly is
51:09 um hasn't been trained on any
51:11 copywritten
51:12 images so if you work within a
51:16 corporation um you can use images that
51:19 are generated in Firefly and Photoshop
51:22 generative fill and all that
51:24 commercially because there's not there's
51:26 not as big a risk as you have with
51:28 something like mid Journey which clearly
51:31 has been trained on on you know artists
51:33 images because somebody leaked the big
51:35 long [ __ ] list of artists that they
51:38 trained that they trained mid Journey on
51:41 so so so you know if if if you've got
51:46 you know ethical issues with how those
51:47 other things were trained then Firefly
51:49 is really good or if you work for a
51:50 corporation that's super Conservative
51:53 then it's really good now is it as good
51:55 good
51:57 no why cuz it wasn't trained on as much
52:02 art so everything's a balance right now
52:06 now if you're like if you're like but
52:08 Kyle that's too much can you just tell
52:10 me which one's the best did you see my
52:13 news post about Adobe and mighty
52:14 networks I did and it's um what did they
52:18 I I forget what it was it's hang on I
52:20 can go see it wait what's that hey I
52:22 learning
52:24 lab
52:30 um hang on
52:32 Danielle yeah there was some news about
52:35 some I think Firefly became commercially
52:37 available and it's free is isn't that
52:40 right
52:52 Danielle oh mighty networks is being
52:54 weird it keeps resetting anyway all
52:56 right whatever I'll figure it
52:58 out monetize images free Firefly that's
53:01 what it was yeah so so go check out
53:03 there there's some some Adobe
53:06 announcements yesterday but one of them
53:07 is Firefly is now free which is great it
53:10 didn't used to be you had to have a a a
53:12 Creative Cloud
53:13 subscription so there you go some of
53:15 these you know some of the prices of
53:17 these things are trending toward free oh
53:19 that's actually a good point okay so
53:21 Dolly 3 um is free
53:23 within Microsoft cop pilot it's it's
53:27 included within the 20 bucks a month for
53:29 chat BT mid Journey cost you money I
53:32 think the minimum is 10 bucks a month
53:34 Korea is absolutely free which is pretty
53:37 amazing because it's a powerful tool and
53:38 it's got more than their real time
53:40 engine it's got background removal and
53:42 it's got an upscaler and it's all free
53:45 Deco here is free now these things that
53:47 are free right now here's a little
53:50 software economics lesson for you the
53:53 reason they're free right now is they're
53:55 trying to get you to use their tool
53:56 instead of mid Journey or instead of
53:58 some other tool so they're they're free
54:00 initially and then basically what
54:02 happens is this when they burn through
54:04 their series a or series B
54:07 money then the series C people come in
54:10 and go yeah about how much money you're
54:12 blowing uh let's go on ahead and start
54:15 charging people for that so these are
54:17 not going to be free forever so so use
54:19 them while you can um okay beautiful so
54:23 we have a
54:24 new oh we got 31 votes okay so in here
54:28 tonight out of 31 votes we've got um 30%
54:34 are
54:35 boomers 50% are gen xers which makes
54:38 sense that's who I am 12% are gen yers
54:41 6% are Jen ear welcome jenz ear U 3%
54:45 don't know which is one person that's
54:48 okay um but but welcome that's pretty
54:51 cool that's that's kind of what I that's
54:52 about what I I would expect I would love
54:55 to see see some some more younger people
54:57 in here but you have an old [ __ ] on you
55:00 know leading the channel so you know I
55:03 think young people are like why is
55:04 grandpa on Tick Tock flick flick
55:08 flick but you know we have people in
55:11 here like Rudy from South Africa he's a
55:14 young a young sprightly fell he's in
55:16 here regularly so if if you know some
55:19 youngans in your life tell them to come
55:21 in here tell them this is an old guy
55:23 he's not bad he's got decent
55:33 energy in fact do me a favor now we got
55:36 139 people in here do me a favor right
55:40 now share this live either with Gen Y or
55:43 Jen zers so some young people in your
55:46 life especially if you know it's going
55:48 to piss them off because I've got
55:49 23-year-old twin boys upstairs they hate
55:52 this AI [ __ ] so so share this thing with
55:55 with uh with young people in your life
55:58 or with people in your life that you
56:00 know don't want to hear about AI anymore
56:02 that that's who we want to get in
56:07 here all right trained um aren't we all
56:12 trained on other people's Arts yeah so
56:14 Brendan you know BT comp
56:17 KSU what you're talking about is the
56:20 reality of of what's actually happening
56:22 the the the larger these training sets
56:27 get the less it makes sense
56:30 that every individual artist within it
56:33 got to be compensated because the you
56:36 know their art was used in the training
56:38 set there's a point at which these
56:40 training sets get so big that they've
56:43 been trained on all of the output of all
56:45 of the humanity and and and the the
56:48 large language models and the image
56:50 models and the music models and the
56:52 animation models and the video
56:54 models they all essentially
56:57 become like the global Library of
57:00 Congress right or like the
57:02 global Museum or the global library
57:06 right that we've all contributed to and
57:08 we all get to benefit from so there's
57:10 absolutely an argument that says
57:13 that 130 people and eight or so robots
57:17 which ones I'll never tell p P's good so
57:21 so I was in here the other night
57:22 bragging about how I can absolutely 100%
57:25 % guarantee no one will ever jailbreak
57:28 your GPT and by the time I finished that
57:30 sentence Pate had jailbroken my GPT
57:33 which was really awesome of him to do
57:35 I'm but I'm actually serious I'm I'm
57:37 glad he did it because I was not making
57:39 a promise what what I was actually
57:41 saying is assume that anything that you
57:43 put in your GPT will be had by the
57:46 public if they really want it and then
57:49 and then Pate made it so proved that
57:52 that was the case I made a Luma to show
57:55 what it's capable of but it was at work
57:59 and scared of HIPPA oh yeah yeah I know
58:02 the listen the hip the Hippa stuff is
58:04 scary right now we've still
58:09 got I think we've probably still
58:14 got we we'll see what open Ai and and
58:17 Google do I I I I don't think Google's
58:20 going to address any of this [ __ ] um gp5
58:25 could address some of this stuff
58:27 where they will get more reliable and
58:30 they will start to have
58:32 reasoning
58:34 and God help us they have some better
58:37 privacy and security
58:39 so my my gut is telling me that we have
58:45 as as like this Collective little rag
58:47 tag bunch of people just trying to
58:49 figure this [ __ ] out probably 18 months
58:53 before the tools themselves are stable
58:56 enough and secure enough that big
58:58 companies really lean into
59:01 them healthc care companies are probably
59:03 going to be 18 months after that right
59:05 so so so we're probably three years out
59:10 from healthcare companies really leaning
59:11 into this generative AI stuff they might
59:13 lean into it on the business process
59:15 side but anything on the patient side is
59:17 going to be
59:18 slower but what that means for all of us
59:21 is that the opportunity for all of us if
59:24 you just get all you need to do in the
59:25 next 18 months is simply get AI literate
59:29 and what that means is be curious
59:33 like honestly dig through this list that
59:36 I have on the AI Salon start with chat
59:39 GPT learn it play with it have fun with
59:41 it go make some you know some uh you
59:44 know figure out what are the differences
59:46 between like perplexity is really good
59:48 at research and Pi's really good at
59:51 conversational stuff and grock from X is
59:55 not good for [ __ ] um and how do you
59:57 learn that well you just go play with it
59:59 and you go figure it out and then and
1:00:01 then start playing with the image
1:00:03 Generation stuff and you'll figure out
1:00:04 what this is good at and you'll have a
1:00:05 lot of fun and you'll get inspired and
1:00:07 you'll realize oh I've always wanted to
1:00:09 be able to be an artist and now I can be
1:00:11 or whatever it
1:00:13 is and just play and then do the same
1:00:15 with the video generation tools Runway
1:00:17 ML and pabs or the two two biggies for
1:00:20 you know generating content and then hey
1:00:23 Jen and D D- ID are good for Avatar
1:00:27 creation and then visa and I should
1:00:30 probably add what was that one that that
1:00:32 we put up the other day
1:00:37 um another video tool I don't know but
1:00:39 visel is good for now um and then
1:00:41 there's music and audio tools and then
1:00:43 there's 3D worlds and objects and then
1:00:45 there's people to
1:00:46 follow and and the the four people that
1:00:49 I have here are there more than four to
1:00:51 follow [ __ ] yes there's a ton more to
1:00:53 follow but these These are like solid
1:00:57 like will deliver the goods week after
1:00:59 week kind of people all
1:01:02 right if you're not a robot click on all
1:01:05 the boxes that contain stop lights
1:01:07 that's good steo but anyway like just
1:01:12 just being in the conversation in the
1:01:14 next 18 months and getting curious about
1:01:16 this stuff and kind of understanding
1:01:18 what these tools make possible is going
1:01:21 to yield you opportunities you cannot
1:01:24 imagine right now and I can't imagine
1:01:26 right now today I had my first live on
1:01:29 Tik Tok Mr
1:01:30 it how do you do it Kyle I understand
1:01:33 now
1:01:35 all yeah so here's the deal Mr
1:01:40 it when I started doing these lives so
1:01:42 so so here I'll tell I'll tell you how I
1:01:45 did
1:01:47 it
1:01:50 um I like this image this this is this
1:01:53 is where we all are right now we're
