aiLL 8/14/23 - Navigating the Wild West Early Days of AI

Description

Chapters

Transcript

0:01 foreign
0:19 [Music]
0:31 are we live are we live
0:34 has anyone joined cat dragon number one
0:38 in the house
0:40 welcome welcome welcome
0:42 um
0:44 hey
0:45 hey
0:48 got a scratchy voice from
0:51 being out in the world welcome to the AI
0:54 learning lab my name is Kyle Shannon I'm
0:57 coming to you live from New York City
1:03 New York
1:04 New York New York it's a hell of a town
1:11 the Bronx is up and the battery's down
1:16 welcome everybody
1:20 this is not my normal place I am in a
1:23 hotel room just a little teeny hotel
1:26 room right next to Times Square
1:28 we're riding right in the middle of
1:30 Manhattan we're right in the middle of
1:31 it all everything's happening
1:33 everything's happening
1:36 it's fantastic Bob tell him what he's
1:38 won
1:40 you want a new car
1:44 welcome everybody Welcome we'll get
1:46 started here what day is today
1:49 Monday right
1:52 I think it's Monday
1:56 a couple of meetings today
1:59 one of the meetings
2:00 the dude we flew out to sea didn't show
2:03 up yay
2:08 I love coveted work commitments
2:10 [Laughter]
2:14 welcome everyone this yeah learning lab
2:15 my name is Kyle Shannon
2:17 a little bit about my background I try
2:20 to do these lives every night
2:21 I kind of missed last night although we
2:23 did a live stream uh from
2:26 downtown Manhattan right near Ground
2:29 Zero it's beautiful gonna walk around
2:31 walk by the old uh
2:34 hey Dallas cooler what's happening
2:37 walked down by my old offices my
2:39 oldagency.com offices we walked around
2:40 Lower Manhattan that was cool and then I
2:42 was gonna do a live last night but I got
2:44 home and I was just toast
2:47 because I've gotten up at like three in
2:48 the morning that morning and it was you
2:51 know by the time I got to my hotel room
2:52 it was like 10 o'clock
2:54 deep in the bowels of Jersey
2:57 so um so anyway so we're back we're live
3:01 tonight it is what time is it eight
3:05 something oh nine o'clock 908 here in
3:08 New York City New York City the way this
3:12 works I'll talk about this so so these
3:14 websites right here if you're new to
3:15 generative AI if you have not played
3:17 with chat CPT go to that top one
3:20 chat.openai.com I'll talk about these in
3:22 a second
3:24 um but just go there and start playing
3:25 right now if you haven't played
3:27 um I'm an entrepreneur I'm the
3:30 co-founder and CEO of storyvine uh we
3:32 are in New York visiting clients and
3:36 talking about all sorts of work they
3:38 want to do
3:39 everyone seems to be excited about
3:43 generative AI our clients are asking us
3:47 for their input on how they should deal
3:49 with Microsoft how they should deal with
3:51 their agencies which is kind of funny
3:53 because we're a
3:54 we're a small but humble a small yet
3:57 Mighty company
3:59 um I don't know if we're Mighty but
4:00 we're we're we are small we punch above
4:02 our weight
4:04 um so so that's exciting I'm also I was
4:08 also an entrepreneur in the early days
4:09 of the world wide web and one of the
4:11 reasons I'm so fired up about this
4:13 generative AI stuff is the parallels
4:16 between the early days of the web and
4:18 the early days of chat GPT and kind of
4:21 the shift from the Geeks and scientists
4:24 and mathematicians owning
4:27 um generative a or owning machine
4:29 learning and AI it you know chat GPT
4:32 sort of shifts it into
4:34 now the rest of us get to play with it
4:36 which is exactly what the World Wide Web
4:37 did for the internet
4:39 so that's that sole proprietor LLC uh no
4:43 we are a C Corp and we have
4:47 six or seven employees and an outsourced
4:49 Tech team we're an 11 year old company
4:53 um
4:54 the uh
4:57 what we do on the channel here if you
4:58 got questions pop them in the comments
5:00 below I do my best to answer all the
5:02 questions so I'll likely fall behind
5:04 so let's let's jump in
5:07 my voice is a little shot oh it happened
5:09 I just lost my
5:10 I lost my dual screen
5:13 hold please
5:19 that's not what I wanted is this what I
5:22 wanted
5:23 effects no that's not what I wanted
5:27 hope please
5:34 where did my
5:37 dual camera go I knew I figured this out
5:39 at one point before
5:47 enhance
5:50 enhance oh there it is there's dual
5:52 camera well not special
5:55 you know you want to know who uh tick
5:57 tock tick tock has some help
5:59 doing their little interface here the
6:01 they they had some help
6:03 and the person that helped them do their
6:06 interface what could that have been well
6:08 well I know who it was it was Satan
6:12 Satan made it completely impossible to
6:14 use well fantastic
6:16 [Music]
6:18 okay
6:20 back to where I was going
6:22 Okay so
6:24 this
6:26 this little list of things if you're new
6:28 to generative AI kind of hang on I got a
6:30 different setup so I'm trying to trying
6:33 to figure this out on the Fly here
6:36 um
6:37 all right there you go so so feel free
6:41 to go ahead and take a screenshot of
6:42 that
6:43 um
6:44 so so these are all sort of variations
6:46 or competitors to chat GPT if you
6:49 haven't played with chat EBT that's the
6:51 official website right there
6:53 then you've got Bing and Claude two kind
6:56 of major Bing is sort of a co-op Edition
6:59 kind of thing that's Microsoft's version
7:02 of gpt4 it's connected to the internet
7:05 it can generate images it can read
7:07 images Claude is a competitor from a
7:10 company called anthropic it's got a
7:13 hundred thousand token context window
7:14 meaning you put in really long prompts
7:17 um so that's pretty good Bard is
7:20 Google's entry into this
7:22 perplexity.ai is a really interesting
7:24 kind of search engine citation GPT kind
7:28 of thing but the
7:29 labs.perplexity.ai they're giving us
7:32 free access to the meta Lama 2 models
7:35 and then Pi is from a company called
7:37 inflection posed from the company quora
7:41 all of these are worth exploring so
7:45 um so that's that another thing that I
7:47 wish I didn't have to talk about but I
7:49 do
7:50 is oops is how do you set up your light
7:54 in a hotel room apparently not very well
7:56 thank you I'll be here all night don't
7:59 forget to tip your waitress try the
8:00 chili
8:02 um
8:05 the the AI learning lab has gotten so
8:08 successful
8:09 that that we've been scammed we've been
8:14 scammed
8:16 so
8:17 there is a
8:19 um a copycat account of the AI learning
8:22 lab that's got two L's in it it's a i l
8:25 l
8:26 so two L's for learning that's not how
8:28 it's spelled
8:29 so if you see that one they're [ __ ]
8:32 scammers they've copied my content
8:34 they've copied the name and they're
8:35 trying to get people to sign up for some
8:37 [ __ ] scammy multi-level marketing
8:40 [ __ ] I will not if I reach out to
8:44 you for multi-level marketing [ __ ] I
8:46 will call it a pyramid scheme
8:48 it'll be crystal clear
8:50 but we don't do that here so if if
8:53 you've been reached out I apologize I
8:55 don't there's no reason for me to
8:57 apologize they're the [ __ ] but
8:59 that's going on so I'm sorry about that
9:02 um if you're brand brand new to this uh
9:05 this whole world and you don't really
9:07 know where to start and you're like I
9:08 don't even know what chat EPT is go to
9:10 that thing prompts.chat
9:12 we'll teach you all about it what these
9:14 large language models are how to talk to
9:15 them it's got a whole list of like all
9:18 these different personas that you can
9:19 tell Chachi BT to act like a
9:21 screenwriter or an accountant or a
9:23 lawyer or whatever all that sort of
9:24 stuff and it's just it'll give you lots
9:26 of ideas of what it makes possible
9:28 if you're into the image stuff those
9:31 things at the very bottom mid Journey
9:33 Adobe Firefly and generative fill within
9:36 Photoshop Leonardo is a really cool web
9:39 interface and from a stable diffusion
9:41 clipdrop.co is a bunch of tools from
9:44 stability AI from stable diffusion
9:47 and that prompt book thing right there
9:51 will give you access
9:53 to lots of knowledge about how to use
9:56 those things and then future PD is just
9:58 a big old list of AI tools there's like
10:00 4 000 in there blah blah blah blah okay
10:03 so that's that the next thing tomorrow
10:05 night is the the AI salon so go to the
10:09 salon.ai we've got a Meetup Group there
10:11 you can sign up for the Meetup group and
10:13 uh
10:15 um you'll get information about when the
10:16 group is it's tomorrow night from five
10:18 to seven mountain time so seven to nine
10:20 East Coast time because of my business
10:24 travelings I don't think I'm going to be
10:26 there for the beginning of the meeting
10:27 but I'll probably join as a participant
10:29 Leah and some one of the other
10:31 organizers are going to host tomorrow
10:32 but we have John Knack from Adobe from
10:37 the AI group at Adobe is speaking
10:39 tomorrow so
10:41 the ones behind Adobe Firefly John Knack
10:44 is is uh one of the leaders in that part
10:46 of the organization he's a really good
10:48 guy I used to work with him
10:50 um he's wicked smart he talks faster
10:52 than I do and uh and he's gonna just
10:56 talk about what they've learned and and
10:58 you know all that sort of stuff so that
10:59 should be a really fascinating
11:01 fascinating thing so go to the salon.ai
11:04 sign up for the Meetup you can also sign
11:06 up for a Discord where you can come play
11:08 hang out ask questions make images
11:11 whatever
11:15 all right and my voice is going I don't
11:17 know how long I'm gonna laugh tonight
11:19 but I'm gonna do my best I'm gonna do my
11:22 best it's gonna be it's you know what it
11:23 is it's gonna be special it's gonna be a
11:26 special night special
11:28 it's gonna be special
11:34 okay let me get to the questions I'm
11:37 gonna scroll up a little bit here and
11:38 see what anyone's asked see if we got
11:40 any trolls by the way if you're gonna
11:42 troll
11:43 at least be creative about it just don't
11:46 give me some boring ass [ __ ] like oh
11:49 isn't it nice that the geriatrics are
11:51 trying to teach people about AI if you
11:54 say that kind of [ __ ] you can just jump
11:55 off a flipping Bridge
11:58 all right hello awesome what are you
12:01 doing in New York I'm out here we've got
12:02 a Pharma client that was like hey could
12:05 you come out here and talk to us about
12:06 all this special AI stuff you're doing
12:08 and whatever this automated video
12:11 technology you have for your company
12:13 whatever you're doing with that we want
12:15 to talk to you about that and they talk
12:17 just like this
12:18 I'm going to talk to you about that
12:21 we're like okay we can come out there
12:24 do you want us to get on a plane we'll
12:26 get on a plane and we'll come out there
12:27 and we'll stay in a hotel
12:29 so we did that and then so we had one
12:31 today in Jersey well no we had two
12:34 meetings today in Jersey although we got
12:35 we got stiffed at one of our meetings
12:38 [ __ ] didn't show up I mean the
12:40 lovely professional uh that I'm sure had
12:43 some of their commitment
12:45 didn't show up
12:48 well the other meeting we had was really
12:49 really good and then tomorrow we've got
12:54 two meetings and then possibly like a
12:57 drive-by with a very important person
12:59 and then we have uh dinner with uh with
13:03 a prospective client tomorrow night and
13:05 we're going to
13:07 um new on Joe Korean Barbecue on 32nd
13:11 Street right off Fifth Avenue in
13:14 Koreatown
13:15 and if you're in New York and you
13:17 haven't been to new one Joe that's a
13:20 mistake you got to go there
13:22 you got to go to new one Joe it's
13:24 amazing it's amazing it's amazing
13:29 all right I use chat CPT job interview
13:32 plug-in to prep for a data analyst