1:01:56 we're we're in a we're in a little inner
1:01:58 tube uh with with our pet llama uh in
1:02:02 the middle of the North Atlantic Ocean
1:02:04 in the middle of a storm that's this is
1:02:07 where we all are right now um but we're
1:02:10 in there together that's what the AI
1:02:12 salon's about is about doing it with
1:02:13 other people that's what this channel is
1:02:15 about here's here's what I did Mr
1:02:20 it
1:02:23 is
1:02:25 I have I have a thing I can do in my
1:02:27 brain I suppose it's a
1:02:29 talent where I can just flip a switch
1:02:32 and then I'll just commit to something
1:02:34 and I kind of commit to it
1:02:36 blindly ah Winston good to see you thank
1:02:40 you for that llama I love that llama the
1:02:43 by the reason there are so many llamas
1:02:45 in all of these images is because for
1:02:49 the first for the early days of this
1:02:50 live every time I would say llama 2 or
1:02:53 llama one from meta Winstead would throw
1:02:56 up that dancing llama and so it became
1:02:58 one of the official animals of the uh of
1:03:00 the AI learning lab
1:03:03 um but the way I did this live was I I
1:03:07 said
1:03:08 well for the most part when I get home
1:03:11 after work I eat dinner watch a little
1:03:13 TV and then I go do Sim
1:03:15 racing and I was doing like an hour and
1:03:18 a half 2 hours of sim racing a night to
1:03:22 the point that I got um Sim RAC 's heel
1:03:25 which I don't know if you've ever had it
1:03:27 but it really sucks it's basically like
1:03:30 a super tight Achilles tendon that that
1:03:32 makes my heel hurt um and it like it's
1:03:35 still pretty bad and I haven't done it
1:03:36 in a while um and so I just said what if
1:03:42 I give up Sim racing and just do this do
1:03:46 these lives and so I said [ __ ] it I said
1:03:49 I'll just do it every
1:03:51 night and I'll tell you Mr it the
1:03:56 probably the first two
1:03:59 months I had regular anxiety and panic
1:04:02 attacks either before the live or during
1:04:06 the live or just and so I just said [ __ ]
1:04:10 it I'm going to do another one [ __ ] it
1:04:12 I'm going to do another one and
1:04:14 sometimes I would sit there for like 10
1:04:15 minutes the little red live button was
1:04:18 sitting there and my heart would be
1:04:20 racing and I'm like I don't know what to
1:04:22 say and I I try to be doing homework and
1:04:25 coming up with plans and then I just go
1:04:27 oh
1:04:32 [ __ ] and that's
1:04:34 just just do it like it doesn't [ __ ]
1:04:41 matter it's Tik Tock this stuff is
1:04:44 ephemeral I mean I'm saving these things
1:04:46 on YouTube but I'm I'm really saving
1:04:48 them on on YouTube
1:04:50 because I know that what's coming from a
1:04:53 technology perspective is is going to be
1:04:55 the ability to take a body of work like
1:04:58 2 to three hour nightly videos for 6
1:05:01 months or a year and do some really
1:05:03 interesting [ __ ] with that I don't know
1:05:04 what it is yet but but that I want to
1:05:07 play with where do you go to learn about
1:05:09 creating
1:05:10 gpts do
1:05:13 channels
1:05:16 I I don't know I do not know what you
1:05:19 mean what do you mean
1:05:21 channels you're an inspiration Kyle
1:05:24 thank you
1:05:29 I I was I was on a call today our
1:05:33 [ __ ] our story Vine app for this one
1:05:36 client is just it's just like Mal
1:05:40 behaving it's just it's being an
1:05:42 [ __ ] and so we're on the we're on the
1:05:46 phone today doing a tech troubleshooting
1:05:49 call with this lovely young woman from
1:05:53 Texas um who who was just super sweet
1:05:56 and trying to help us out and at one
1:05:57 point I said something and she said she
1:06:00 said um how did she say it she goes she
1:06:03 goes um she goes well I hope this helps
1:06:06 sir and I was like like don't call me
1:06:09 sir she was taking her back cuz she's
1:06:11 from Texas they're like that's how I was
1:06:13 raised I call people sir uh so yeah when
1:06:17 I hear him in Inspiration it hits me
1:06:19 kind of like that like sir
1:06:22 like custom G P PT resources in the
1:06:26 salon I'm going to send you more newbies
1:06:29 I talked about wait what was
1:06:31 that I don't know what I'd be doing
1:06:34 right now if you hadn't started doing
1:06:35 this so grateful for you oh thank you
1:06:37 very much Danielle I appreciate that um
1:06:39 so grateful you pushed on yeah listen I
1:06:41 mean here here's the deal here's the H
1:06:44 the honest to God truth of why I pushed
1:06:46 through that
1:06:49 is I've been in this movie before right
1:06:52 I started one of the early digital
1:06:54 agents in
1:06:56 1994 and I also started a group called
1:06:58 the worldwide web artist Consortium
1:07:00 which was a New York City based you know
1:07:02 anyone who's interested in the worldwide
1:07:04 web join this little group and that
1:07:06 little group turned into like a 2500
1:07:09 person you know Juggernaut that anyone
1:07:12 in the New York City they called it
1:07:14 silicon alley at the time
1:07:18 right silicon alley reporter that's me
1:07:22 and my co-founder of agency.com
1:07:26 um anyone who was
1:07:30 in at all interested in the web was a
1:07:33 part of this group in New York City and
1:07:35 we would meet every other week in the
1:07:36 Sony building on Fifth
1:07:39 Avenue we and Sony gave us this massive
1:07:43 it was a conference room that held 300
1:07:45 people and we would regularly fill it
1:07:48 the the table in the center of the
1:07:51 conference room sat 150 people and then
1:07:53 we had chairs around it we would fill
1:07:55 that room
1:07:57 regularly so I know the value of
1:08:00 community when technology is this early
1:08:03 in raw and so so that's what this was
1:08:06 was like I just I feel like I need to
1:08:09 get as many people as possible curious
1:08:12 about this [ __ ] not afraid of it or if
1:08:15 they're afraid of it just willing to go
1:08:16 [ __ ] it and play anyway and then I want
1:08:19 to find a place for them to coales and
1:08:21 we're in a different time now so we can
1:08:22 do that online so the AI Salon is now
1:08:25 this Global community and it's it's just
1:08:28 amazing and and so is this channel it's
1:08:29 Global hey dude that's Kyle these Tik
1:08:32 toks keep me going ah thank you very
1:08:34 much lunatick you're I mean you're
1:08:37 kicking butt I'm I'm really just super
1:08:39 impressed with what you're doing it's
1:08:41 it's just
1:08:42 amazing um I remember the first time I
1:08:46 was called Sir now about 25 years ago I
1:08:48 was looking around who what exactly I
1:08:52 definitely have sir you know sir
1:08:54 sideburns and a sir hairline and and I
1:08:58 this is sort of more like a ma'am haird
1:09:01 but you know
1:09:04 whatever it did not look this bad for
1:09:07 most of the day I will have you know
1:09:10 uh I do the same thing when someone
1:09:12 calls me ma'am I know it totally threw
1:09:15 me like and I turned beat red they were
1:09:17 they were both like oh my God you turned
1:09:19 red I was
1:09:22 like cuz I just it was just such a an
1:09:24 instinctive thing don't call me sir man
1:09:27 I felt
1:09:28 [Laughter]
1:09:31 bad oh
1:09:34 man
1:09:35 uh best tool for flowcharts and org
1:09:38 charts
1:09:41 um I don't know there's a bunch of
1:09:43 plugins for that I don't I don't I don't
1:09:46 know that world I I would think at this
1:09:48 point there's probably some dedicated
1:09:50 tools for that stuff that I've been
1:09:53 doing you know flowcharts and or charts
1:09:55 for a while that have now added some AI
1:09:57 stuff I just don't know them I I would
1:09:59 say go to um go to Mike diesel go to
1:10:03 futur
1:10:23 pedia.com
1:10:24 like I like I have yet to see something
1:10:27 inside a large language model that does
1:10:29 anything sophisticated with
1:10:32 flowcharts but I'm not looking for that
1:10:34 so there might be something out there I
1:10:35 don't know
1:10:40 about all right so what do we want to do
1:10:43 tonight
1:10:47 um so oh someone asked about making gpts
1:10:50 maybe we'll make a GPT
1:10:52 tonight I can could do a little bit of
1:10:55 uh it's already 10:00 damn oh I started
1:10:57 late that's why it's Friday night though
1:11:00 [ __ ] it let's
1:11:03 party get my get my iced coffee going
1:11:06 let's get some energy shall
1:11:09 we you can make money with
1:11:20 ch by the way this thing you can make
1:11:23 money
1:11:25 what is this
1:11:27 about I don't bring it out a lot and
1:11:29 it's it's kind of an inside joke at this
1:11:31 point but here's why here's why this
1:11:35 exists
1:11:39 um there's a lot of Tik Tok channels out
1:11:41 there and there's a lot of channels out
1:11:43 there there's a lot of this on there's
1:11:44 not so much on LinkedIn there's a fair
1:11:47 amount on Facebook there's a fair amount
1:11:49 there's a lot on YouTube and there's a
1:11:51 lot on Tik Tok and it's
1:11:59 idiots going you can make money with
1:12:02 ch and they lean into the screen like
1:12:05 this and they twist their [ __ ] heads
1:12:08 around and it just it grats in my soul
1:12:12 every time I hear it because it I think
1:12:14 it it is
1:12:16 doing everybody a
1:12:20 disservice how they're how they're
1:12:22 making their money is selling you
1:12:25 courses on how much money you can make
1:12:27 with chat
1:12:28 GPT they're not making money with chat
1:12:31 GPT now they likely use chat GPT to
1:12:34 write the shitty course that they're
1:12:36 selling
1:12:39 you but how are they making money
1:12:41 they're not making money with chat GPT
1:12:44 they're making money with the fact that
1:12:47 people want to get rich
1:12:50 quick so they've figured out a problem
1:12:53 to solve
1:12:54 and they found a new to tool to solve it
1:12:58 and they put together these slick
1:12:59 [ __ ] channels you make
1:13:03 money to sucker people
1:13:05 in and so it does a disservice to the
1:13:10 tool it does a disservice
1:13:13 to I one of the things that I think that
1:13:16 that attitude that tool first
1:13:19 attitude does is it scares people away