13:34 interview and loved it so helpful oh
13:36 that's pretty cool awesome
13:37 which one the the job interview plug-in
13:40 oh that's really cool
13:43 all right
13:46 oh yeah
13:47 Dallas cooler the other thing I did in
13:50 New York I didn't come here to do this
13:51 but I knew I did I knew I was coming
13:53 here for a client
13:56 um but I also knew that I could get my
13:59 world ID and I did I got my world ID so
14:03 I got scanned by the orb my irises are
14:06 now
14:06 in the in the blockchain I don't think
14:10 they are but uh it was cool it was a
14:13 little anti-climactic because the the
14:15 place was called it wasn't called The
14:17 Grove what was it called the canvas I
14:19 think it was called the canvas
14:21 and so I'm like oh it's the canvas I
14:23 thought it was like some hipster work
14:25 Collective or something like that no it
14:28 was a clothing store
14:29 just like a boutique clothing store in
14:32 in a in the uh the subway entrance down
14:36 at uh Ground Zero at the World Trade
14:38 Center
14:39 um is the it's this beautifully stunning
14:42 building and so I went in there and it's
14:45 basically just a big mall of really
14:46 expensive shops right so so this store's
14:49 there and then you just walk up and
14:50 sitting next to the cash registers the
14:52 silver orb and the and the the you know
14:55 the women at the clothing store is like
14:58 oh yeah that thing we don't know
14:59 anything about that but you know make
15:01 your little app do what it does and just
15:03 look into it that was it and I looked
15:07 into it had like a little ring that you
15:09 know showed progress as it was scanning
15:11 my irises and at some point I had to
15:12 make them wider because it kind of
15:14 stopped she goes oh you may need to open
15:15 your eyes wider so you know took my
15:17 glasses off opened my eyes wipe
15:20 and then when it completed the ring it
15:22 it went somewhere and then it it uh
15:25 pinged on my phone in the app and it
15:27 said um it said
15:29 analyzing your uniqueness and it was
15:32 like this little bar going across I mean
15:34 you're verified
15:36 and so now I have this virtual kind of
15:39 nft ID card that you can flip backwards
15:42 and forwards and it's Anonymous which is
15:44 kind of cool so it doesn't have my name
15:45 on it just says Anonymous verified human
15:48 with my you know I've got a private key
15:50 for my identity
15:52 so the idea is that
15:54 there'll be a whole
15:56 um infrastructure built where developers
15:59 can create applications where you have
16:01 to have one of those IDs to participate
16:04 so imagine a version of Twitter
16:07 that was World ID verified where you
16:10 knew that everyone in there was a
16:11 verified human and then imagine tying
16:13 that to another application layer where
16:16 you could say not only is that human
16:18 verified but this piece of content that
16:20 they created is also verified on the
16:22 blockchain
16:23 and we can also verify that it hasn't
16:25 been modified I think that stuff's going
16:27 to become really really important so
16:29 anyway yeah I did that too
16:31 so that was kind of cool
16:32 welcome sole proprietor nope we talked
16:35 about that Joy hail hey Joy hackers
16:39 heaven have you used worm GPT I haven't
16:42 I've heard about it that one and there
16:44 was chaos GPT for a while or wasn't it
16:46 called chaos GPT the the sort of nasty
16:49 ones that are going to go out and do all
16:50 sorts of hacking and destroy the world
16:52 yeah we'll see we'll see I mean here's
16:55 the good news hackers are gonna hack
16:56 right hackers are going to jailbreak
16:58 [ __ ] hackers are gonna hack hackers are
17:00 going to turn autonomous agents loose
17:02 just to see what they can [ __ ] up
17:04 but then the good news is is that the
17:06 the good guys have access to these tools
17:08 too so the good guys are going to go
17:10 [ __ ] up some hackers so you know it'll
17:13 be entertaining just hopefully they
17:15 don't
17:16 do something so bad that it turns
17:18 everyone off to this stuff because that
17:20 could slow it down for a year or two
17:23 but I don't think it's going to slow
17:25 down for a year or two I just don't
17:27 here's the thing about all these tools
17:29 here see about all these tools see all
17:31 these these are just like the chat GPT
17:32 ones these are just the words
17:36 um they ain't going away they ain't
17:38 gonna be uninvented
17:39 um Chachi PT
17:41 saw to that when they got 100 million
17:44 users in six weeks now
17:47 chat EPT also in the past two months the
17:50 usage has gone down
17:53 does that mean it's over no
17:55 it just means you know hey there's other
17:58 options a lot of those other options are
18:00 free
18:02 um and B people are trying to figure out
18:05 you know is this valuable to my life or
18:07 not this software is really new it's
18:09 really thin meaning it's not very
18:12 sophisticated right now it's a lot of
18:13 spit and duct tape so um some people if
18:17 if it's not ready for prime time they'll
18:18 just sort of turn off to it so
18:20 so that's what's going on there I think
18:22 what type of consulting jobs do you see
18:24 in AI
18:26 um I think all sorts of them um I've
18:29 been doing
18:30 um
18:31 you know I've been doing stuff
18:33 within my own company right trying to
18:35 figure out what's the value proposition
18:37 look like what's the future state of the
18:40 technology look like what's the product
18:42 roadmap look like
18:44 but in pitching that to my clients and
18:48 you know establishing some credibility
18:50 like I know a little bit about what I'm
18:52 talking about a lot of those clients now
18:54 are saying hey can you help us figure
18:56 this out
18:57 um so if you if you start getting AI
19:02 literate if you start getting good with
19:03 these tools and start talking about it
19:05 just put it out there talk about it at
19:07 dinner parties talk about it at work
19:09 talk about it on LinkedIn especially
19:13 um people start reaching out to you so
19:15 that's so that's one thing you can do
19:17 you can teach workshops
19:19 um there are a lot of agencies now
19:21 selling automations so they're they're
19:23 using tools like zapier or make or
19:26 bubble you know automation low code no
19:28 code tools
19:30 um
19:31 to make automations and so there's a
19:33 bunch of agencies starting up going hey
19:35 we can make you a Content generator or
19:37 whatever it might be whatever automation
19:41 um companies want to be straighter or
19:42 whatever it might be whatever automation
19:46 um companies want to be so I'm seeing a
19:48 lot of automation agencies
19:52 um
19:53 what types of consulting jobs
19:58 if you start to establish yourself out
20:00 there as someone who's thinking about
20:01 this stuff
20:03 um
20:04 just some opportunities will rear their
20:08 heads I was I was working with a group
20:10 the other day and this guy said hey can
20:12 you stay after Kyle and I said sure
20:14 and he's got some sort of opportunity
20:17 where it's a large company they're
20:18 having a 2 000 person meeting they want
20:20 to do some really cool [ __ ] with AI
20:22 there's not much time to turn it around
20:25 and I said well see if they're willing
20:28 to pay for it if they are let's go see
20:29 if we can build something so we might
20:31 have to build something in like two
20:32 weeks
20:35 I love it
20:37 have you noticed whatever you do a
20:40 Google Search now gives you a generative
20:42 AI response as well is that just in
20:44 general Google I have not noticed that I
20:46 know they had
20:48 didn't they have like search.google.com
20:50 or some
20:52 some [ __ ] like that hang on let's see
21:03 um what
21:06 I got one
21:08 what
21:11 what is the average speed
21:15 of an unladen swallow
21:28 no I got nothing here
21:32 I got nothing I think there's I think
21:34 there's there's a uh
21:37 a version of this that
21:40 um yeah hang on hang on I gotta move
21:43 something I gotta flip something around
21:45 I've got my computer
21:48 on a on a pillow and it's sort of
21:50 leaning toward toward me so the
21:54 the computer is
21:57 is doing bad things it's still a bad
22:00 thing it's making me cranky this is not
22:03 my setup
22:05 I just knocked the phone off the little
22:08 receiver yeah
22:09 uh
22:11 come on what are we doing very good are
22:14 we good everybody good
22:16 all right
22:19 okay
22:21 [Music]
22:25 um oh I'm gonna light where's my light
22:32 got a little indirect bounce lighting
22:34 over here huh see that gives me a little
22:36 glow it makes me a little dead it's a
22:39 little cold but whatever
22:42 um what was I saying oh have I noticed
22:44 um so the one thing I have noticed is
22:47 when you go to Bard
22:49 um
22:50 Bard if you ask a question that that it
22:53 would make sense to incorporate a link
22:55 like you know what what are some shoe
22:57 companies that make a certain kind of
22:59 running shoe
23:01 um it's incorporating links into the the
23:04 GPT responses I've noticed that but I
23:07 haven't noticed normal Google doing
23:09 doing their thing that doesn't mean
23:12 they're not doing it I just my default
23:14 search engine now is Bing
23:17 which a year ago I would have bet your
23:20 money my default search engine would
23:21 never be Bing but you know
23:27 it is because it's got gpt4 in it so I'm
23:30 liking it
23:32 okay let's see
23:34 how are we
23:38 okay
23:41 have you used worm GPT I have not Kyle
23:44 Shannon in the Big Apple what up new
23:47 joke
23:48 in New York
23:50 it's really good here I love New York
23:54 the people are amazing oh you know what
23:55 I did so so one thing Monique and I my
23:59 business partner and I were here and
24:02 um we went down to Saint Mark's Place
24:06 um there's a there's a Japanese
24:07 restaurant there that you know you you
24:09 you went into it and all the tables were
24:12 kind of at floor level and and you would
24:14 sort of sit on the floor and your legs
24:16 would go under the table and this is
24:18 very traditional Japanese thing so we
24:19 went to that restaurant
24:21 and we're all excited to have this
24:23 traditional Japanese restaurant
24:24 experience they've completely gotten rid
24:26 of all that traditional [ __ ] it's just
24:28 regular tables now it's like
24:32 but then so we went down there and we're
24:35 staying up up near Times Square and um
24:38 we um we got those City bikes the
24:41 electric ones those were a blast there's
24:43 bike Lanes everywhere in New York now if
24:45 you haven't been here in a while there's
24:46 bike Lanes everywhere so he rode up from
24:47 Lower Manhattan to Times Square on these
24:50 electric bikes those things are fast
24:52 so it was you know it's a little sketch
24:55 because you know you're flying around at
24:57 like 40 you know not 40 miles an hour
24:59 probably 25 miles an hour but you know
25:01 close proximity to people in cars with
25:05 no helmet so but it was fun
25:08 um your area of expertise can be
25:10 enhanced and augmented by AI absolutely
25:12 agree Winston I think these things are
25:14 ability amplifiers
25:17 um if you have not played with them if
25:19 you have not had what I call your Kevin
25:21 McAllister moment with AI which is
25:26 play with it until you have until it
25:28 does something that you just you
25:30 literally can't process
25:34 and then once you have that moment then
25:36 you kind of dig deeper and you realize
25:39 um this is
25:41 you'll be like oh that's pretty
25:43 impressive but you know at least it
25:45 can't do what I do
25:47 and then it will
25:49 and then you get nervous and that and
25:52 that begins your crack addiction to this
25:55 stuff
25:58 all right tick tock just jumped me to
26:01 the bottom of the questions because it
26:03 loves me
26:04 oh let's see
26:11 uh what are we talking about debugging