1:13:24 it makes generative AI seem like nft
1:13:28 Bros you can make money with nfts you
1:13:32 can make money with blockchain you can
1:13:34 make money with
1:13:41 ch
1:13:45 no you make money by solving
1:13:49 problems you make money by being
1:13:52 creative in how you help other people do
1:13:55 something and if you do that good enough
1:13:59 they pay you
1:14:01 money it's called
1:14:03 business there is not a [ __ ] shortcut
1:14:07 for it period end of
1:14:11 story just
1:14:13 not occasionally someone will trip onto
1:14:16 something and they'll be the first one
1:14:18 to generate a 100,000 t-shirt designs
1:14:22 and get that up on print of
1:14:25 and you know overflood the market for a
1:14:28 week and a half and take the screenshot
1:14:31 that they probably photoshopped anyway
1:14:33 and said I made $100,000 in one week
1:14:36 with ch
1:14:39 JP just give me $29.95 I'll Teach You
1:14:42 How per
1:14:50 week and so it makes it it makes it all
1:14:52 seem like a sleazy [ __ ] Ponzi and
1:14:54 it's not this
1:14:56 shit's these tools are like the most
1:14:58 profoundly powerful tools in the history
1:15:02 of
1:15:03 humanity they're deadly
1:15:05 serious like the the they're going to
1:15:08 change
1:15:09 everything
1:15:11 everything from science to medicine to
1:15:15 the nature of work to the nature of how
1:15:18 we value work to what you do to what's
1:15:21 possible to what you do like it's going
1:15:22 to do all of that
1:15:24 like gar [ __ ] teed and it's not going
1:15:27 away and it's not overhyped in fact in
1:15:30 one of those Corners is a video series I
1:15:32 did called underhyped where I've got
1:15:35 like 18 videos in there talking about
1:15:37 why this is a big deal and why I think
1:15:39 it's a bigger deal than people realize
1:15:42 so so the you can make money with Chad
1:15:44 GPT crowd as far as I'm concerned can go
1:15:46 suck an
1:15:48 egg I understand why they're doing it
1:15:51 you know trying to make a buck it's just
1:15:53 a sleazy way to make a
1:15:57 buck I don't even want to talk about
1:15:59 gpts that killed mine oh that sucks
1:16:02 Joker yeah because see there there the
1:16:06 so what we're what we're in the middle
1:16:08 of right now Joker I would hang in there
1:16:10 because because uh and I know you will
1:16:17 um open AI right now is trying to figure
1:16:20 out how to be be a consumer application
1:16:24 company they've been a research lab for
1:16:28 five six seven years I don't know when
1:16:30 they were founded but for a while they
1:16:33 did not expect Chad GPT to be a hit and
1:16:35 it was and so they're now dealing with
1:16:39 like being a consumer application
1:16:41 company and they're really [ __ ]
1:16:42 shitty at
1:16:44 it
1:16:48 um
1:16:50 and so so if you look at their terms of
1:16:53 service for making
1:16:55 gpts there were three areas that
1:17:00 they basically
1:17:02 said financial services and Healthcare
1:17:05 were two of
1:17:06 them right and then you know sex stuff
1:17:09 was the other one there might have been
1:17:10 one other but but basically they're
1:17:15 avoiding the Deep liability holes
1:17:18 because they don't they're they don't
1:17:20 they don't these tools are not
1:17:22 sophisticated enough right now for them
1:17:24 to put themselves on the liability hook
1:17:27 for financial service advice or a bad
1:17:31 diagnosis and and a you know a
1:17:33 malpractice lawsuit coming at them so
1:17:36 they're just not going to let you make
1:17:37 those
1:17:38 gpts they'll get their [ __ ] together
1:17:40 eventually so I'd hang in there but yeah
1:17:42 that's really frustrating yeah that
1:17:44 sucks I'm sorry about
1:17:46 that ah anyway let me look at some
1:17:49 comments here let's get let's get some
1:17:50 comments going is he is he ever going to
1:17:53 is is he ever going to read my comment
1:17:54 he just seems to be talking does this
1:17:56 man ever stop talking all right I've
1:17:59 already got a twin up good yeah just
1:18:01 keep just keep [ __ ] flooding them in
1:18:02 there just slightly change the
1:18:05 language good for you yes do that
1:18:10 um but Kyle I want to I haven't got
1:18:13 familiar with
1:18:14 gp4 please slow down with GPT 5 I know I
1:18:19 know I'm right there with you listen I
1:18:25 so I I I I am
1:18:27 still I would still consider myself in
1:18:30 the middle of a semi existential
1:18:34 crisis that that started on November 6th
1:18:38 of
1:18:41 2023 and it's funny November 30th
1:18:44 2022
1:18:46 is when I sort of lost my mind and
1:18:49 realized oh my God the world just
1:18:50 changed this is AI is you know
1:18:53 generative AI is like the next worldwide
1:18:55 web um and then almost a year later
1:19:00 November
1:19:01 6th 2023 open AI has their developer
1:19:05 conference where they launch gpts and
1:19:09 they also launch multimodal chat
1:19:12 GPT and the reason I'm in a bit of an
1:19:15 existential crisis
1:19:20 is um
1:19:24 chat gp4
1:19:27 is signific it's a significantly
1:19:30 different
1:19:32 tool than the chat GPT 4 that existed in
1:19:37 October like like like it
1:19:41 changed just into a completely different
1:19:44 thing CU it went true multimodal where
1:19:47 it can see images and it can make images
1:19:50 and it can understand talking and it can
1:19:53 talk talk and it can program and it can
1:19:56 and it can do all this together and then
1:19:58 I'm starting to look at tools like
1:20:00 Leonardo that started as a very simple
1:20:03 image generation tool and now it's an
1:20:05 image generation tool and it's this
1:20:07 real-time composer tool and it's a movie
1:20:09 making tool and and so I'm watching
1:20:13 every one of the individual little
1:20:14 simple tools that we've been playing
1:20:16 with are getting more and more and more
1:20:18 capabilities and they're also starting
1:20:20 to blur like this and then and then I
1:20:23 looked
1:20:24 at the the Google Gemini launch even
1:20:30 though it was you know people made fun
1:20:31 of that marketing video because they
1:20:33 they they accentuated the positives
1:20:37 Andor lied as as some people said it's a
1:20:40 marketing video so every [ __ ]
1:20:42 marketing video is a lie what do you
1:20:43 what do you hi you don't think they're
1:20:45 going to make it look as good as [ __ ]
1:20:47 possible but
1:20:49 whatever but if you look at what Gemini
1:20:52 Ultra is and what what's coming the one
1:20:54 thing that's really clear to me is it's
1:20:57 not what we
1:21:00 know generative AI is today it's a
1:21:04 different thing it's a more visual
1:21:07 thing
1:21:10 um yep it's everywhere people buy buy it
1:21:14 chat GPT prompts for 25 bucks or 47
1:21:16 bucks and a million dollars exactly yeah
1:21:19 exactly and and here's the thing Marina
1:21:22 those
1:21:24 those prompt sets that you buy will the
1:21:26 prompts work yes would the prompts work
1:21:30 if you just type some [ __ ] in to chat
1:21:32 gbt
1:21:34 yes would they be as perfect maybe maybe
1:21:37 not maybe yours would be better AI is
1:21:40 literally magic changed my mind so
1:21:42 probably a potato Pate Pate p and I have
1:21:45 this uh this debate on a on a regular
1:21:48 basis in here um I am I am of the is it
1:21:53 Arthur C Clark or wait who said hang on
1:21:57 I think it's Arthur C
1:21:59 Clark hold
1:22:03 please
1:22:05 um who
1:22:15 said
1:22:19 uh Arthur C Clark yeah so Arthur C
1:22:24 Clark in
1:22:27 1962 I think three maybe said any
1:22:30 sufficiently advanced technology is
1:22:32 indistinguishable from
1:22:36 Magic that's where we are we are
1:22:38 absolutely there what what hit me within
1:22:41 within I don't know an hour of using
1:22:44 chat GPT was I've been reading about the
1:22:49 Turing test for all of my
1:22:52 life team magical AI
1:22:55 exactly um you know what paint will tell
1:22:57 you and P's not incorrect P what Pate
1:22:59 will tell you is this is a mathematical
1:23:01 equation this is a probability engine
1:23:03 he's not
1:23:04 wrong my contention is it is such an
1:23:07 advanced probability engine that
1:23:11 it um performs as magic wait P Pat's
1:23:15 banging in wait wait Bard is an AI
1:23:19 assistant Gemini is an AI model that now
1:23:22 powers behind oh yeah right oh that's
1:23:25 true all right yeah it's this is this is
1:23:28 the similar so this is the similar
1:23:31 conversation to you know here's gp4
1:23:34 here's chat GPT right so yeah I I'll get
1:23:38 my terminology correct yes correct are
1:23:40 your lives recorded in case I missed one
1:23:43 they are recorded if you go to uh
1:23:45 YouTube my YouTube channel is UM AI
1:23:48 learning lab dtt for- Tik Tock DT TT
1:23:53 someone already had AI learning lab on
1:23:55 YouTube
1:23:57 bastard and I could have gotten it I was
1:23:59 there early enough but I was just lazy
1:24:01 and didn't grab it so I'm a dumb dumb so
1:24:04 it's AI learninglab
1:24:08 dtt are the are the the back episodes of
1:24:12 these um I didn't put the I did not put
1:24:15 the video up of the night that I gave
1:24:17 out my credit card info although I could
1:24:21 because I've since cancelled that thank
1:24:23 you Pate for for being Eagle eyed and
1:24:27 going uh hey dum
1:24:31 dum oh that was that was some fun that
1:24:34 was some fun
1:24:36 action um what are we talking
1:24:40 about oh any oh this shit's magic
1:24:43 convinced me otherwise yeah this
1:24:46 is when I when I first started using
1:24:48 chat
1:24:51 GPT
1:24:53 it was
1:24:55 instantly what I thought what I was told
1:24:58 for years like 10 years that chatbots
1:25:03 would
1:25:04 be for years we were told that Siri
1:25:07 would be this that Alexa would be this
1:25:09 that AI powered chatbots would be this
1:25:13 and AI powered chatbots were never
1:25:15 anything more than logic trees that
1:25:17 broke the minute you ask them anything
1:25:21 remotely out of BS of their task what
1:25:24 they've been programmed to do if you
1:25:27 look at a lot of the layoffs that have
1:25:30 been happening in the big tech companies
1:25:32 a lot of the layoffs are happening in
1:25:35 departments where people were writing
1:25:37 logic trees like Alexa logic trees and
1:25:40 Siri logic trees because large language
1:25:43 models just [ __ ] answer the
1:25:45 question so the first the the first
1:25:48 thing that popped in my head when I used
1:25:50 chat GPT was the Turing test is done
1:25:54 like we're like we we're we're done and
1:25:57 and and what I what it was funny when I
1:25:59 first started using it I said you know
1:26:00 what's going to happen is there's going
1:26:02 to be a bunch of people come out of the
1:26:03 woodwork and they're going to be like
1:26:04 well we haven't really hit it yet cuz
1:26:06 and I saw I saw this article like a
1:26:08 month after I was thinking that where
1:26:10 this guy was like well I need to be in a
1:26:12 conversation for four to five sustained
1:26:14 hours and not like he started qualifying
1:26:18 like the Turing test hadn't been passed
1:26:19 because it hadn't hit his very stringent
1:26:22 definition definition of it I'm like
1:26:24 okay if we're into semantic arguments
1:26:26 now about nuances of the Turing test
1:26:28 then we [ __ ] passed it right um
1:26:31 weren't we going to build a GPT yeah we
1:26:33 probably were I I'll get i'll get back
1:26:34 to that and build a GPT
1:26:37 um so so Turing test if it hasn't been
1:26:42 passed we're like you know
1:26:45 I interacting with chat GPT is is more
1:26:48 like interacting with an intern than it
1:26:50 is interacting with an expert system
1:26:53 right you have to guide it you have to
1:26:55 coach it it knows a lot it it's it's
1:26:58 like got good book smarts but it's you
1:27:00 know simpleton so you got to guide it a
1:27:03 lot
1:27:04 so what do you want champy hang on
1:27:07 Champy's had it what do you want you
1:27:10 want to go out to
1:27:12 Mama oh you want
1:27:16 in oh nobody let champ
1:27:21 be
1:27:23 he wanted in do you want to sing for the
1:27:26 people do you want to do a little
1:27:28 singing you want to sing for the people
1:27:31 huh do you want to sing or you just want
1:27:34 to come say
1:27:35 [Music]
1:27:51 hi
1:27:55 [Music]
1:28:07 yeah that's a good do oh there's a good
1:28:09 doggy oh there's a good doggy did you
1:28:11 miss Daddy is that what's going on you
1:28:13 miss Daddy
1:28:14 oh oh there a good doggy all
1:28:20 right chat GPT magic all these things
1:28:24 are magic they they are
1:28:28 it's so part of it is
1:28:30 just we've now got
1:28:33 tools that are as good as or better
1:28:38 than the technology that was in our
1:28:40 Sci-Fi movies when we were
1:28:44 kids and Todd Todd really good point
1:28:47 just keep in mind it's the worst worst
1:28:49 it will ever be right now
1:28:51 exactly so we've got in our grubby
1:28:54 little hands and we don't have to be
1:28:55 technologists to do this the p in GPT
1:28:58 stands for
1:28:59 pre-trained meaning we don't need to be
1:29:02 data scientists to use machine learning
1:29:06 that's what it's been for two decades
1:29:08 well we've got this large machine
1:29:10 learning supercomputer that you're
1:29:12 welcome to use if you bring your own
1:29:13 data well what data do I bring but all
1:29:16 of it bring all of your expert set
1:29:20 data and then we will teach how to take
1:29:23 your mathematical waiting and hire
1:29:26 someone like Pate who actually
1:29:28 understands what all that [ __ ]
1:29:29 means bring your data come into our data
1:29:32 center sweat with the
1:29:35 oldies and then we'll make you a little
1:29:37 expert system that's what it was for
1:29:40 years it's not that
1:29:42 anymore Sam mman and now every company
1:29:45 on the
1:29:47 planet said let's train these things on
1:29:50 every piece of data we can get our hands
1:29:52 on the more data we throw in it the more
1:29:54 compute we throw at it the better it
1:29:55 gets let's
1:29:58 go and so these things are functionally
1:30:02 magic they are technically quite simple
1:30:04 and very flawed and janky pieces of [ __ ]
1:30:09 but they're Magic Sing champy Sing
1:30:12 champy Champy's a rock
1:30:18 star it's also per pervasive with the
1:30:21 intended unintended to slant politically
1:30:24 of its handlers yeah I think I think one
1:30:27 of the things that we'll
1:30:29 see Sam Alman kind of kind of inferred
1:30:32 this in in
1:30:34 uh in his talk at Davos this week where
1:30:38 you know they were asking him about chat
1:30:41 gp5 and what what he Bas so he implied a
1:30:44 couple of things he he
1:30:46 implied these things are going to get
1:30:48 much better
1:30:50 reasoning and so so and if if if you
1:30:55 followed any of the debacle after they
1:30:58 fired Sam Alman about qar and the rumors
1:31:01 about
1:31:02 qar and and that it was basic it's
1:31:05 basically a reasoning engine that sits
1:31:07 on top of chat
1:31:09 GPT if even a small part of that is true
1:31:12 chat GPT is going to be able to start
1:31:14 checking its own work and the minute it
1:31:16 can start checking its own work it gets
1:31:18 better really fast it can start
1:31:20 improving itself so not only will it not
1:31:23 make mistakes on your stupid little
1:31:25 question it's going to train itself to
1:31:27 get better and better and better really
1:31:28 really fast um so it's it's going to
1:31:32 reason and and uh and you know it's it's
1:31:37 it's just going to be it's just going to
1:31:38 be a completely different thing so
1:31:40 anyway how long until we can understand
1:31:42 what dogs are actually saying thinking
1:31:44 you know it's funny Todd I just
1:31:47 saw yesterday or the day
1:31:51 before
1:31:54 they're they're starting to do that
1:31:56 right now so so animal biologists are
1:32:00 starting to map you know you know
1:32:04 existing knowledge of sounds of birds
1:32:07 and animals and calls and things like
1:32:09 that and they're training large language
1:32:12 models on it and so like it's like it's
1:32:16 coming like that sort of shit's
1:32:18 coming because understanding understand
1:32:22 ing animal language is no different than
1:32:24 what they did with there's there's a
1:32:26 paraplegic out of
1:32:28 Italy um where they trained they had him
1:32:32 they they did an fmri scan and they told
1:32:35 him to think about moving his legs he
1:32:37 was you know had a spinal cord in injury
1:32:39 and you know at some point you know
1:32:42 below that couldn't move his
1:32:44 legs can't move his legs and so but they
1:32:48 had him think about moving his legs and
1:32:50 then they trained a large language model
1:32:52 on that
1:32:54 data and then they
1:32:56 sent it's [ __ ] insane they they
1:32:59 created a Bluetooth bridge that sent the
1:33:04 the M the generative AI training of his
1:33:08 brain when he thinks about moving his
1:33:10 legs around the injury into
1:33:13 electrodes at the base of his spine and
1:33:17 he's
1:33:19 walking so that that's no different
1:33:24 than an animal you know communicating
1:33:28 something or saying something like it's
1:33:30 it's it's pattern recognition right so
1:33:33 it's just different data and that's
1:33:36 that's why all these tools if if if one
1:33:38 of the things you're wondering is like
1:33:39 why are we seeing all these tools like
1:33:41 it like okay I understand that chat GPT
1:33:45 open AI figured out how to make words do
1:33:47 this but why are we concurrently seeing
1:33:49 images like image generation tools and
1:33:52 music generation tools and deep mind
1:33:55 just invented 2.2 million new materials
1:34:00 380,000 of which are stable
1:34:03 crystals in Material Science it's like
1:34:06 an
1:34:07 800-year increase in scientific
1:34:09 breakthroughs
1:34:11 overnight why are we seeing all these at
1:34:13 the same
1:34:15 time because it's the same basic
1:34:19 technology they're just throwing
1:34:20 different data at it so if you throw
1:34:23 music data at it it makes music if you
1:34:25 throw image data at it it makes images
1:34:28 if you throw brain scan data at it it
1:34:30 does brain scans it understands them it
1:34:33 understands language it understands you
1:34:36 know they're taking fmri data right now
1:34:38 and having people watch videos and then
1:34:40 they're recreating the videos based on
1:34:43 the brain
1:34:45 waves I wrote a screenplay about
1:34:48 this the technology the technology
1:34:52 for my science fiction screenplay I
1:34:54 wrote in
1:34:57 1992 the foundations for it are being
1:35:00 experimented with in Labs today that
1:35:04 could actually do what's in my
1:35:05 screenplay it's [ __ ]
1:35:09 insane it's just it's just [ __ ]
1:35:16 Bonkers Tik Tok give me a pat on the
1:35:19 back for going 90 minutes sorry chappie
1:35:21 due to your father
1:35:23 insistence that AI is Magic I cannot
1:35:25 support your chicken habit oh Champion
1:35:28 you you don't get no more chicken he's
1:35:30 going to get skinny Pate you can't let
1:35:32 my dog get skinny he's got to be that
1:35:35 wide so I can put wine glasses on his
1:35:37 back and he can walk around the house
1:35:39 carrying wine glasses somehow my parrot
1:35:42 used Alexa to call me and leave a
1:35:44 message oh that's [ __ ]
1:35:49 awesome actually oh you you know what
1:35:52 would be
1:35:54 cool you've got a
1:35:58 Parrot you know what would be
1:36:04 cool so I'm in the chat GPT app
1:36:18 right I've got a Parrot and I want you
1:36:20 to help me train my parrot say things
1:36:23 can you give me 10 really interesting
1:36:26 ideas of things we could teach the
1:36:28 parrot to
1:36:33 say absolutely training a parrot can be
1:36:36 really fun here are some interesting
1:36:38 ideas for phrases one good morning
1:36:41 sunshine it's a cheerful way
1:36:44 to no I mean whole categories like if I
1:36:47 wanted to train a parrot to quote
1:36:50 Shakespeare or
1:36:52 explain quantum mechanics and I wanted
1:36:55 you to train him all day long every day
1:36:59 could you do thinking about categories
1:37:01 of phrases here are some creative