26:14 tensorflow no I'm not that deep into it
26:16 I'm not that deep into it I'm talking
26:19 about generative AI more think post chat
26:22 gbt
26:23 as companies and people try to figure
26:26 out how to use these tools more so than
26:28 companies and you know things trying to
26:31 create these tools
26:33 so
26:35 um
26:35 this is what I did in the early days of
26:37 the world wide web and the World Wide
26:39 Web first came out I was not you know on
26:42 the W3 Council coming up with HTML tags
26:45 I was using HTML tags to help Fortune
26:48 500 companies build their websites and
26:50 so that's kind of where I am with this
26:52 stuff as well so my expertise is more in
26:54 using the stuff once it kind of hits
26:56 that
26:57 Prime you know ready for prime time
26:59 which this is just hitting it so we're
27:01 very early if you think you're behind on
27:03 this [ __ ] you're not we're very early on
27:06 it but
27:07 it's moving fast it's moving way faster
27:10 than you know it's moving way faster
27:12 than you know
27:15 [Music]
27:18 I'll see you there
27:20 did you get kicked out of your office no
27:23 I'm in New York I'm in a hotel room a
27:26 salty little hotel room just off Times
27:28 Square
27:29 maybe I'll go for a walk later just
27:32 people watch watch the [ __ ] Insanity
27:34 of Times Square
27:36 good to know that you're a verified
27:37 human thank you please experience
27:39 explain your T-shirt please it is
27:44 New York New York established 1898
27:48 Original Big Apple
27:51 I don't think this shirt was bought in
27:54 New York and I actually kind of hated
27:55 wearing it out because if you
27:58 if you wear a shirt that says New York
28:00 City and you live in New York City
28:02 you're a tourist
28:05 so
28:06 but I live in Denver now so my wife
28:08 bought this at some previous trip she
28:11 did and so this was the only t-shirt I
28:13 had in my bag so there you go
28:16 so I just went out of the tourist
28:19 all right any way to control a device
28:22 via AI like Google Assistant
28:25 um
28:26 I don't know what you mean by control of
28:28 Voice or control a device
28:32 um
28:34 you could I mean I mean you could
28:36 certainly you could certainly I mean
28:37 here's the deal chat GPT knows all the
28:41 programming languages so you could
28:43 certainly take something like a
28:44 Raspberry Pi and have it generate a
28:47 bunch of code for you and then you could
28:49 hook an API on the front end of that so
28:52 you could talk to chat gbt and then
28:56 if you did the automation right it would
28:58 send
29:00 um commands to your Raspberry Pi so yeah
29:03 I would think I would think you could
29:05 mock something like that up
29:07 pretty easily and
29:10 um even if you're not super technical
29:13 you need to be a little technical but if
29:14 you can figure out a Raspberry Pi which
29:16 is
29:17 that's not that tough and then
29:20 um Chachi BT can even teach you how to
29:22 do that just tell chat gbt say Hey
29:24 listen
29:25 I want to get into microcontrollers and
29:28 I got myself a Raspberry Pi
29:32 4.0 point x Point r
29:35 and it'll know what that is and then
29:38 just go tell it tell it to teach you how
29:42 to do that [ __ ] so yes I think you
29:44 probably can I don't know anything about
29:46 that
29:47 that's now you're getting into Robotics
29:49 and things like that is your salon
29:50 tomorrow after party in New York City oh
29:52 that's actually a good idea so the salon
29:54 is tomorrow
29:56 um if you want to go to the salon go to
29:59 the salon.ai the first link there is the
30:01 Meetup link so sign up for the Meetup
30:03 and you'll get the login information for
30:06 the Google meet tomorrow night
30:08 um John Knack from Adobe is speaking
30:11 um I've got a dinner with a client so
30:14 I'm probably not going to make the
30:15 beginning of it but I'll sort of join as
30:17 a participant later but Leah my co-host
30:20 is co-hosting with another another one
30:22 of the organizers and uh yeah we'll be
30:24 Off to the Races
30:26 tomorrow from seven to nine Eastern five
30:30 to seven Mountain four to six Pacific so
30:34 yes but thank you for asking Casey have
30:37 you wait I asked you several weeks ago
30:38 about implementing AI in a high school
30:40 classroom I ended up giving a two-part
30:44 lesson that ended with students creating
30:46 a list of do's and don'ts oh that's
30:47 really cool
30:49 and I've been showing them each day ways
30:51 to use it responsibly
30:53 responsibly so far so good that's super
30:56 cool and I'm curious so I love that
31:00 newsy floozy by the way 9.5 out of 10 on
31:02 the name like the name a lot the solid
31:04 it's so it's solid it makes you smile
31:08 you got your Rhymes you got your
31:10 alliterations it's intriguing you want
31:12 to know more
31:14 um
31:15 uh yes zero my t-shirt is New York
31:18 I would be curious to know if there if
31:21 their list of do's and don'ts the the
31:23 high school students list of do's and
31:24 don'ts changes
31:26 once they start using it right like like
31:29 if they're new to it I would imagine the
31:31 list of do's and don'ts might be
31:32 different than if they're you know in a
31:34 month so be interesting to see how that
31:36 evolves But but so first off like
31:39 [ __ ] Kudos so rather than hiding from
31:43 this you're bringing it into the
31:45 classroom you're getting them engaged
31:47 critically thinking I like this is what
31:50 I talked about when when you ask the
31:51 question if I if I recall correctly is
31:54 the most important skill moving forward
31:56 is critical thinking
31:58 because these things are going to do all
32:00 the Tactical execution so it's not about
32:02 memorizing [ __ ] anymore it's not about
32:04 learning these you know specialized
32:06 individual skills it's about okay we've
32:09 got tools that can do all the Tactical
32:11 execution what do we need to do we need
32:14 to like conduct the symphony we need to
32:17 say
32:18 what do we want to create what does
32:20 success look like
32:22 um if these things generate a bunch of
32:24 content
32:26 how do you know if it's any good I don't
32:27 know you look at it you use your [ __ ]
32:29 brain
32:30 and so I think you're you know this
32:33 feels very right to me right you're
32:35 heading down this journey of okay kids
32:38 these tools are here let's start
32:40 exploring them let's figure out what
32:42 works let's figure out what doesn't
32:45 what's it good for what's it not good
32:47 for
32:49 and and you know you come back three
32:52 months from now it's going to be a
32:53 completely different answer I love that
32:55 I think that's awesome
32:59 um so
33:00 Kudos man oh Casey is also a teacher
33:03 awesome
33:06 hello hello hello
33:10 nice yeah yeah that's fantastic
33:14 newsy floozy are you a teacher me too
33:16 Sony is a teacher oh that's right Sonya
33:18 I think you're in New York Sonia aren't
33:19 you I'm in New York I'm in New Jack City
33:22 um
33:23 Hotel I am at the
33:27 Hyatt House I think I'm
33:31 I don't know where it is somewhere near
33:32 Times Square
33:34 you could cross business wait you could
33:37 cross business analysis with generative
33:39 AI use cases and do a small pilot
33:42 project yep that's it that's a really
33:44 good
33:45 a good thing to do
33:47 um with code interpreter if you've not
33:49 played with chat gpt's code interpreter
33:51 the data analysis you can do with it is
33:54 stupid powerful and stupid easy
33:58 and also frustrating when you're trying
34:00 to do something really specific because
34:02 it kind of goes off and starts prompting
34:04 itself and
34:06 you ask it a simple thing and it's like
34:08 let me try 16 ways to do that but it's
34:11 it's it's it's amazingly it's staggering
34:14 to watch you're just watching this
34:16 machine essentially start reasoning like
34:19 like code interpreter from chat GPT I
34:23 mean gpt4 is is just mind-blowing in and
34:27 of itself but then code interpreters
34:28 kind of Step Above that because you kind
34:31 of get a glimpse into the future of what
34:33 these things are going to do and right
34:35 now it's early enough that you can kind
34:37 of it's all slow and you can see the
34:40 code that it's writing and you can see
34:42 it making mistakes and trying to unmake
34:44 the mistakes
34:46 but you can kind of see when when this
34:49 [ __ ] gets sophisticated
34:51 you'll just ask it for [ __ ] and it'll go
34:55 aye aye Captain I'm off I'm off I'm
34:58 gonna go get you your tickets you're
35:00 gonna have a fantastic time at the Rock
35:02 concert in that obscure town you want to
35:04 go to
35:05 and off it'll go here you go
35:09 here's your tickets
35:11 many companies are willing to peel off a
35:14 few bucks for at least exploring this I
35:16 yeah more and more are
35:19 um
35:19 are starting to do that we did a we in a
35:23 workshop that I just gave we did we
35:24 showed a chart of
35:26 um
35:28 it was a McKinsey report on marketing
35:31 this was from back in April or May on
35:35 marketing and sales that McKinsey did
35:39 and 70 of the respondents said that they
35:43 expected in the future their companies
35:45 would be using AI you know a lot or or a
35:49 lot a lot right and and there was a
35:51 there was a companion chart that
35:52 basically said what are you doing now
35:54 and it was like not very much and so a
35:57 lot of companies right now are expecting
35:59 they're moving in this direction they
36:01 don't quite know how to do it yet but no
36:04 one does
36:06 you know the fact that I the fact that
36:09 I've got some Pharma Executives asking
36:13 me for advice on what their agencies
36:16 should be doing
36:18 um is telling right so
36:21 it's it's we're very early
36:24 very very early I heard chat GPT is
36:26 going bankrupt
36:28 um I don't know where you heard that is
36:30 that news today
36:32 I doubt that's the case I'm sure they're
36:34 spending
36:36 daggering amounts of money
36:39 um
36:39 chat GPT bankrupt
36:45 news about chat GPT bankrupt
36:51 declining users and competitors
36:55 could make the
36:57 is reportedly heading toward bankruptcy
37:00 tech.co what are the sources here
37:04 I don't see any real
37:11 Times of India
37:14 who is this Tech talk
37:22 this is Yahoo news
37:34 um
37:38 yeah I just I you know while it's
37:41 possible I I don't
37:44 I don't see that happening I don't see
37:46 Microsoft letting that happen for one
37:49 thing they put in 10 billion dollars Sam
37:50 Altman said he thinks they need to raise
37:52 a hundred billion to do to get to AGI
37:56 artificial general intelligence
37:59 um
38:02 even even with their reduced usage like
38:07 they're not making money right now on
38:09 the 20 bucks a month from us that
38:10 they're just using that to cover their
38:12 costs
38:13 um they're going to make their money on
38:15 apis and and that sort of [ __ ]
38:17 so
38:19 um I don't know I wouldn't it like it's
38:22 possible I'm sure at their current burn
38:25 rate I'm sure they're burning through
38:26 staggering amounts of cash Sam Altman
38:29 said when when uh when Chachi VT first
38:32 launched and it got way more popular
38:34 than they expected it to he said the the
38:37 costs of of what it was costing them he
38:40 said it was eye-watering or something
38:43 like that it was just you know
38:45 crazy turn on bard
38:48 why what's
38:49 oh turn on Bard in
38:52 in uh
38:54 in Google search hang on
38:57 google.com
39:00 you mean go to bard.google.com yeah
39:02 that's not really a search though
39:05 let's see
39:12 I don't see a way to turn on Bard here
39:16 but if you mean bard.google.com that's
39:18 not really search that's just bard
39:21 uh or do you want me to show you Bart
39:24 I'm happy to do that as well so let me
39:26 just chat you but you've got lots of
39:27 stuff Jason rusy