1:37:03 categories one Shakespeare quotes teach
1:37:06 your parot so you could just turn Alexa
1:37:10 conversation on with your parrot and
1:37:12 just have it train it all [ __ ] day
1:37:15 and you just come now your parot going
1:37:17 to talk like whatever voice you
1:37:20 choose
1:37:23 oh my God that'd be hilarious you could
1:37:25 have the smartest parrot and and you
1:37:28 could also hook it up to an
1:37:30 uino and and have it you know do do
1:37:33 positive reinforcement for the parrot
1:37:35 where he gets a little seed every time
1:37:36 he says something hook it up to your
1:37:38 Arduino how but you're like but I don't
1:37:40 know how to program an Arduino sure you
1:37:44 do
1:37:48 watch chat gp4
1:37:56 I'd like to program an
1:37:58 Arduino connected to
1:38:03 a voice recognition
1:38:06 module that every
1:38:09 time the voice says good job you learned
1:38:14 another
1:38:15 phrase it activates a reward system that
1:38:20 feeds a snack to a
1:38:24 parrot can you please help me write that
1:38:39 program you could do this with children
1:38:42 you could do this with
1:38:45 dogs command good job the command good
1:38:49 job constant should correspond with with
1:38:51 the ID you assigned a good job
1:38:59 command
1:39:01 yeah that's right we're hacking pet
1:39:04 ownership now
1:39:08 kids oh for every hook that now has ADHD
1:39:12 to just interrupts when a thought comes
1:39:15 hey I resemble that
1:39:18 remark wo oh man that's really
1:39:33 [Music]
1:39:37 funny all right oh we want to talk about
1:39:41 gpts let's do
1:39:43 that make the parrot talk to a GPT do we
1:39:46 have to add parrot to our Zoo not yet if
1:39:49 we'll see if this one sticks cuno I
1:39:51 think I think I think parrots at this
1:39:54 point are just flying around outside the
1:39:57 zoo I don't think they've they've
1:39:58 decided to uh to roost but we'll we'll
1:40:01 see if this one
1:40:03 sticks I'm telling you man I the auto
1:40:08 parrot
1:40:11 trainer because because what you should
1:40:13 really do with the auto parrot trainer
1:40:15 is not do it on chat GPT but you should
1:40:16 do it on um on the
1:40:19 unrestricted llama 2 model that that has
1:40:22 no safety guard rails on it and just
1:40:24 teach it to cuss in every possible
1:40:27 language the the most vile things you
1:40:30 could imagine
1:40:36 saying and and and train it to say those
1:40:40 things when people say things like ply
1:40:42 want a cracker go [ __ ]
1:40:49 yourself this is this is is why you
1:40:51 should not have theatricals anywhere
1:40:54 near animal training uh opportunities
1:40:59 theatricals are just not to be trusted
1:41:01 with this kind of
1:41:03 [Laughter]
1:41:07 power oh I love how much fun you're
1:41:11 having with this that's
1:41:13 good
1:41:15 sometimes sometimes this channel is
1:41:17 really just me entertaining myself I
1:41:19 think most of the times this channel is
1:41:20 just me entertaining
1:41:22 myself oh my God that's
1:41:30 funny Joker there are two PODS of whales
1:41:33 on different coasts in Norway he thinks
1:41:36 they speak different
1:41:38 dialects and a third went together yeah
1:41:41 there you
1:41:43 go I mean seriously we're we're in
1:41:47 for so Google Deep Mind now
1:41:54 Google Deep
1:41:55 Mind
1:41:58 has they created 400,000 protein folding
1:42:02 predictions that are proving to be
1:42:04 correct they created 2.2 million new
1:42:07 materials 380,000 which of St are
1:42:11 stable which is like a MAG order of
1:42:13 magnitude increase in materials we
1:42:16 didn't know existed so and if you're
1:42:18 like well what's a crystal mean and what
1:42:20 what does materials mean think
1:42:23 batteries there is some sort of room
1:42:26 temperature superconductor there was
1:42:28 just a sort of false hit on that out of
1:42:31 Korea about three months ago that if
1:42:34 they hit it it changes everything it
1:42:37 changes power transmission it changes
1:42:39 Battery Technology like it's the kind of
1:42:41 thing like if they get the right
1:42:42 crystallin structures going that have
1:42:45 the right properties we'll have
1:42:47 batteries that you can charge in 30
1:42:49 seconds that last for four
1:42:52 weeks under load like that kind of
1:42:58 [ __ ] and then deep mind just
1:43:01 announced yesterday or was it today
1:43:04 might have been today that it's now
1:43:07 doing sort of like Advanced High School
1:43:10 level geometry it's
1:43:13 solving like you know geometry geometry
1:43:17 competition
1:43:20 questions
1:43:22 at like a like some crazy [ __ ] insane
1:43:25 hit
1:43:28 rate
1:43:30 we're it there's
1:43:33 just it's just going to be one thing
1:43:36 after another where because so many
1:43:38 people are focused on this stuff now
1:43:40 they're just pointing these machines at
1:43:42 all the different problems that we've
1:43:44 got and they're going to be like oh
1:43:45 here's a pile of solutions go deal with
1:43:48 that oh here are a bunch of ways we
1:43:50 could cure this genetic disease they I
1:43:53 think the FDA
1:43:55 today just approved the first Sickle
1:43:58 Cell genetic treatment that was AI
1:44:02 created AI generated AI assisted
1:44:05 something like that is that blandon's
1:44:07 caramel K it is exactly what that is
1:44:09 Amelio's
1:44:14 wife
1:44:16 um and I think I from from what I read
1:44:19 from the article that that FDA thing was
1:44:21 is is the first of its kind right it's
1:44:23 the first
1:44:25 cure that was generated in this way and
1:44:28 we're going to start to see just bang
1:44:30 bang dominoes boom boom boom
1:44:34 boom I'm messing around with a new AI
1:44:37 text to video an image to video website
1:44:40 called pix verse PX verse let's go look
1:44:44 at
1:44:50 it
1:44:51 pix
1:44:54 verse create breathtaking videos
1:45:02 [Music]
1:45:13 with there's there's pretty good
1:45:16 coherence
1:45:17 there this one looks like it's a little
1:45:20 janky let's
1:45:24 go waves are always weird their waves
1:45:27 look
1:45:32 okay
1:45:35 huh I wonder what they're using I wonder
1:45:38 if this is oh is this from
1:45:42 X oh that's that's just their X
1:45:49 Channel
1:45:54 4K upscale
1:46:09 [Music]
1:46:17 [Music]
1:46:19 feature
1:46:24 huh
1:46:25 interesting
1:46:30 interesting do they talk about does
1:46:33 anywhere talk about what it is what's
1:46:37 this Discord Community Showcase
1:46:42 home Discord Community Showcase
1:46:49 home I'm guessing this is sitting on top
1:46:53 of stab stable video diffusion but I
1:46:56 don't know it looks a little more
1:46:58 coherent than that but I I can't tell
1:47:01 just because of
1:47:04 um just because of
1:47:08 uh they're just showing you know they're
1:47:10 showing the best examples so you can you
1:47:13 can cherry-pick these things to give you
1:47:15 really good examples but it's hard it
1:47:17 it's hard to get predictable results
1:47:19 with any of these things loging
1:47:22 successful
1:47:23 create can I upload an image image to
1:47:34 [Music]
1:47:38 video let's take one of these
1:47:42 wacky we'll take that
1:47:46 dude
1:47:49 yeah let's take that
1:47:56 dude uploading image to video
1:48:00 seed strength of motion
1:48:03 prompt see what it does with no
1:48:07 prompt the system is operating at full
1:48:12 capacity all these free tools are like
1:48:14 uh we didn't think anyone would actually
1:48:16 use it this much we don't know what
1:48:18 we're doing here
1:48:22 well we'll come back to that that it
1:48:23 looks interesting it was worth worth
1:48:25 checking out um I would say I would say
1:48:28 if if you want one that it looks like
1:48:30 it's a bit more baked um just as an
1:48:33 overall tool um
1:48:36 Leonardo their their image to video
1:48:39 stuff looks pretty good and then
1:48:44 um Deco here looks pretty good there's I
1:48:47 don't like I don't know man all there's
1:48:49 just so much as moving so much is moving
1:48:52 so fast it's so hard to
1:48:55 know this is kind of the existential
1:48:57 crisis it's like I don't know where to
1:48:59 put my focus I don't know what to learn
1:49:01 right now you're like but Kyle you're
1:49:04 supposed to be telling us what to learn
1:49:05 I don't know [ __ ] start with chat gbt
1:49:07 and just keep going like I don't
1:49:11 know
1:49:14 I'm I wish I had more better answers for
1:49:17 you I wish I could quit you how where
1:49:21 can I input and manipulate my own
1:49:23 drawings for
1:49:26 charts Twisted Crone lead of Vico
1:49:32 um I don't quite know what you mean
1:49:34 manipulate drawings for
1:49:36 charts so so start uploading it into GPT
1:49:41 vision and ask it to describe it first
1:49:44 yeah that's that's a good place to start
1:49:46 but if she's saying she wants it for
1:49:49 charts you might not need
1:49:52 AI like if you're looking for if you
1:49:54 want to do charts that have Turtles
1:49:57 stacked on top of one another to
1:49:59 represent bar charts there's charting
1:50:02 software that'll do that it's not always
1:50:04 it's not always um you don't always need
1:50:08 ai ai is just a tool charts for artistic
1:50:12 per purposes oh yeah then definitely
1:50:14 here we let's go do that we'll go play
1:50:16 with that right now so here's here's my
1:50:19 the the my go-to right now is just keep
1:50:22 learning what what chat GPT makes
1:50:24 possible so let's see let's go let's go
1:50:28 to the internets and we'll
1:50:32 say visually
1:50:35 compelling
1:50:39 chart who's the dude
1:50:46 tufty like here's tuft's famous chart
1:50:49 right that's the
1:50:51 what was this from the Battle of
1:50:55 something French Russian campaign I
1:50:58 don't know some
1:51:00 War but that's the that's the Edward
1:51:02 tufty that's from his
1:51:04 book Let's see we got anything good in
1:51:07 here infographics let's go grab some
1:51:10 infographics see if we can grab
1:51:12 something interesting um that one's cool
1:51:16 all right oops so I'm just going to
1:51:20 screen shot
1:51:23 it and we'll grab all the text cuz
1:51:27 you'll you'll see this thing gets this
1:51:29 gets
1:51:34 quite it's quite something what it can
1:51:36 do all right so we're over here at chat
1:51:39 J I'm just going to paste that in you
1:51:41 can also upload it oh and I have to
1:51:44 upload it my screenshot dumped it to the
1:51:48 desktop there it is okay
1:51:52 so there it is I'm going to say
1:51:55 describe this in as much detail as
1:52:04 possible this also starts to get into
1:52:06 the accessibility thing I don't know if
1:52:08 if if the guy that was in here earlier
1:52:10 that said he was blind and this stuff's
1:52:12 given him
1:52:13 accessibility um the ability to describe
1:52:15 stuff like this in
1:52:17 details just I it's just amazing
1:52:21 um all right countries prize categories
1:52:24 chords color
1:52:25 coding okay so so now I'm going to say
1:52:29 please give
1:52:33 me five new ideas for how
1:52:39 to make this
1:52:43 chart even more visually
1:52:49 appealing
1:52:52 one of the cool things about having an
1:52:55 image generator inside a large language
1:52:57 model is that you can have the whole
1:53:00 conversation about your ideas about what
1:53:03 you want to do before you actually make
1:53:04 the image and then when it makes the
1:53:06 image it takes all of that [ __ ] into
1:53:08 account which is mid uh mid Journey
1:53:12 can't do that stable diffusion can't do
1:53:15 that but but Dolly can so what's it it
1:53:19 say Interactive elements know 3D or
1:53:21 animated visualization harmonized color
1:53:26 palette so I'm going to
1:53:29 say
1:53:32 recreate this chart with a
1:53:38 harmonized color
1:53:42 palette
1:53:48 and 3D elements
1:53:55 and now it
1:54:00 should oh
1:54:07 no when I used 3D it thought I wanted it
1:54:10 to actually create a 3D thing so I'm
1:54:12 going to
1:54:19 say recreate it with a
1:54:23 harmonized color
1:54:35 [Music]
1:54:39 palette are we going to do it I'll
1:54:42 generate the image circular layout okay
1:54:45 here we go so now it's generating the
1:54:47 image so this is a way you could do it
1:54:49 you can upload anything and you could
1:54:51 also upload you
1:54:54 know two unlike images a chart and a
1:54:58 turtle and tell it to do something like
1:55:00 there you go there's a so you know
1:55:03 there's where we
1:55:04 started and then there's kind of how it
1:55:06 interpreted that and you could say well
1:55:09 I don't like that or I do like that or
1:55:10 add this or add that and just keep just
1:55:12 keep
1:55:15 iterating Jeffrey Lombardo yeah that's
1:55:19 right
1:55:21 um all right it's getting to be it's
1:55:23 getting to be late here so let's let's
1:55:27 let's
1:55:28 start let's start winding this thing
1:55:31 down I'll start my 30 minute
1:55:34 goodbye sooner than
1:55:39 [Music]
1:55:44 later is this the nature paper with the
1:55:47 m materials prediction yeah Jasmine it's
1:55:50 it's um yeah it was the nature paper um
1:55:54 that had um from Google Deep Mind um
1:55:59 that that basically said here are 2.2
1:56:02 million new materials go play with them
1:56:04 and here are 380,000 that are
1:56:07 stable and as of a month month and a
1:56:11 half ago I think 800 of those stable
1:56:14 Crystal had been verified in the lab so
1:56:17 so because they're all there they know
1:56:19 all the crystal structures they know
1:56:20 they're stable now the labs can just go
1:56:24 test which ones actually can they
1:56:26 reproduce and do they exhibit the the
1:56:29 properties that exist so like the the
1:56:32 the speed of acceleration in Material
1:56:35 Science in the next year is going to be
1:56:36 [ __ ] Bonkers 30 minute
1:56:39 [Laughter]
1:56:42 goodbyes all right um what's the
1:56:45 difference between GPT and gpts okay
1:56:48 let's go do that interesting question so
1:56:51 chat
1:57:07 GPT I'm trying to think if I if I want
1:57:10 to go to my my box metaphor
1:57:14 so so so
1:57:18 gpts are
1:57:22 one of the things that open AI has
1:57:23 successfully done and I think them
1:57:28 calling their Bots gpts is smart because
1:57:32 everyone's going to start calling
1:57:35 anything that's a large language Model A
1:57:37 GPT
1:57:39 right
1:57:41 um so from a branding
1:57:44 standpoint GPT is going to kind of be
1:57:47 like Xerox or Kleenex where it actually
1:57:50 was from a brand um but it's going to
1:57:53 become synonymous with the generic thing
1:57:55 so so right out of the gate there's
1:57:58 confusion
1:58:00 already
1:58:02 between gp4 which is the
1:58:06 underlying foundational model right the
1:58:09 underlying large language model and chat
1:58:12 GPT which is an application sitting on
1:58:15 top of
1:58:17 gp4 which is different than
1:58:21 gpts that are little specialized
1:58:24 versions of chat GPT so wait I do need
1:58:27 to do my boxes God dang it okay and I
1:58:30 and I actually have
1:58:32 okay oh God this is this is just painful
1:58:38 okay
1:58:40 [Laughter]
1:58:43 box this is
1:58:45 gp4 see it's empty right
1:58:48 now all you do is fill it with all of
1:58:52 the knowledge of the public internet for
1:58:54 the past four decades and close the box
1:58:57 and that's
1:58:59 gp4 okay the foundational model it's got
1:59:04 all this good goodness in it but you the
1:59:06 only way you can get it in in into it is
1:59:08 through apis application programming
1:59:10 interfaces right so you make requests
1:59:14 into this thing it does its magic and it
1:59:16 sends [ __ ] back
1:59:18 out um
1:59:21 before chat GPT came out access to the
1:59:25 apis for this thing were available for
1:59:27 to Developers for years there was gpt1
1:59:31 gpt2 gpt3 there were all different kinds
1:59:35 of fine tuned models within them and
1:59:38 developers could build [ __ ] on top of
1:59:39 these things right so Jasper is one of
1:59:43 those companies copy AI is one of those
1:59:45 companies notion with all their fancy AI
1:59:49 [ __ ] now is one of those companies so
1:59:50 there's all sorts of companies that were
1:59:51 building stuff on top of
1:59:54 this chat
1:59:58 GPT it's the blue
2:00:00 one that's chat GPT is an app that sits
2:00:05 on top of this this app makes it really
2:00:07 easy to use all of the functionality
2:00:10 within here rather than giving us all of
2:00:12 the options they just say here's your
2:00:14 little chat box and here's a submit
2:00:16 button and now there's like here's a
2:00:19 paperclip button to upload [ __ ] and
2:00:21 that's what this thing
2:00:24 does a GPT is like a little mini version
2:00:28 of one of these that just kind of sits
2:00:29 on top of it that we get to build so
2:00:32 there are now 3 million of these little
2:00:34 mini things that are basically just
2:00:39 little specialized versions of this main
2:00:42 chat GPT all of these are just calling
2:00:45 the information sitting inside the Box
2:00:48 all right
2:00:50 so
2:00:51 that's gpts and
2:00:54 then like
2:00:58 uh
2:01:01 um how
2:01:03 do crickets make
2:01:08 sound and it'll just answer that
2:01:11 [ __ ] right so you can ask it anything or
2:01:14 you can say uh make me a
2:01:18 picture of
2:01:20 [Music]
2:01:21 um
2:01:25 of a caramel
2:01:29 corn kernel on a
2:01:38 tongue and it'll make you
2:01:41 that and
2:01:43 so it can kind of do anything which is
2:01:46 one of the intimidating things about it
2:01:48 is you don't necessarily know
2:01:50 what to do we do need to wait for
2:01:53 the this is a ridiculously helpful
2:01:56 visual that's good Amanda h i
2:01:59 i i i mean this is a professional video
2:02:03 Studio don't get me wrong um we Cho we
2:02:06 choose to use Simple props like garbage
2:02:10 uh because it it is easier for people to
2:02:14 understand not because this is an
2:02:16 amateur hack
2:02:17 operation oh wait wait why could I
2:02:23 not try
2:02:34 again um so anyway so so chat GPT is the
2:02:38 general tool and then we'll go we'll go
2:02:41 look at a specific
2:02:43 GPT no it's for whatever reason it's not
2:02:46 it apparently a caramel corn kernel on a
2:02:50 tongue is too pornographic or something
2:02:53 whatever um okay so I'm going to go look
2:02:56 at uh 70s let's see
2:02:59 70 7s muscle car machine so this is a
2:03:04 GPT this is this is the second GPT I
2:03:07 ever made gpts are really easy to
2:03:10 make um I'll show you how to make them
2:03:12 in a
2:03:13 second and this one's sole purpose in
2:03:16 life is to make you a post
2:03:20 of a 70s muscle car you can hang in your
2:03:25 bedroom you're like why would anyone
2:03:27 want to do that and then you might also
2:03:29 ask well wait couldn't you just do that
2:03:30 in chat GPT you could but who would do
2:03:34 that no one so this is something you
2:03:36 stumble on my metaphor for gpts right
2:03:39 now it it hit me today what gpts are
2:03:42 like because they're so early and
2:03:44 they're so janky and they're so easy to
2:03:46 make gpts right now are kind of like
2:03:49 like when you're leaving Walmart and
2:03:51 they have the vending machine where you
2:03:52 can get the little plastic cup that has
2:03:54 a plastic toy in it you get it for your
2:03:57 kid to shut him up here take the little
2:04:00 thing don't choke on it you little
2:04:02 bastard that little toy that's what gpts
2:04:05 are like like chat GPT is the whole
2:04:08 vending machine and and our the little
2:04:10 gpts we're making there's like one or
2:04:12 two cool ones in the whole machine and
2:04:14 the rest of them are just janky pieces
2:04:16 of [ __ ] so that's what that's what this
2:04:18 is this is a janky piece of [ __ ] but
2:04:19 it's a fun one and so I've got one
2:04:22 little button here and it says I'm ready
2:04:23 to design my dream car so I click that
2:04:27 button and it says fantastic let's start
2:04:29 designing your dream car the first step
2:04:31 is to to grab your preferred model so
2:04:34 let me see I'll do a I'll do a uh I'll
2:04:37 do a 71 Corvair which is not a muscle
2:04:41 car but it's kind of
2:04:43 funny the Chevrolet cor Corvair is an
2:04:46 interesting choice especially for a
2:04:48 muscle car since it's known for its
2:04:51 unique rear rear engine layout and the