39:29 what's this news about it going bankrupt
39:32 we just talked about that is a career in
39:33 AI profitable
39:35 um I don't know I think I think uh
39:38 we're all gonna have careers in AI like
39:41 like like where we are right now is
39:45 very similar to the early days of the
39:46 World Wide Web where you could ask you
39:48 know is a career in the world wide web
39:50 profitable in 1994 or five
39:53 I don't know if it becomes something
39:56 sure
39:58 um same thing with this like I I think
40:00 because
40:02 nobody knows what to do with these right
40:05 now everyone's trying to figure it out
40:07 so right now if you can figure this
40:09 shout [ __ ] out even a little bit people
40:11 will pay you money
40:13 if you're a clever business person and a
40:15 good salesperson they'll pay you a lot
40:17 of money
40:20 if you're not good at selling your
40:22 services and you can't figure out how to
40:24 make money
40:26 using Photoshop then it's you're gonna
40:29 have similar problems doing this but if
40:31 you've figured out a way to solve
40:33 problems and get people to pay you to do
40:35 that then you're going to be doing great
40:37 with these things if you can get your
40:39 head around them so
40:41 um I think a career in AI is inevitable
40:44 and I think the sooner you jump on
40:47 getting literate with this stuff and
40:49 figuring out how you can take your skill
40:51 set and solve other people's problems
40:53 using these tools to make yourself
40:55 better
40:56 smarter faster all that sort of stuff
40:59 done great then you will and yes I think
41:02 it will you will be able to charge a
41:04 premium for some amount of time
41:06 because these things are very
41:08 complicated and very impactful
41:12 um if I play it out even just three
41:14 years into the future
41:16 um people are going to be desperate to
41:17 have people that understand what in the
41:20 [ __ ] are these things and how do we do
41:22 them
41:23 all right
41:26 some comments in life were filtered to
41:28 protect the community's experience don't
41:30 be jerks don't be jerks in the AI
41:33 learning lab
41:34 um do me a favor if you would share the
41:36 live uh I just want to get some more
41:37 people in here if we can share the live
41:39 if you're willing to and and hit the
41:41 hearty heart button the the double tap
41:43 thing we'll get some likes going in here
41:45 the other thing you should know
41:48 I try to answer all the questions and I
41:50 can tell because I'm tired and my voice
41:51 is shut I'm rambling tonight so sorry
41:54 about that but you know that's that's
41:56 part of the charm of the channel isn't
41:58 it if you know what I mean it's like you
41:59 know what we do sometimes we just talk
42:01 we just talk we just have little
42:02 conversations
42:04 um
42:06 and uh what was I gonna say I am now 58
42:10 messages behind so if I haven't gotten
42:12 to your message
42:13 it's not because I'm a douche and it's
42:14 not because I'm an avatar it's because I
42:17 haven't gotten your message when I try
42:19 to read them all because I'm a
42:21 conscientious Tick Tock Live host
42:24 what's going on with the hair tonight I
42:26 got a different product in it I got some
42:28 sort of mud that I carry in my little
42:29 travel pack seems like it's holding
42:31 pretty good I mean you know it's it's
42:34 still we got there's some issues don't
42:36 get me wrong but it's hanging in there I
42:38 got a little Buster Poindexter thing
42:40 going but this time of night that's to
42:42 be expected
42:43 all right let's see to show students how
42:47 to use AI
42:49 as a writing assistant that will break
42:50 prompts down into several
42:53 smaller questions and then answers to
42:55 the prompt also generate ideas newsy
42:57 floozy I love what you're doing I think
42:59 that's really really flipping awesome
43:00 Grim Reaper wrote something in Russian
43:03 or some sort of that kind of language
43:07 is that Slavic it's not Slavic is it
43:09 it's
43:11 whatever Russian is but but it might not
43:15 be Russian it just looks in that
43:16 neighborhood of languages uh
43:19 I don't know how to copy and paste
43:21 comments but but we can copy and paste
43:23 that comment if if someone wants to copy
43:25 and paste that Russian comment and
43:26 translate it for me assuming it's not
43:28 some
43:29 evil nasty douchey thing
43:32 um I'm happy to look at it
43:46 sorry I'm trying to find my where I was
43:52 somebody put their phone number in there
43:54 not sure what that's about
43:57 not sure that that's super wise but
43:59 that's okay writing analysis writing
44:01 analysis for GPT is far superior to
44:05 Claude correct I agree gpt4
44:09 I think Claude and GPT 3.5 are close
44:13 maybe clouds inch and up above it
44:17 um gpt4 is definitely better but but
44:21 Claude you can put in a 75 000 Word
44:25 document and it'll just do it or you can
44:28 have a really long conversation with it
44:30 and it won't lose its memory for 75 000
44:32 words
44:33 so Claude has advantages I I'm telling
44:36 you I will pull [ __ ] off this list
44:39 when
44:41 it doesn't have something that's unique
44:43 to the others so chat EBT has gpt4 code
44:47 interpreter custom instructions and
44:51 plugins so it's the leader it's the best
44:54 right now thank you Luna stick welcome
44:57 Moon even you Moon comedian you
45:00 um Claude has a hundred thousand
45:03 token context something oh thank you for
45:05 the Hardy heart I like the hearts hearts
45:08 are sweet I like turtles
45:10 um Bing is connected to the internet
45:12 Bing can also read images Bing can also
45:15 generate images Bing is also what's
45:18 going to be the basis or not being Bing
45:21 chat
45:21 the basis for
45:24 um the co-pilot product that's going to
45:26 be inside all of office 365. Bard is
45:29 from Google it can read images it's
45:31 connected to the internet and it doesn't
45:33 suck as bad as it used to and is
45:35 probably going to get much better
45:36 because they kind of [ __ ] up and are
45:39 having to play catch-up to those guys
45:41 and so they're spending a lot of money
45:45 um perplexity is a really interesting
45:47 company they they've got good searches
45:50 with citations and they're letting us
45:51 play with the Llama 2 models for free at
45:54 labs.perplexity Pi just raised 1.3
45:57 billion dollars and Pi is an incredibly
46:00 conversational bot which if you want to
46:03 do work is a pain in the ass but if you
46:05 want to have a conversation it's
46:07 actually really good
46:08 it's really good at conversations and
46:11 then Poe you can play with multiple
46:12 models and you can it's got a bot
46:14 builder in it all of these have
46:15 something to play with
46:17 so yes I agree but
46:20 don't discount the others just because
46:22 one's better at this than the other
46:24 at some point someone's gonna roll a
46:27 bunch of this [ __ ] together we're going
46:28 to have kind of a master GPT that does
46:32 a lot of these things well and at that
46:33 point I'll jettison some of these things
46:36 but right now worth worth and again
46:40 don't think of any of these things as
46:42 the final state
46:44 these things are all brand [ __ ] new
46:47 this list used to be three like three
46:50 weeks ago I only had three things on
46:51 this a lot of this stuff's new
46:56 so just know that it's going to change
46:58 it's going to change It's Gonna Change
47:00 Gonna Change It's Gonna Change It's
47:03 Gonna Change
47:04 all right
47:06 oh wait by the way whoever just said
47:08 that writing analysis uh MIM okay your
47:12 name is annoying
47:17 I'm glad I don't stutter your your name
47:21 is designed to make people stutter is
47:22 that what's going on if it's that it's
47:24 clever
47:26 um here's a trick I heard that I think
47:28 is really good
47:30 if you have a really long document
47:32 use Claude
47:34 to summarize it and and kind of reduce
47:36 the word count say say summarize this
47:39 you know basically make this have less
47:42 words
47:43 and make it more efficient and then you
47:46 can make it small enough that you can
47:47 take that summary to chat GPT and use
47:50 chat gpt4 to do the true analysis of it
47:53 so use Claude as a distiller and then
47:56 gpt4 to actually do the analysis that
47:58 might be a cool way to do large
47:59 documents best of both
48:03 all right our school has blocked chat
48:06 GPT but not Claude ah interesting I'm
48:09 preparing to push for chat GPT to be
48:12 unblocked good yeah do that the other
48:13 thing news floozy I guarantee you they
48:15 haven't blocked Pi
48:17 um for your students that have uh
48:20 iPhones I think there might be I think
48:22 Pi does Pi have a
48:24 I don't know you can go check see if
48:26 they have an Android app but but the
48:29 conversational
48:30 um nature of of Pi could be really
48:34 interesting for your students because
48:36 it's so different like what you get from
48:38 chat GPT and what you get from PI are so
48:40 different there might be something
48:42 interesting they could do some sort of
48:44 analysis of what's different about it
48:47 right and how are they doing what
48:48 they're doing and how might they have
48:50 prompted pi to act like that and then
48:53 you could even go to oh you you don't
48:56 you can't you don't have chat gbt yet
48:57 well when you get chat GPT back
49:00 Chachi PT has its custom instructions so
49:03 you could go
49:05 write custom instructions for chat GPT
49:07 to have it act like Pi that would be an
49:10 interesting exercise
49:13 does your wife pack for you uh laughing
49:16 my ass off so ADHD I'm totally ADHD uh
49:20 yes she did help pack for me and yes and
49:23 she actually had she had secret clothes
49:26 hidden away in case I go to New York in
49:28 the summer so it's like khakis I look
49:30 like I'm from [ __ ] Connecticut I have
49:32 like a like a polo shirt and khakis and
49:35 I yeah I just look like yeah
49:40 me wanting Raspberry Pi
49:43 yeah exactly well I know I know that's
49:47 they call it Raspberry Pi what are you
49:49 gonna do
49:51 tis the season too the raspberries are
49:54 coming pretty soon I asked mid journey
49:57 to generate a chart of all edible
49:59 berries that produce something really
50:00 nice that's super cool
50:02 um if you haven't seen there's there's a
50:04 there's a cool prompt in mid-journey
50:06 that is is one that
50:09 um
50:10 that makes me very happy I'll show it to
50:13 you
50:16 okay if you haven't oh wait I might have
50:19 slow crappy internet here in the hotel
50:21 what's going on
50:23 come on
50:24 what are we doing
50:27 are we working
50:29 slash imagine
50:34 I don't know something's
50:36 something something's gone South here
50:47 uh let me quit Discord and see
50:51 if I relaunch it if it is less shitty
51:01 do you know what a noling photograph is
51:03 okay
51:05 n o l l i n g
51:10 checking for updates I think my internet
51:13 is poopy
51:15 how's the live is this live okay give me
51:18 give me some hearts if it's okay
51:23 uh uh or is it laggy yeah I don't think
51:26 that's gonna work
51:28 all right I got a couple of Hearts so it
51:30 seems like it's okay all right
51:32 um it seems like my laptop is not happy
51:35 it's it's just not happy
51:37 it does not seem to like
51:41 whatever internet they got going on here
51:45 and very disturbing it's very disturbing
51:48 all right
51:51 what are you gonna do
51:53 um
51:54 if you're in mid-journey you can also do
51:58 the I know there's some some lures for
52:00 this for stable diffusion
52:03 um
52:04 type noling photograph and then a topic
52:08 so like Nolan photograph VW Bug noling
52:13 photograph hunting supplies nulling or
52:17 just noling k-n-o-l-l-i-n-g
52:22 um
52:23 and so noling photograph is one where
52:25 they take all the parts
52:28 they'll take like in a like a they'll
52:30 take an assembled something and
52:31 disassemble it and lay all the parts out
52:33 in a really organized
52:35 um
52:36 you know uh grid layout and then they
52:39 photographed that that's what noling is
52:42 so you can get mid-journey to do nulling
52:45 on all sorts of [ __ ] like mushrooms or
52:48 that you know you said all the different
52:50 berries or whatever it was and it did a
52:52 pretty good job do do a nulling
52:54 photograph of all the berries and it'll
52:56 lay them out really nice for you
52:59 um Grim Reaper's still giving us some
53:00 Russian I I don't think I could can I
53:03 copy
53:07 I can pin this comment I can translate
53:09 it
53:11 oh Grim Reaper I can read your stuff
53:13 look at that we got I'll tell you what
53:16 the AI just did the AI just I can read
53:19 Russian I can read Russian now right now
53:21 why did chatbt invent new poisons but
53:25 not a cure for cancer well okay GP chat
53:28 gbt didn't invent [ __ ]
53:32 if he didn't invent [ __ ]
53:34 human beings
53:36 used these tools
53:39 to do bad [ __ ]
53:41 they also used these tools to create
53:44 Alpha fold and created four hundred
53:47 thousand proteins in two days that is
53:50 going to completely revolutionize drug
53:53 discovery
53:55 and they probably will find a cure for
53:58 cancer so it these tools didn't do [ __ ]
54:02 people did [ __ ] if people choose to use
54:05 these tools that are more powerful than
54:09 anything we've had in the history of
54:10 humanity for bad [ __ ] well then
54:14 whatever motivates them to do that
54:16 that's their thing
54:18 part of what this channel is about is to
54:21 try in this the little way we can to
54:24 bend how we use this [ __ ] toward the
54:27 good bend it toward the light
54:31 because I personally think and why I'm
54:33 spending so much [ __ ] time like on my
54:36 trip out where I'm [ __ ] exhausted
54:38 spending hours a night doing this
54:41 is that
54:43 individual contributions that use this
54:46 for the good right now are gonna have a
54:49 disproportionately
54:53 um positive impact on the future
54:55 trajectory of this technology
54:58 I really believe that like I really like
55:01 deep down believe that
55:03 are there [ __ ] that are going to do
55:05 bad [ __ ] with technology yeah of course
55:08 we're human
55:10 some of us are [ __ ]
55:12 some of us have ulterior motives
55:16 some of us are being paid to [ __ ] [ __ ]
55:18 up
55:20 and the rest of us are just trying to
55:21 live our lives
55:24 and if we can use these tools to live
55:26 our lives a little bit better [ __ ]
55:29 far out
55:30 and I think
55:33 if we bend it toward the light
55:35 maybe we outweigh
55:39 those other actors
55:42 I noticed the detail under each Berry
55:44 was made up of gibberish yeah so right
55:47 now
55:48 um
55:50 um
55:52 what's it called text text not so good
55:56 but I did find out from someone on Tick
55:59 Tock I thought what was her name
56:01 I forget it but
56:03 hopefully she'll be in here tonight
56:05 um
56:07 there's a thing you can do at
56:09 clipdrop.co which is stabilities thing
56:12 clip drop
56:15 dot Co
56:16 let's see if my internet is yeah the
56:19 internet's just hosed on this machine
56:21 let me try turning off Wi-Fi and turning
56:23 it back on
56:25 maybe we got a little we got a little
56:27 situation here
56:29 we got a little situation yeah isn't
56:32 that special oh you know who's running
56:34 the internet here at the hotel it's
56:37 Satan Satan's doing it now look at all
56:40 those all those unwashed people looking
56:42 at pornography in their little hotel
56:44 rooms well and that's special
56:46 maybe they're eating up all the internet
56:47 with their little porno pornographic
56:50 pictures of the the loot and lascivious
56:52 acts no not special wow wow okay create
56:58 stunning visuals in seconds
57:00 you can make money watch at gbt wait so
57:03 that worked and then
57:06 um
57:08 oh someone just commented in my tick
57:11 tocks
57:12 they said they believe
57:14 websites will be obsolete soon I
57:17 actually kind of agree with that
57:19 and I'll tell you why in a second I'll
57:22 tell you why
57:23 when my voice is really tired I can do I
57:26 can do low low low low how low can you
57:29 go
57:30 well let me tell you boy
57:34 all right the the internet just sucks
57:36 here all right so we're not gonna deal
57:37 with that
57:38 um if you go to clipdrop.com
57:42 who hasn't said that the stuff under the
57:45 image
57:46 so ion China if you go to clipdrop.co
57:49 and use
57:51 um what's it called
57:53 uh
57:56 is it sketch
57:58 no it's uh it's called doodle
58:01 doodle to image or something like that
58:03 anyway one of the tools on clipdrop.com
58:07 lets you doodle an image and then and
58:10 then write a prompt and it will use your
58:12 doodle as kind of the composition of the
58:15 of the image it does text really well
58:17 like like not really well it does text
58:21 decently predictably so I did the other
58:24 night I did like I did a square and then
58:27 I had a little square inside it with a
58:29 little triangle above it so I I did like
58:31 a sign hanging in a window and I put
58:34 Help Wanted sign in a shop window
58:37 and it generated three images and one of
58:41 them absolutely said help wanted in
58:43 clear text the one was total gibberish
58:45 and one was mostly help wanted
58:47 so I think with the sketch tool at clip
58:50 drop you can start to do some predictive
58:51 text it's not going to be that long
58:54 until they crack that nut they'll start
58:56 layering in
58:57 some of the stuff adobe's doing with
58:59 their with in Firefly they've got a type
59:01 tool there where they're taken full-on
59:03 fonts and like re-rendering [ __ ] on top
59:05 of them at some point they're going to
59:06 start layering that stuff together in a
59:09 prompt and you know sort of dynamically
59:10 compositing it or
59:12 or something there's also technology
59:14 that's they can make stuff more
59:16 predictable now so there's some cool
59:17 stuff coming
59:21 but right now text sucks so you still
59:24 gotta take [ __ ] into Photoshop and fix
59:26 it
59:26 I he that that's the thing about these
59:28 tools oh there's lots of people here
59:30 welcome everybody my name is Kyle
59:31 Shannon this is the AI learning lab I'm
59:34 not my normal uh thing I'm in a hotel in
59:36 New York because I'm here on business
59:37 I'm tired as [ __ ] my voice sounds like
59:39 [ __ ] well I don't know if it sounds like
59:41 [ __ ] but it just sounds like it sounds
59:44 like I've been chewing on gravel yeah
59:47 you know Tom Waits uh yeah let's go
59:54 why do I do this I don't know
59:59 yeah anyway that okay
1:00:03 welcome everybody
1:00:06 she's my raspberry pie that's nice can
1:00:08 you explain why it's so obsessed with
1:00:11 obfuscating text yeah I can actually
1:00:14 so it's not obsessed with obfuscating
1:00:17 text it is not aware of anything do you
1:00:21 want to see how
1:00:22 images are created
1:00:27 I have a presentation for that
1:00:34 let's see
1:00:49 yeah I had a feeling so I've got my my
1:00:53 problem with my little mac is
1:00:56 I've got
1:00:59 um
1:01:00 too much [ __ ] on it so my big files get
1:01:04 dumped up to iCloud
1:01:08 do I have anything
1:01:11 I have it anywhere no I don't
1:01:15 I can't tell you how to do it well so so
1:01:17 here's the deal I'll I'll talk to you
1:01:20 how
1:01:22 the image generation things work
1:01:24 so
1:01:26 the way image generation works
1:01:29 there's two models that are involved
1:01:31 there's the large language model right
1:01:34 and the way a large language model works
1:01:36 is you put in your prompt and then it
1:01:39 understands those words and it it
1:01:42 doesn't it it actually doesn't
1:01:44 understand those words it turns those
1:01:46 words into a mathematical probability of
1:01:50 the very next word that should be the
1:01:52 response
1:01:53 or the first word of the response and
1:01:55 then it takes your prompt plus that word
1:01:57 and it creates the probability for the
1:01:59 next word so it is
1:02:04 if you distill down what it's doing it's
1:02:06 actually a quite it's quite a simple
1:02:08 mathematical function
1:02:11 your prompt turns into a mathematical
1:02:14 probability it creates a word creates a
1:02:16 word creates a word creates a word and
1:02:18 it writes all that [ __ ] and it looks it
1:02:20 looks like
1:02:21 it's sentient and it looks like it's
1:02:23 reasoning and it looks like it's smart
1:02:25 and it looks like it says it's a
1:02:27 probability engine it's got the
1:02:29 awareness of what it did the same
1:02:32 awareness that your calculator has when
1:02:35 you type two plus two
1:02:37 and it gives you four well how did it do
1:02:40 that it must be conscious no it's a
1:02:41 [ __ ] calculator
1:02:43 so so it's literally like that it's a
1:02:47 fancy [ __ ] ass version of that but
1:02:49 but that's what it is
1:02:51 so that's that's the the language part
1:02:54 of of things like mid-journey and stable
1:02:56 diffusion the image part at least for
1:02:58 the the diffusion model works like this
1:03:00 Gans work a different way there's
1:03:02 there's different ways to do it but for
1:03:03 the most part what we're playing with
1:03:05 right now are diffusion models what
1:03:07 diffusion models do is they take a bunch
1:03:09 of images that have been tagged with
1:03:11 words okay so there's a cat in that one
1:03:14 there's a dog in that one there's a desk
1:03:16 in that one there's right
1:03:18 and then they they run every single
1:03:20 image that's been tagged through a noise
1:03:24 filter they they noise it so it starts
1:03:27 as an image and then it turns it into
1:03:28 random noise
1:03:30 through a particular noise algorithm
1:03:32 okay
1:03:33 then when you type in your prompt it
1:03:36 basically goes and says go find me all
1:03:39 of the
1:03:41 um
1:03:43 the the
1:03:45 images
1:03:47 that have that were tagged with any
1:03:49 words that are relevant to that prompt
1:03:51 so it's using the large language model
1:03:52 as a filter to filter out right so maybe
1:03:55 you go from a billion images that are in
1:03:58 the training set and you filter that
1:04:00 down to
1:04:01 you know a hundred thousand
1:04:04 it's not seeing the images the images
1:04:06 are these noised random noise patterns
1:04:09 and so when you put in your prompt it
1:04:11 filters it down and then it averages
1:04:13 those hundred thousand noisy images into
1:04:16 a single noisy image
1:04:18 and then it uses that same noise
1:04:20 algorithm to denoise it
1:04:22 toward toward whatever it looks like
1:04:26 it's not aware of all of the pieces of
1:04:30 individual grains of sugar if you think
1:04:32 of every picture in the training set as
1:04:34 a grain of sugar but you dump it all
1:04:36 into cake batter mix and Stir It Up with
1:04:39 eggs and butter and [ __ ] and put that in
1:04:42 the oven
1:04:43 you don't know which grain of sand
1:04:45 generated to that so all it's doing is
1:04:48 denoising some random pixels into
1:04:50 something and so the reason text is
1:04:55 garbled is because it might know that it
1:04:58 needs text but it's meshed together you
1:05:01 know many many images
1:05:04 and is you know kind of just denoising
1:05:07 them it doesn't know what's in there
1:05:09 it's got no awareness of what it's doing
1:05:11 it's just a mathematical
1:05:13 calculator
1:05:15 so which I like once you understand how
1:05:18 they work I think it's even more
1:05:19 remarkable that they do what they do but
1:05:21 but that's why I can't do text so in
1:05:23 order for it to do text they're gonna
1:05:25 have to do some more specialized
1:05:27 algorithms that maybe say okay you know
1:05:30 start denoising the image and then go
1:05:33 look at the text on it and maybe you
1:05:35 have a separate layer that's doing text