2:04:53 fact that it can't take Corners at speed
2:04:59 um it doesn't exactly fit in the typical
2:05:02 muscle car category would you like to go
2:05:05 with one of these I'm going to say
2:05:10 um I want a pumped up Resto mod
2:05:15 Corvair as if it were a muscle car from
2:05:18 the 70s
2:05:25 it was also only produced until
2:05:35 1969 I'd like it to be mint green and
2:05:38 glossy with black
2:05:48 stripes
2:05:50 it should have drag slicks and chrome
2:05:56 mags and Fire coming out of the exhaust
2:06:01 pipes and so now it's going to make us a
2:06:03 little
2:06:05 poster I
2:06:08 think yeah there we
2:06:12 go so that's a GPT a GPT is just a
2:06:15 little specialized version of a thing
2:06:17 the cool things you can do with them you
2:06:19 can train
2:06:21 them you can train them on your own
2:06:24 data so if you had a PDF of drink oh
2:06:29 that's actually really [ __ ] cool come
2:06:32 on now sure it's a it's a 16 cylinder
2:06:38 it's it's apparently a 16 cylinder
2:06:40 engine in it but you
2:06:46 know and apparently it's going in
2:06:48 reverse cuz the Flames are shooting
2:06:52 forward the Corvair the only drag
2:06:55 dragster that that drags
2:07:01 backwards but I mean doesn't that remind
2:07:03 you of like you know those 70s stickers
2:07:05 of the crazy
2:07:06 Dragsters and it look it almost spelled
2:07:09 Corvair
2:07:12 [Music]
2:07:14 right oh man but but if you had a uh a
2:07:18 PD F of you know a book you wrote or of
2:07:23 specific recipes for certain kinds of
2:07:25 food your grandma made you could make a
2:07:27 GPT that just searches on the knowledge
2:07:31 that you put in it right so so super
2:07:34 super powerful stuff okay I just turned
2:07:37 59 today and I'm still excited about the
2:07:39 GPT explosion dude I'm 58 and here's the
2:07:42 deal like we are absolutely in a revenge
2:07:45 of the Gen xers moment where one of the
2:07:47 things I'm consistently
2:07:53 seeing I'm so discouraged from word
2:07:56 wrestling with chat gp4 oh with it with
2:07:58 its safety guard rails I know it's a
2:08:00 pain in the ass um one of the things I'm
2:08:04 consistently seeing is that is that you
2:08:07 know all Elder Millennials and gen xers
2:08:10 are [ __ ] lit on fire about this
2:08:13 generative AI stuff and um the younger
2:08:17 Generations don't seem to care as much
2:08:19 about it I the only thing I can think of
2:08:22 is we were all there we like we all got
2:08:25 to experience what happened with the
2:08:28 worldwide web and and you know we watch
2:08:30 the internet come onto the scene and
2:08:31 transform everything and so I think
2:08:34 there's I think there's a
2:08:35 palpable visceral holy [ __ ] this is one
2:08:38 of those things
2:08:41 um but uh but but yeah you you should
2:08:44 absolutely get excited about this stuff
2:08:46 because it's it's crazy simple to do so
2:08:47 here's how you make them you're like
2:08:49 you're not going to believe how easy it
2:08:50 is to make these things um
2:08:53 so explore gpts right and then this is
2:08:57 the GPT store so you can go find things
2:08:59 in here
2:09:02 um and then you can go look at your gpts
2:09:05 but in the upper right hand corner it
2:09:07 says create so you click on
2:09:09 Create and then this interface there
2:09:12 there's sort of two modes on this left
2:09:14 hand side there's create and configure
2:09:16 I'll get to configure in a second and
2:09:18 then over here is the preview pane or
2:09:20 they also call this the playground which
2:09:22 is confusing as [ __ ] cuz they also call
2:09:24 the developer area the playground so
2:09:27 this is a different playground it says
2:09:30 preview they call it pre playground what
2:09:32 it doesn't [ __ ] matter
2:09:34 anyway go back to evidence a that that
2:09:37 um open AI does not know how to be a
2:09:39 consumer software company yet so the way
2:09:42 you make gpts when you're in create mode
2:09:46 you just talk them into existence and
2:09:48 because I'm using uh Voice to Text for
2:09:51 it to do the typing I'm literally going
2:09:53 to talk this into
2:09:56 existence so
2:10:06 um
2:10:11 okay I want to create a GPT wait where's
2:10:14 my cursor hang
2:10:17 on I want to create a
2:10:20 GPT that asks
2:10:26 someone
2:10:29 for how they're
2:10:31 feeling
2:10:34 emotionally and then
2:10:38 responds
2:10:40 with describing a car from
2:10:44 history that best represents that
2:10:46 emotion and why
2:10:49 and then
2:10:52 renders a photograph of that
2:10:55 car that looks like it should be in a
2:11:03 magazine that's
2:11:05 it it's now building
2:11:08 it so so like literally anything that
2:11:12 you like it's
2:11:17 done um
2:11:19 tell me about your current
2:11:24 mood oh wait it's so we we've got to fix
2:11:27 these things but um so I'll say um hang
2:11:31 on I'm feeling a little
2:11:34 blue and would love a little pickme
2:11:41 up feeling blue how about a 1963
2:11:46 Volkswagen Beetle this car is Iconic for
2:11:48 its unique contended positive free
2:11:49 spirit
2:11:50 Adventure let's create a photorealistic
2:11:53 image of a 1963 Volkswagen
2:11:59 Beetle oh wait did it oh
2:12:03 regenerate I think I broke
2:12:06 something is it still
2:12:14 going I can't tell if it's still
2:12:17 rendering
2:12:24 um I'm feeling
2:12:36 blue something's
2:12:40 broke
2:12:42 oh I might have shitty internet right
2:12:47 now so now what's it going to do the
2:12:50 Bugatti type 57cc from the late
2:12:55 1930s is going to make us feel
2:13:00 better do you think as soon as it hits
2:13:03 FIFA
2:13:06 games kids will start releasing
2:13:10 AI I don't I don't quite understand the
2:13:16 question you you mean when AI gets
2:13:19 integrated into like EA Sports kind of
2:13:22 games um I think what's going to what's
2:13:25 more likely to happen that's pretty
2:13:26 slick looking
2:13:28 car I think what's likely going to
2:13:30 happen is that um everyone's going to be
2:13:33 able to make custom games that are like
2:13:36 AAA quality games in the next 3 to five
2:13:38 years by just doing what I'm doing like
2:13:41 speaking this into existence so anyway
2:13:43 so so now that this thing's made now I
2:13:45 can so let's see it says great choice
2:13:47 let's give this GPT a name how about
2:13:49 emotion Wheels how about let's see
2:14:06 um let's call it your Automotive
2:14:17 therapist so we're going to give it a
2:14:19 name and now it's going to generate a
2:14:21 profile picture so like it's doing all
2:14:24 the work right it gave it a name it's
2:14:26 going to give it the description it
2:14:27 wrote those pre-done questions there's a
2:14:30 little C thing I'm going to
2:14:36 say that image is really boring I want
2:14:39 it to be super sexy and make Fe people
2:14:42 feel
2:14:45 good so it's going to make us another
2:14:47 image
2:15:10 perfect okay now it's going to ask us to
2:15:13 start refining this thing and we're
2:15:15 going to stop here and I'm going to tell
2:15:16 you why in a second so I'm going to go
2:15:17 over here I'm going to reload this so
2:15:19 now we have the cool icon your
2:15:22 Automotive therapist I empathize with
2:15:25 emotions and describe a historical car
2:15:27 that matches with a photo that's a
2:15:29 horrible description we're going to go
2:15:30 fix that but first we're going to save
2:15:32 this thing and so if I save it to
2:15:35 everyone that's going to make it
2:15:36 available in the GPT store if I save it
2:15:38 to anyone with a link then I can you
2:15:40 know just send it to people only me
2:15:42 means only this account can play with it
2:15:45 so I'll save this as open we're going to
2:15:46 give it a category like
2:15:50 uh lifestyle good enough we confirm
2:15:57 that so there we go so that works now
2:16:00 I'm going to go fix some things in it
2:16:01 I'm going to go edit GPT I'm going to go
2:16:03 to configure now and we're going to
2:16:05 change the
2:16:06 description tell me your
2:16:10 emotion
2:16:12 and
2:16:17 aisle give you a picture of a car that
2:16:23 will make you feel
2:16:28 better and then I'm going to go um hash
2:16:33 therapy hashtag
2:16:37 um uh
2:16:43 Gearhead
2:16:45 hashtag
2:16:47 um uh
2:16:50 uh whatever Gearheads fine okay so now
2:16:54 in my instructions I'm going to
2:16:58 say wait well we're going to do one
2:17:00 other thing so we've got these
2:17:01 pre-written questions and it it it wrote
2:17:03 them in such a way that it's like when
2:17:05 you click on one of those buttons you're
2:17:07 actually asking the GPT that question
2:17:11 so when you ask how are you feeling
2:17:13 today it will actually answer that so
2:17:16 we're going to have it say we're GNA
2:17:18 click here and say
2:17:21 um let's get
2:17:25 started and then I'm going to get rid of
2:17:27 all these other
2:17:29 things so the only thing someone's going
2:17:31 to be able to click on over here is
2:17:33 let's get started okay and we'll do
2:17:36 exclamation point or how about let's get
2:17:41 rolling because it's a car get it so
2:17:44 let's get rolling and then I'm going to
2:17:46 say up up here this is our instruction
2:17:49 so this is our prompt so I'm going to
2:17:52 say follow these
2:17:58 rules if
2:18:03 someone says quote
2:18:09 let's get
2:18:13 rolling ask
2:18:16 them how
2:18:18 they are
2:18:24 feeling all right I'm going to put a
2:18:25 bullet point in front of this because
2:18:27 we're going to give it a couple of
2:18:30 rules oops did I mess that up yes okay
2:18:35 uh ask them how they're feeling and then
2:18:37 we're going to also go bullet point
2:18:41 um after they tell
2:18:46 you make the image first and
2:18:51 then
2:18:52 describe the
2:18:55 car you gave
2:18:59 them to
2:19:03 reflect their
2:19:08 emotion and why and then I'll give it
2:19:11 another bullet point I'll say
2:19:15 um provide a a
2:19:19 Freudian
2:19:21 analysis of
2:19:25 how
2:19:27 their emotion is tied to
2:19:33 that car model
2:19:36 color and year and your
2:19:41 reasoning in three
2:19:45 simple bullet points