1:05:37 in the context of that image so they'll
1:05:39 be more sophisticated ways to generate
1:05:41 images in the future but
1:05:43 I will leave that to the mathematicians
1:05:45 I mean whoever figured this [ __ ] out is
1:05:48 it's amazing
1:05:52 it's amazing it's just amazing it's
1:05:54 amazing what time is it it's getting
1:05:56 late here in New York City it's getting
1:05:59 late late my voice is going
1:06:03 sorry to hear that it's banned there oh
1:06:06 I know that sucks the chat GPT police
1:06:08 that really sucks
1:06:09 hi I need to improve my English go to
1:06:11 these things they're really good at it
1:06:14 put in your language oh you know what
1:06:17 you can do uh
1:06:19 Lati latiz Latisse
1:06:22 here's what the Doula tease
1:06:25 in your language in your native language
1:06:27 go to these things
1:06:29 and have it create custom
1:06:33 um tutorials for you
1:06:35 the way you like to learn like if you
1:06:37 like to learn with flashcards have it
1:06:39 give you flash cards if you like to
1:06:41 learn with you know something like Babel
1:06:43 or you know one of those things
1:06:46 um have it give you lessons and it will
1:06:48 and I would even imagine you could go to
1:06:50 Pi and just talk to pie
1:06:53 [Music]
1:06:55 Translate
1:06:59 all right Grim Reaper medical
1:07:00 corporations won't let Chachi PT come up
1:07:03 with a new cure for diseases you know
1:07:05 why well yeah that capitalism right they
1:07:07 you know like
1:07:09 Pharma companies don't want to cure for
1:07:11 cancer
1:07:12 again
1:07:13 medical corporations are not
1:07:18 going to not let chat GPT come up with
1:07:21 it
1:07:24 what big corporations will do if they
1:07:26 want to squash something is they'll let
1:07:28 again chatgpt is not coming up with this
1:07:31 [ __ ] human beings prompting chat you're
1:07:33 coming up with this [ __ ] if one of those
1:07:35 people comes up with a cure of cancer a
1:07:38 cure for cancer
1:07:39 but those corporations might do is buy
1:07:42 that technology and then bury it
1:07:45 but they're not they're not stopping
1:07:48 this [ __ ] from doing it and the fact
1:07:49 that meta open sourced llama 2
1:07:54 means that those big ass corporations
1:07:57 can't do that like let's say that the
1:08:01 big Pharma companies could strong arm
1:08:03 open Ai and Claude and Bing and barred
1:08:10 and Pie and Poe and everyone else that's
1:08:13 coming
1:08:15 and they can never do anything that
1:08:18 would cure cancer well you still got
1:08:20 llama 2 out there in the public domain
1:08:23 without any safety guard rails on it
1:08:25 they can't they can't control that
1:08:27 those companies control it
1:08:30 after the invention not before
1:08:33 I think yes exactly I told them it's a
1:08:37 living document and we will revisit it
1:08:39 throughout the year that's cool any good
1:08:42 AI tools for short form video editing
1:08:44 other than video or Opus I don't know I
1:08:47 I haven't
1:08:48 I've been doing a lot of that I I would
1:08:51 just go to futurepedia.io go to the
1:08:54 video section there and you know learn
1:08:57 just start playing and just know that
1:09:00 most of them are going to be a waste of
1:09:01 your time there's probably a lot of
1:09:03 video tools but there might be something
1:09:05 good in there
1:09:08 um videos the video is the one that I've
1:09:10 used
1:09:12 but even that I mean the problem with it
1:09:14 is they aren't cheap they're really
1:09:16 [ __ ] expensive which
1:09:18 I I suppose it's expensive in compute
1:09:21 but I have a feeling it's just expensive
1:09:23 because they can right now there's not a
1:09:24 lot of competition
1:09:28 [Music]
1:09:30 did your meeting go well one went really
1:09:33 well really new big opportunity and one
1:09:37 the [ __ ] dude didn't show up we flew
1:09:40 out here to meet with two people one of
1:09:42 the guys today didn't [ __ ] show
1:09:47 that like
1:09:50 if there was some sort of emergency
1:09:52 great
1:09:53 if there wasn't that is [ __ ]
1:09:56 classless
1:10:02 there will be manipulation with AI Grim
1:10:05 Reaper yeah
1:10:07 I don't know what you mean you mean with
1:10:09 content yeah there's gonna be a lot of
1:10:11 manipulation
1:10:12 thought you were wearing a Trekkie shirt
1:10:14 now it's New York City
1:10:17 the Indian time says they may go
1:10:18 bankrupt by the end of the year is the
1:10:20 end is the Indian times
1:10:22 um
1:10:23 respectable I don't I mean I don't know
1:10:25 I don't know that sounds it sounds
1:10:27 respectable
1:10:28 um it's possible that that open AI goes
1:10:31 bankrupt by the end of the year
1:10:33 I can't imagine that Microsoft or anyone
1:10:38 else is going to let them go bankrupt
1:10:40 unless Microsoft has some poison pill
1:10:43 that says if they go bankrupt they own
1:10:45 it or some [ __ ] like that which is
1:10:47 possible depending on their investment
1:10:49 rights
1:10:52 I just don't see it happening investors
1:10:55 will come in it's it's too important a
1:10:56 technology
1:10:58 all all of this competition would not
1:11:00 exist had chat CPT not did what it done
1:11:04 what it did did what it done did what it
1:11:07 did
1:11:08 I'm tired
1:11:10 uh 10 billion over how long 10 billion
1:11:14 see I think most of the 10 billion from
1:11:16 Microsoft was in compute credits
1:11:19 so
1:11:21 I would imagine that they burnt through
1:11:24 a lot of that 10 billion dollars already
1:11:28 cost 700 Grand daily yeah yeah so so so
1:11:31 they're they're blowing through the cash
1:11:33 right every time we do one of our stupid
1:11:35 [ __ ]
1:11:37 write a poem that sounds like a cowboy
1:11:41 got a rock in his shoe every time we do
1:11:43 that it costs them a couple of bucks
1:11:45 like it ain't cheap to run these things
1:11:48 um they're too important they're they're
1:11:50 too important to let it die in my
1:11:52 opinion I like that just seems like a
1:11:55 doomsday kind of sensationalist
1:11:58 thing like if if listen if I haven't
1:12:02 read the article if there's a quote from
1:12:03 Sam Altman saying we're in trouble great
1:12:06 but if this is just one of those
1:12:07 speculative things that's probably just
1:12:09 someone trying to buy some time to build
1:12:11 a competitor to chat gbt
1:12:15 uh
1:12:16 uh yeah 700 Grand a day is a lot to
1:12:19 operate something
1:12:20 they'll cut back first quarter spending
1:12:22 don't expect much from AI yet
1:12:26 um this is a great intro to wait this is
1:12:28 great for introduction and profit taking
1:12:32 wait most tickers wait what this is
1:12:36 great for introduction and profit taking
1:12:38 most tickers 10 from highs
1:12:44 I don't quite know what you mean but I
1:12:46 kind of know what you mean yeah so
1:12:49 so yeah user just dropped off with
1:12:51 Chachi BT
1:12:54 things are down a bit
1:12:56 wait listen
1:13:01 if what we get to play with today is
1:13:03 this powerful
1:13:05 do you know the [ __ ] that they're
1:13:07 playing with
1:13:09 that we don't
1:13:10 get to see I mean Sam Altman said
1:13:13 there's something they developed at chat
1:13:15 GPT they will never release it's that
1:13:18 powerful
1:13:21 what is that
1:13:23 the CEO of stability AI says the image
1:13:26 generation tool that they're playing
1:13:27 with internally it's stability AI is
1:13:30 generating images at 60 frames a second
1:13:35 so it's doing 60 frame a second video
1:13:38 that's the tool they're playing with
1:13:39 inside open AI right now or in inside
1:13:42 stability AI right now the [ __ ] that
1:13:45 that we haven't seen yet is
1:13:48 like okay this stuff's not ready for
1:13:51 prime time okay great it's not
1:13:54 but you wanted to write some social
1:13:55 media copy it'll [ __ ] knock it out
1:13:59 it'll get you to 80 of what you need in
1:14:02 five minutes and then you can spend the
1:14:03 rest of the hour refining it
1:14:08 I don't know man I don't know I would I
1:14:10 would not
1:14:12 piss on the grave of generative AI just
1:14:15 yet just because there's a little
1:14:16 there's a little Summertime Blues
1:14:19 and and there's not as many users as
1:14:21 there were in the in the frothy
1:14:24 aftermath of it launching
1:14:29 um have you tried out
1:14:30 goblins.tools yet goblin.tools yet yeah
1:14:32 I have so I would show [ __ ] here but I
1:14:35 can't
1:14:36 unless my internet has gotten better now
1:14:39 it really hasn't look see
1:14:42 it working let me let me try one more
1:14:43 time
1:14:45 maybe they reset The Rounders because
1:14:47 people were complaining
1:14:49 no
1:14:50 no it's still shitty
1:14:52 um yeah goblin.tools is a to-do list
1:14:55 maker that you just put in a to do
1:14:59 and then you hit the magic wand button
1:15:01 and it will break it down like a full-on
1:15:04 project manager it's pretty amazing
1:15:07 um
1:15:08 it with my ADHD to-do lists give me
1:15:11 hives and goblins to-do lists really
1:15:14 give me hives because it's like you know
1:15:17 I just want to accomplish this it's like
1:15:18 here's the 73 steps you need to do to
1:15:21 accomplish that yeah I'm good I'll just
1:15:24 not get [ __ ] done
1:15:25 I'm going to write a book not get [ __ ]
1:15:28 done
1:15:31 the way of the the way of the add wizard
1:15:37 let's say DD people we get [ __ ] done
1:15:39 when we [ __ ] want to get it done and
1:15:42 it's always at the last minute
1:15:46 prefrontal cortex block of cheese from
1:15:49 here forward
1:15:50 nothing but net
1:15:56 I'll take a cross-section here a little
1:15:58 brain scan it's just like
1:16:05 great Bondo hairdo yeah I know it's the
1:16:09 it's it's that mud [ __ ] like like it's a
1:16:11 little teeny tin and you don't think
1:16:12 it's doing anything but then you put it
1:16:14 in and it is it's like [ __ ] Bondo
1:16:16 it's pretty funny thank you oh just
1:16:19 jumped me to the bottom hey dang it dang
1:16:21 it tick Jack why are you doing that to
1:16:23 me it's just makeup it's just rigm
1:16:25 friggin man let's rock them suck them
1:16:27 you son of a [ __ ]
1:16:33 I want to meet Mr Tick Tock and I want
1:16:35 to say when I report someone for copying
1:16:39 my [ __ ] account and trying to scam my
1:16:41 followers and I report that four times
1:16:44 why is that okay why are you just saying
1:16:47 yeah they're fine what are they doing to
1:16:49 you
1:16:50 yeah and then why did you jump me to the
1:16:53 bottom of the comments when I'm just
1:16:55 trying to scroll through them why why
1:16:56 why
1:16:58 someone programmed that
1:17:01 it should be fired right now
1:17:03 [Music]
1:17:04 all right I think I found my place oh
1:17:06 there was the Bondo in the hair comment
1:17:08 the Russian translation is you two can
1:17:11 make the Russian translation you can
1:17:14 make money watch every day that's pretty
1:17:17 good Tobias was shaking man
1:17:19 all right Luna stick good evening
1:17:24 if you hold down a comment there's an
1:17:26 option to translate he said AI will
1:17:28 bring manipulation yep got it I figured
1:17:31 it out
1:17:32 old guy figured out tick tock
1:17:35 stop the presses
1:17:37 want to stop the presses mean well see
1:17:39 in my day there were a thing called
1:17:41 newspapers and they had these things
1:17:44 called presses where they pressed ink
1:17:47 into paper
1:17:49 and then if there was breaking news
1:17:51 someone would yell stop the presses and
1:17:55 then they would literally stop the
1:17:57 presses
1:17:58 that's what it means
1:18:00 yeah [ __ ] Jen's ears with your God
1:18:03 damn it everything's there instantly all
1:18:06 the time
1:18:08 it's not how it used to be
1:18:13 dang it
1:18:15 sup buddy chatgpt4 welcome welcome
1:18:18 welcome Emilio's wife many dinosaurs
1:18:21 awesome there thank you welcome
1:18:23 everybody let's just joined
1:18:26 um I am definitely many questions behind
1:18:29 down below down in the below times more
1:18:31 than 100 questions behind so if it seems
1:18:35 like
1:18:37 I am just talking about random [ __ ]
1:18:41 that's why that's why because I'm trying
1:18:44 to answer all the questions
1:18:45 isn't that good is mid-journey not good
1:18:48 anymore it's not on the list it's on the
1:18:50 list
1:18:50 it's just not on this list
1:18:53 oh wait when was the last time you were
1:18:55 here it's that Buffalo okay so so here's
1:18:57 what here's what happened it's that
1:18:59 Buffalo mid Journey used to be on this
1:19:02 list but this list only had three chat
1:19:05 GPT things on it and now it's got seven
1:19:09 so so I made one page for chechi PT [ __ ]
1:19:15 and then I made another page for
1:19:17 prompts.chat the prompt book for images
1:19:21 and future pedia and then I put all the
1:19:23 image generating tools down at the
1:19:24 bottom here so I got mid Journey Adobe
1:19:27 Firefly Leonardo and clips uh
1:19:30 clipdrop.co which is stable diffusion
1:19:32 stability ai's little experimental
1:19:35 portal
1:19:37 um so yeah a good eye
1:19:39 uh no it did not go away mid journey is
1:19:43 still the [ __ ] and no one's caught it
1:19:45 yet and I don't think they will for a
1:19:47 while but stable diffusion XL is close
1:19:49 but it ain't there
1:19:52 um but yeah it's it's we we have a
1:19:55 situation where there's there's just too
1:19:57 much [ __ ] here but but we need it there
1:20:00 until further notice thank God I can
1:20:03 change my name back tomorrow what how
1:20:05 did you get stuck changing your name to
1:20:06 chat gpt4 and why can you not change it
1:20:09 back is this just a special tick tock
1:20:11 tick tock being [ __ ] special kind of
1:20:13 thing
1:20:15 Chad gbt4 changing to Claude Dr J coming
1:20:18 in with a solid one there Grim Reaper is
1:20:21 saying a bunch of random stuff in
1:20:23 Russian like why does chat GPT share
1:20:26 poison info yeah I saw that Winston
1:20:27 thanks
1:20:29 um just watch your verified video too
1:20:31 bad you didn't get the free crypto
1:20:33 thanks Gary yeah I know yeah Gary
1:20:35 ginsler thanks Gary thanks yeah you know
1:20:38 what I feel so much safer that I didn't
1:20:40 get that free money that Sam Altman
1:20:42 wanted to give me I I just it warms my
1:20:45 heart to know that I'm not going to be
1:20:46 ripped off by someone who wants to give
1:20:49 me money thank you you're the best
1:20:55 bitter you betcha
1:20:58 all right the chat GPT doesn't cure
1:21:00 cancer
1:21:02 not sure why he's using Russian I don't
1:21:04 know why he's using Russian maybe he's
1:21:06 just trying to see if I can figure out
1:21:07 how to translate [ __ ] but I figured it
1:21:09 out there I figured it out because I'm
1:21:12 like an old guy with tech chops but not
1:21:15 real technical chops but like I can
1:21:17 point at [ __ ] and click on it
1:21:20 you don't think that takes skill
1:21:24 it doesn't it's not the point that's not
1:21:27 the point
1:21:29 all right
1:21:30 good day Christian Martinez yes the
1:21:33 shrewd one heart Dr Jake heart heart
1:21:36 heart Joy heart heart
1:21:38 thanks for the hearts this live is great
1:21:40 no lag when you switched cameras oh
1:21:42 really
1:21:44 weird
1:21:45 huh
1:21:47 fine life is good many hearts
1:21:51 no lagging for All of Me
1:21:54 better here than in Colorado isn't that
1:21:56 [ __ ] awful I'm in a hotel hotel Wi-Fi
1:21:59 is notoriously shitty and the [ __ ] I buy
1:22:02 at home that I spend way too much money
1:22:04 on
1:22:05 is crappier than this
1:22:07 the hell
1:22:08 why am I chewing my cuticle
1:22:11 because it's [ __ ] tasty
1:22:13 stop picking on me weirdos
1:22:17 Zen AI oh I can't look at the internet
1:22:19 because the in what okay you know what
1:22:21 I'm gonna do I'm gonna restart my
1:22:23 machine that's what I'm gonna do because
1:22:24 that's gonna fix it
1:22:26 that's gonna fix it there is more than
1:22:29 one way to get the internet I could also
1:22:31 tether it to my phone couldn't I
1:22:34 let's let's restart it and see what what
1:22:36 happens
1:22:38 all right show me how to make money
1:22:40 online with AI you can make money with
1:22:42 chat
1:22:44 so Giovanni Hendrickson Henriksen
1:22:48 um
1:22:50 I have a point of view on this
1:22:54 and my point of view is that making
1:22:56 money with AI or making money with chat
1:22:59 gbt
1:23:01 is not
1:23:03 it ain't about the tool it ain't about
1:23:06 the tool
1:23:10 what do you mean by that Kyle what do
1:23:12 you mean it ain't about tool here's what
1:23:14 I mean about that
1:23:16 um
1:23:18 what did you say there show me how to
1:23:20 make money online with AI the answer to
1:23:23 that question
1:23:25 Photoshop beta failed to quit why okay
1:23:28 we force quit it
1:23:30 screw it we're out of there
1:23:31 [Music]
1:23:33 make money online with AI
1:23:37 you can replace AI with any other tool
1:23:40 and the answer is exactly the same make
1:23:42 money online with Photoshop make money
1:23:44 online with HTML make money online with
1:23:48 CSS make money online with project
1:23:50 management make money online with
1:23:52 shovels
1:23:55 they're all tools
1:23:58 how you make money
1:24:01 is you figure out a problem someone has
1:24:07 why is it not using my fingerprint what
1:24:10 just happened to my machine did one of
1:24:13 you [ __ ] hack me and screw up my
1:24:15 machine
1:24:16 I'll be pissed you're in here speaking
1:24:18 Russian
1:24:21 someone's probably hacking me
1:24:23 taking me down
1:24:25 all right that thing's back let's see if
1:24:27 let's see if uh
1:24:30 where's
1:24:33 here's clip drop let's see if we get
1:24:35 some some internet back
1:24:37 um so so how you make money with any
1:24:40 tool
1:24:41 is you figure out a problem someone has
1:24:43 you figure out what your skills are
1:24:45 you identify that problem you're like I
1:24:48 can fix that problem
1:24:50 and many times it's like I can fix that
1:24:53 problem in ten thousand dollars and if
1:24:56 you fix that problem good enough and
1:24:58 it's worth ten thousand dollars people
1:25:00 give you money they give you the ten
1:25:01 thousand dollars and then you go fix
1:25:03 that problem
1:25:04 and with this generative AI [ __ ] maybe
1:25:06 you can fix that problem faster for five
1:25:09 thousand dollars
1:25:11 so you can go into the market and you
1:25:12 can say hey
1:25:13 they're fixing that problem for ten
1:25:15 thousand dollars I'll give you double
1:25:18 the results for half the money but you
1:25:21 gotta sign today and then you just made
1:25:24 five thousand dollars and you use chat
1:25:26 GPT to do it faster than those people
1:25:29 doing it by hand
1:25:30 and then you're laughing all the way to
1:25:32 the bank and you're going I made money
1:25:34 with her
1:25:37 but you didn't start with Chacha Peppa
1:25:39 did you you start with the problem so
1:25:43 I know that's a smart ass answer but
1:25:44 that's actually the real answer and the
1:25:47 reason I make fun of people that say
1:25:48 You're Gonna attach if you take
1:25:50 is that
1:25:52 how they're making their money
1:25:55 is with fear
1:25:58 wait a minute someone has a secret for
1:26:00 how to make money with chat gbt I can't
1:26:02 [ __ ] figure it out
1:26:04 they're gonna give me a course
1:26:07 and it's only 299 dollars
1:26:11 per session amortized over 18 months
1:26:15 and if I sell that course I can make
1:26:18 money no you're making money selling
1:26:21 courses
1:26:23 that's the that's the oldest trick in
1:26:25 the book They're not making money with
1:26:27 Chachi Pate other than the fact that
1:26:29 they wrote the course for chatgpt about
1:26:31 how to sell chat gbt courses to make
1:26:33 money for Chachi see how circular it is
1:26:36 it's circular there's pyramid schemes
1:26:38 and there's circular marketing schemes
1:26:42 circular marketing
1:26:44 and if they got really clever they'd
1:26:46 make it a ah the bear gave me the dangly
1:26:49 lights I love the dangly lights
1:26:51 um if they were really clever they would
1:26:53 turn their circular marketing logic into
1:26:56 a multi-level marketing scheme and they
1:26:58 could have a pyramid scheme on top of
1:27:00 circular marketing logic and they would
1:27:03 really make money watch activity if
1:27:05 you're at the top of the pyramid
1:27:09 you see how bitter I am
1:27:15 you know why
1:27:16 because people like that make me [ __ ]
1:27:18 crazy
1:27:26 go generate some [ __ ] value in the
1:27:28 world and people give you money for it
1:27:32 there's no easy answer
1:27:35 none
1:27:37 none
1:27:39 you've gotta do the [ __ ] work you
1:27:42 know what I'm saying
1:27:43 you come in here you expect the easy
1:27:45 Road out
1:27:47 what are you gonna do
1:27:49 you gotta [ __ ] walk a mile
1:27:52 then you gotta walk another mile
1:27:54 then you gotta walk another mile
1:27:56 what have you done then
1:27:58 you walk three miles you're [ __ ]
1:28:00 thirsty and your feet hurt
1:28:01 you still haven't made a [ __ ] dollar
1:28:07 walk a thousand
1:28:09 figure out problems solve them get
1:28:12 literate with this [ __ ] because here's
1:28:14 the deal
1:28:15 let's say what's your name oh Giovanni
1:28:18 Hendrickson
1:28:20 let's say you're really good at project
1:28:21 management learn how to use these tools
1:28:24 to do project management better than
1:28:26 your competition
1:28:28 and you probably can
1:28:30 these things are ability amplifiers
1:28:32 what's the [ __ ] that you're good at that
1:28:35 you enjoy doing that you can talk about
1:28:39 and then figure out how to do it with
1:28:42 this cheaper better faster
1:28:44 you know that old adage you can have a
1:28:46 good faster cheap you could only have
1:28:48 two of the three
1:28:49 I think with these tools you can have
1:28:52 three of the three
1:28:53 and right now
1:28:55 in today's world
1:28:57 it you can use these tools to make money
1:29:00 and not even tell anyone you're using AI
1:29:04 you just walk in and go hey you know how
1:29:06 your competitors say you can have two of
1:29:08 the three good fast and cheap
1:29:10 I'll give you all three
1:29:14 what's that mean
1:29:15 it means I'll deliver what they deliver
1:29:19 twice as fast for half the money
1:29:26 try me try me for a month
1:29:28 see if I can't do it
1:29:30 and then do it and if you can do it then
1:29:32 do it again
1:29:33 and then do it again and then do it
1:29:35 again
1:29:36 and when they pick their jaw up off the
1:29:38 floor going how the [ __ ] did he do that
1:29:41 you'll be on to the next thing
1:29:43 and you will have made money which at
1:29:45 gbt my light just went dim it went dead
1:29:49 do I have a
1:29:51 I think I have a thing
1:29:53 I think I have a thing it doesn't really
1:29:54 matter
1:29:55 I gotta wrap this up anyway I'm tired
1:29:59 tired
1:30:01 [Music]
1:30:15 all right Green Acres is the place to be
1:30:20 hi living is uh something I see
1:30:26 anyone
1:30:27 Green Acres
1:30:35 [Music]
1:30:37 there see that I'm not shiny again I'm
1:30:40 all shiny and blue I'm like I'm like
1:30:43 corpse over here and
1:30:45 something over there okay
1:30:49 cures for cancer already exists but I'm
1:30:51 not a healthcare professional yeah
1:30:52 there's that new there's LK 99 and then
1:30:55 there's some other numbered like G gr
1:30:58 something or G something R 1066 or I
1:31:02 don't know what it is there's supposed
1:31:03 to be some pill that cures all the
1:31:05 cancers yeah listen
1:31:07 we're gonna get free energy we're gonna
1:31:09 have levitation we're gonna have flying
1:31:11 cars we're gonna have cures for cancer
1:31:13 we're gonna solve climate change we're
1:31:14 gonna just just
1:31:16 figure people
1:31:19 the world can't be made in a day but it
1:31:21 can be made in a year with these things
1:31:24 or not
1:31:26 all right let's see what time is it 10
1:31:29 30.