2:19:51 points okay all right so we gave it some
2:19:54 stupid rules but but so so now what I'm
2:19:57 doing again what I'm doing here is I'm
2:20:00 programming but I'm programming in the
2:20:02 English language and you can do this in
2:20:04 whatever language you speak right our
2:20:07 native
2:20:09 language is now going to become the
2:20:13 primary interface for all computer
2:20:16 systems all of them
2:20:19 and it and it won't just be language it
2:20:21 will be take a picture of something it
2:20:24 will be you know talk into a computer it
2:20:27 will be look you know point the camera
2:20:30 somewhere it will be um a real-time
2:20:35 conversation where it will
2:20:37 dynamically um identify different
2:20:40 languages being spoken and automatically
2:20:42 translate between all of
2:20:46 them all right so so now I'm going to
2:20:48 update our
2:21:02 GPT Brandon usage guidelines oh Freudian
2:21:06 analysis so let's see provide
2:21:09 a uh
2:21:12 humorous
2:21:16 humorous mock
2:21:20 anal
2:21:21 uh
2:21:24 analysis of how their emotion is tied to
2:21:28 that car model okay this should work so
2:21:30 what I just ran into there was uh in in
2:21:33 their usage guidelines you can't do
2:21:36 anything that is um medical or financial
2:21:41 because those are regulated environments
2:21:43 right so Freudian
2:21:44 analysis okay so let's get rolling
2:21:49 so now it should ask us how we're
2:21:50 feeling glad to hear that how are you
2:21:51 feeling today let's explore those
2:21:54 emotions and find a historical car that
2:21:56 resonates with them
2:21:59 I'm feeling
2:22:02 poopy but
2:22:06 hopeful uh uh uh uh uh creating image
2:22:09 error creating image
2:22:12 why oh poopy God dang
2:22:15 it oh maybe not let's see there's an
2:22:19 issue generating at the moment however
2:22:20 let me still describe the car it CHS a
2:22:23 67 Beetle I think Beetle Volkswagen
2:22:26 Beatles are going to be solid solid
2:22:29 recommendations here um try again with a
2:22:34 different
2:22:37 model creating image error creating
2:22:41 image okay so something's broken so
2:22:43 let's go back to our little
2:22:45 GPT what did we say I said said if
2:22:48 someone says let's get roll okay after
2:22:50 they tell
2:22:53 you
2:22:58 um choose a
2:23:04 car
2:23:08 and make a 16 by9
2:23:13 wide
2:23:16 photo
2:23:18 of that car as if it were
2:23:24 photographed for a glossy
2:23:34 magazine after you make
2:23:38 the photo
2:23:42 comma oops
2:23:46 photo
2:23:52 describe the car you get blah blah blah
2:23:54 BL okay let's see if that works I don't
2:23:56 know why sometimes they just Peter
2:23:58 out this is this goes back to the the um
2:24:03 one of the technical definitions of a
2:24:06 lot of these applications that we talk
2:24:08 about a lot in this channel these things
2:24:10 are janky pieces of [ __ ] right now let's
2:24:13 get
2:24:14 rolling and just accept that as part of
2:24:16 the experience because they're going to
2:24:18 get so much better so fast that you're
2:24:21 going to be really bored a year from now
2:24:22 when these things are perfect you're
2:24:24 going to be like remember when this
2:24:25 things used to suck and we actually had
2:24:27 to use our brains to make them do cool
2:24:29 [ __ ]
2:24:31 um I'm a little blue but
2:24:37 hopeful your Automotive therapist
2:24:40 feeling a bit blue requires a d d d
2:24:43 creating an
2:24:46 image and like it told me it's going to
2:24:49 you know styled like a glossy magazine
2:24:51 photograph so I'd go in and tell it not
2:24:53 to do that so so testing these things is
2:24:55 an important part of this right make it
2:24:57 test it make it test it make it test it
2:24:59 just just keep playing keep playing so
2:25:01 oh nice 65 you're kidding me 65 baby
2:25:05 blue Mustang convertible I actually had
2:25:07 a baby blue 65 Mustang Fastback that was
2:25:11 very close to that
2:25:13 color humorous mock analysis the
2:25:15 emotional horsepower just like your
2:25:17 emotions the Mustang is powerful
2:25:19 convertible top represents your openness
2:25:21 to brighter days ahead retro Vibe the
2:25:24 Vintage aspect of the car hints it
2:25:26 Nostalgia a subtle reminder of every
2:25:27 emotion like every car all right so
2:25:30 that's
2:25:32 gpts good Lord people I said I was
2:25:36 saying goodbye like half an hour ago and
2:25:37 then you made me make
2:25:39 gpts what are you doing what are you
2:25:41 doing to me look at my
2:25:43 hair that has not gotten
2:25:46 better
2:25:50 all right
2:25:51 people we going to do this I'm going to
2:25:53 get out of here okay so here here's the
2:25:55 thing if you're new here
2:25:58 go
2:26:00 there if you're new to AI Show and Tell
2:26:03 please okay if you're new to ai go to
2:26:06 the salon. click on the first link it
2:26:09 says the AI Salon Community this is a
2:26:12 community that that where the people
2:26:15 that hang out in this channel also hang
2:26:17 out here when when I'm not live and it's
2:26:20 a really remarkable Community where we
2:26:22 help each other you can go there you can
2:26:24 talk about AI no one thinks you're a
2:26:26 weirdo you can be a beginner you can be
2:26:28 a 30-year vet it doesn't matter anyone's
2:26:30 welcome how many PDFs can I load load at
2:26:33 once I think five uh no I think 20 20
2:26:37 for for the knowledge base I don't know
2:26:40 the size limit I think it's 100
2:26:42 megabytes per or 100 megabytes
2:26:45 total um and I'm going to go check out
2:26:49 the show and tell
2:26:55 Channel R don't
2:26:59 choke oh
2:27:07 wait see I can't say anything in here
2:27:10 where it turns into
2:27:13 evidence that's lovely I supplied a
2:27:16 photograph of a GPT to produce a
2:27:18 professional photo apparently I'm not
2:27:20 recognized as a
2:27:23 55-year-old that's good Becky looking
2:27:25 looking sweet it's it's also not very
2:27:28 photo it's not a photograph right it's
2:27:30 an illustration but it's beautiful
2:27:33 lovely that's
2:27:35 great a i Salon Zoo oh that's from uh
2:27:39 parrots want to be included in the zoo
2:27:42 that's from uh Firefly Adobe Firefly
2:27:45 very
2:27:46 cool
2:27:47 actually yeah we we should be playing
2:27:49 with Adobe Firefly to get some of the
2:27:51 groovy uh groovy titles in here you know
2:27:54 what would be fun is to is to pop some
2:27:56 of some of these things from Adobe
2:27:58 Firefly on top of images made in
2:28:04 Del Kyle stuck in one of his many rabbit
2:28:06 holes that's that is just too [ __ ]
2:28:09 close to
2:28:12 reality time to go RV in nice oh that's
2:28:16 beautiful all a conspiracy I tell you
2:28:19 who's who gets in
2:28:22 next really
2:28:25 good why does his chest look like a butt
2:28:28 hole sort of or what I don't know what
2:28:31 that is but it's not I think I can get
2:28:33 permanently banned from this channel for
2:28:36 that bird that's beautiful actually
2:28:39 ooh Cory Sandler I think you're right I
2:28:42 think I think parrots might need to get
2:28:44 in the show but you're right the other
2:28:46 animals are going to going to be very
2:28:47 jealous that's that's like that's like a
2:28:50 serious [ __ ]
2:28:52 showoff if I were a llama I'd be pissed
2:28:55 at that for one thing it's it's about
2:28:58 six times the size of a
2:29:00 llama oh those are beautiful actually I
2:29:03 I'm thinking AI Salon taken over by
2:29:06 parrots yeah I think it's good I think
2:29:08 they're
2:29:09 in yeah it's funny the baby parrot's not
2:29:13 as cute I think the oh that's beautiful
2:29:16 Dag all right all right what was it like
2:29:20 before 2024 we had to type type search
2:29:24 the
2:29:26 web poly want a
2:29:31 GPT oh my God sweating with the
2:29:36 oldies oh man all right you guys are
2:29:39 awesome awesome awesome awesome okay
2:29:42 everybody sleepy Mr it I know me too
2:29:44 okay I'm going to get out of here I'm
2:29:46 going to go Hy hdate and get some sleep
2:29:48 good night Kyle thank you for all you
2:29:49 give us absolutely um you're a great
2:29:52 friend to us all thank you I appreciate
2:29:54 that Dr J Becky I'm so done typing
2:29:59 Beretta oh Beretta but it was a white
2:30:01 bird yeah that's right n never mind
2:30:03 let's see good night Kyle thank you for
2:30:05 all you give us yep
2:30:07 absolutely
2:30:09 uh all right who's in here forever
2:30:11 hooked Robert Rossy thank you as always
2:30:13 Rudy good to see you Rudy saigo hope
2:30:16 all's well down there in South Africa I
2:30:18 guess you're just you're getting ready
2:30:20 for D or something right or lunch I
2:30:23 don't know what time it is there but
2:30:25 it's it's not word it is here U no never
2:30:29 mind it's still rendering oh yeah it's
2:30:30 something rendering what was it it
2:30:32 doesn't
2:30:33 matter oh what was the show it was uh
2:30:37 Columbo I survived my first mod night
2:30:39 Cam kin you did great awesome thank you
2:30:41 very much for modding thank you to the
2:30:42 mods thank you to the Irregulars thank
2:30:44 you to my subscribers thank you to
2:30:46 anyone that hangs out on a Friday
2:30:48 [ __ ] night we had like 165 170 people
2:30:51 in here there's 147 in here right now
2:30:55 that blows my mind um thanks for hanging
2:30:57 out I hope this is worth your time um
2:31:01 Winston good to see you uh Brendan thank
2:31:04 you for what you do I appreciate it Joey
2:31:07 pie steo you to man peace out dude uto
2:31:11 good to see you Lord digital Gods
2:31:14 sporting the turtles tonight very ni
2:31:17 Zack and allaha have a great night thank
2:31:18 you thank you Tobias to you
2:31:20 too Tracy thank you thank you for uh
2:31:24 coming and hanging out and doing all
2:31:26 that stuff Pate as always there's
2:31:28 analogy here between parrots and AI
2:31:31 models well don't you know this pate
2:31:33 parrots are
2:31:38 magic parrots are harbingers of magic
2:31:41 Pate that's why they have to be in the
2:31:43 ai ai learning lab
2:31:45 Zoo
2:31:48 all right everyone I'm out of here peace
2:31:50 out have a good night I'll see you
2:31:53 tomorrow