1:31:30 what am I doing with my life laughing my
1:31:33 ass off church lady well I'm not special
1:31:44 hmm
1:31:45 isn't that terrific isn't that terrific
1:31:48 so the machines are going to do your
1:31:50 work for you whatever happened to good
1:31:52 day's work for a good day's pay
1:31:53 everybody wants to get rich with chat
1:31:56 JPay yeah okay
1:31:57 oh wow wow
1:32:00 [Laughter]
1:32:03 Satan runs the internet everywhere
1:32:08 oh
1:32:09 [Music]
1:32:12 stable doodle that's what it was called
1:32:14 Emilio's life yes clip drop stable
1:32:16 doodle
1:32:17 clipdrop.co that's very good info thanks
1:32:19 yeah I wish I could show it to you
1:32:20 because it's cool oh wait what's going
1:32:22 on over at the yeah salon
1:32:25 everybody figuring it out oh there's my
1:32:27 little doggy
1:32:29 um
1:32:32 reload do I have internet oh
1:32:36 I might have internet
1:32:38 wait
1:32:39 it's really slow
1:32:42 no
1:32:45 no they still sucks it still sucks still
1:32:48 bad
1:32:49 it's still bad am I on the right Wi-Fi
1:32:51 yep
1:32:53 [Music]
1:32:55 hello from Bulgaria yuhi
1:32:58 yuhi I think that's right
1:33:01 hi
1:33:02 the Bears here awesome
1:33:06 Greta gloves the prompts.chat link you
1:33:08 shared has been very helpful thanks
1:33:10 enjoying and learning from the lives
1:33:12 that's awesome thank you oh look
1:33:14 clip drop just showed up
1:33:17 here's stable doodle all right
1:33:20 oh
1:33:21 the internet's working let's go play
1:33:24 hahaha
1:33:28 oh but I I actually granted thank you so
1:33:31 much for acknowledging that that helps
1:33:33 because
1:33:34 the whole point of me showing all that
1:33:36 [ __ ] and repeating it over and over on
1:33:38 these lives is that I think this stuff
1:33:40 helps so by you letting me know that it
1:33:42 helps that's actually great so thank you
1:33:44 for that okay so someone earlier asked
1:33:47 about why doesn't mid Journey make text
1:33:50 well and I explained why it doesn't do
1:33:52 it but
1:33:53 last night no two nights ago someone on
1:33:56 the live said but stable doodle does
1:33:58 good text and I'm like that can't be
1:34:00 true
1:34:01 but Auntie Jane I didn't think image
1:34:04 diffusion models could do text well
1:34:07 well Jimmy Okay so
1:34:10 here's what I did I went
1:34:13 you're gonna watch my
1:34:15 lovely illustration illustration skills
1:34:18 one of the reasons that I'm so [ __ ]
1:34:20 excited about mid-journey and things
1:34:21 like that is this is about the level of
1:34:24 my illustration skills
1:34:25 yeah
1:34:27 all right so
1:34:29 we have a
1:34:35 we have a sign in a window okay
1:34:38 and now we're going to type in a prompt
1:34:41 um
1:34:42 let's see
1:34:43 um
1:34:45 um
1:34:48 buy one get
1:34:51 one
1:34:53 free
1:34:55 sign in shop window
1:34:59 all right so hopefully this will say buy
1:35:01 one get one free
1:35:04 so let's see let's see what happens that
1:35:08 actually seemed like pretty peppy
1:35:09 response I think I got full on internet
1:35:11 here how exciting is that
1:35:16 now watch it'll just sit here for hours
1:35:19 I'm also thirsty and I ran out of water
1:35:21 what am I gonna do what am I gonna do
1:35:23 people
1:35:25 what's happening
1:35:28 I'm right near Broadway
1:35:32 I could go out and just sing at the top
1:35:34 of my lungs right now because you're in
1:35:35 New York City and people wouldn't even
1:35:37 bat a [ __ ] eye
1:35:39 here we come to save the day
1:35:43 I'd probably get Applause for that one
1:35:46 all right I don't know what's happening
1:35:47 with this
1:35:49 we'll move on to another question you
1:35:52 speak like a voice actor nearly Punk
1:35:54 well thank you very much oh who did
1:35:58 who deal was I loved it
1:36:00 hey everybody hi Kyle in New York City
1:36:01 hey bear what's happening
1:36:05 um I appreciate the the uh the voice
1:36:07 comment
1:36:08 Barry was Barry's last name my voice
1:36:11 teacher at Penn State would would
1:36:14 appreciate that comment
1:36:16 he taught me to took talk good words
1:36:19 with the good voice
1:36:22 talk with words in a good way use your
1:36:25 words Luke I am your father
1:36:31 we speak in tongues why do we speak in
1:36:34 tongues why not
1:36:38 um
1:36:39 if it seems like I'm getting Punchy
1:36:41 that's because I'm exhausted
1:36:43 [Laughter]
1:36:45 and my vocal cords feel like someone ran
1:36:48 a cheese grater over them
1:36:51 but when that happens I can talk nice
1:36:53 and low like this
1:36:56 welcome to the AI learning lab how can I
1:36:59 help you
1:37:00 well isn't this fantastic
1:37:03 aren't people that talk like that all
1:37:04 the time wild I used to play uh poker
1:37:07 with Patrick page who's a Broadway actor
1:37:09 and he just talks like that he's like
1:37:12 hey could you could you hand me the
1:37:14 chips you're like Jesus it's like every
1:37:16 time he talks it scared the [ __ ] out of
1:37:18 me
1:37:20 he's an amazing guy though
1:37:23 apparently clip drop is just not gonna
1:37:25 [ __ ] cooperate I'll come back to it
1:37:27 but whatever
1:37:30 what are you gonna do what are you gonna
1:37:31 do it's not my setup and even though my
1:37:34 setup at home is a little janky it works
1:37:36 it works my hotel set up here I got my
1:37:40 computer up on a pillow it's very sad
1:37:45 oh
1:37:50 all right stable doodle is fantastic
1:37:55 everyone's welcoming the bear the bear
1:37:58 came in got some hugs nice concache PT
1:38:02 think of something new is it creative
1:38:04 um it okay well hmm
1:38:08 thinks of something new in a similar way
1:38:11 that humans think of something new in
1:38:13 that
1:38:15 we're all large language models so the
1:38:19 the G in GPT stands for generative so
1:38:22 when you give Chachi BT a prompt
1:38:25 the thing it's giving to you
1:38:28 is original content it's generating word
1:38:32 by word a response to your prompt that
1:38:36 is just by math different than what
1:38:39 other people are going to get
1:38:42 um
1:38:43 if you prompted in a clever way it will
1:38:45 give you new ideas and you know like
1:38:49 like if you think about
1:38:52 how
1:38:54 when you listen when you listen to
1:38:57 rappers talk about how they think about
1:38:59 language and how they
1:39:01 [Music]
1:39:02 um
1:39:03 put phrases together and Rhymes together
1:39:05 and they figure out how to break up
1:39:07 language in ways that they can they can
1:39:09 make Rhymes out of words that don't
1:39:11 actually rhyme but if they you know
1:39:14 stack them in certain ways and then you
1:39:16 also hear them talk about well you know
1:39:18 one technique I is you know I'll combine
1:39:21 this concept with a completely you know
1:39:23 nonsensical concept and then I'll see
1:39:25 what comes out the other side you can do
1:39:27 that with these large language models
1:39:30 you can say
1:39:32 you know
1:39:33 um
1:39:35 take jungian archetypes and combine
1:39:39 those with
1:39:42 Latin names of fungi
1:39:46 and you know write me a story where all
1:39:51 of the characters are named after
1:39:53 mushrooms and you know the jungian
1:39:56 archetypes are put in there and it will
1:39:58 [ __ ] do that
1:40:00 has anyone done that before I don't know
1:40:03 probably not
1:40:07 is that creating something new well I
1:40:11 would say that you have created
1:40:13 something new
1:40:14 using a tool
1:40:17 that if you had enough time to research
1:40:21 all of the inputs that you just
1:40:23 described take all of the young Ian
1:40:25 archetypes and all of the Latin names
1:40:28 for fungi and turn that into a story
1:40:31 that matches the Star Wars plot
1:40:35 it will do that you could go research
1:40:38 all that [ __ ] you could do that for
1:40:39 yourself and it would probably take you
1:40:41 a year
1:40:43 and you can just say to these things go
1:40:45 [ __ ] do that
1:40:49 that's how creativity works right
1:40:52 but you as the human are the ones still
1:40:55 prompting it and goading it and figuring
1:40:58 out thinking creatively again I think
1:41:00 creative thinking is the skill
1:41:02 moving forward that is the valuable
1:41:05 skill
1:41:06 tactical execution
1:41:08 not as valuable creative thinking
1:41:10 strategy
1:41:12 curation
1:41:13 taste
1:41:19 so yeah I think it can be creative when
1:41:21 it's paired up with a creative human
1:41:24 ask it a shitty question it will give
1:41:26 you shitty answers ask it a really
1:41:28 creative question structured in a really
1:41:30 creative way
1:41:32 it'll give you stuff that no one's ever
1:41:34 [ __ ] seen before and it'll do it like
1:41:35 that
1:41:37 so yes I think it can do something new
1:41:42 we are large language models
1:41:45 that is a large language model
1:41:48 in fact architecturally how the neural
1:41:51 net Nets work
1:41:53 why they are the way they are they tried
1:41:56 to mimic our neurons
1:42:02 now our neurons do stuff in a much more
1:42:05 sophisticated way than these things do
1:42:07 and there's many more of them but
1:42:11 I know what's going on
1:42:16 you know
1:42:17 is anything a human does truly creative
1:42:20 isn't it all derivative from [ __ ] you've
1:42:22 learned in your life
1:42:24 yeah but that's human okay
1:42:27 it's language it's words words words
1:42:30 words words
1:42:32 words words
1:42:39 clip drop you disappoint me
1:42:43 my hotel internet has disappointed me
1:42:47 bah humbug I'm gonna try again
1:42:50 I'm doing a window
1:42:53 oh no
1:42:55 oh let me do it look I just did creative
1:43:00 all right it's a room now it's a room
1:43:05 I don't know what I'm doing
1:43:08 oh yeah this could still work
1:43:11 [Music]
1:43:16 help wanted
1:43:21 sign in shop window
1:43:25 all right we'll see if it does its thing
1:43:28 I have a new car oh not me law
1:43:31 huh what
1:43:33 what about when it was asking me which
1:43:35 boxes have traffic lights
1:43:38 I don't know what you mean is that some
1:43:41 sort of inside reference to
1:43:42 Schrodinger's cat is that a quantum
1:43:44 physics joke
1:43:46 or am I just a dum-dum I might be a
1:43:49 dum-dum it's possible I'm a dum dum am I
1:43:52 a dumb dumb
1:43:57 there's 100 people in here and I'm
1:43:59 whacking my stupid [ __ ] head with a
1:44:00 bottle all right
1:44:02 um thanks that was helpful explanation
1:44:04 on image generation oh by the way I'm
1:44:06 very far behind on my questions here
1:44:08 very far behind because I was talking
1:44:10 about image generation like 45 minutes
1:44:12 ago so
1:44:14 if I'm not answering your question
1:44:16 that's because
1:44:18 now I'm an old man I'm just an old man
1:44:23 I do my best
1:44:25 I get on The Tick Tock and I do my best
1:44:31 I click on interface elements and I've
1:44:33 read the manual the manual is it's dry
1:44:37 reading but it's very informative
1:44:39 so I've read The Tick Tock manual and uh
1:44:42 it's a it's uh you can get it at the
1:44:44 library uh I mine's slightly out of date
1:44:47 my mine was from 1982 was the was The
1:44:50 Tick Tock manual that I'm I'm it was it
1:44:52 was the Tic Tac manual that was
1:44:55 that was how to make Tic Tac candy
1:44:58 um
1:44:59 right anyway
1:45:07 does NASA runs their AI on d-wave
1:45:11 quantum computers maybe I don't know
1:45:14 I ain't been to NASA
1:45:16 I don't have access to their Quantum
1:45:18 Computing lab and and and I ain't said
1:45:21 to them hey NASA are you running your
1:45:24 Quantum compute are you making a few
1:45:27 reversed engineered alien technology
1:45:29 with your the rocket Parts you got off
1:45:34 the moon from the moon base that you
1:45:36 didn't go to the moon but there was
1:45:38 rocket Moon based alien people on the
1:45:41 moon that you didn't go to I'm confused
1:45:44 now could you use your quantum computer
1:45:46 to help me figure out which conspiracy
1:45:49 theory is real or not because it's the
1:45:51 polar ice caps have the hollow Earth and
1:45:54 but it's it's it's a flat Hollow Earth
1:45:57 so wait if it's a flat Earth with a
1:46:00 Hollow Middle is is Earth a donut
1:46:03 I think or it's a donut in the Moon is
1:46:06 cheese with alien technology that we
1:46:09 didn't go to from the Mars with the face
1:46:12 and the pyramid on Mars
1:46:15 I'm so confused
1:46:19 I
1:46:21 I try to believe it all and but then
1:46:25 um
1:46:26 I think the Earth's donut I think I read
1:46:28 that I think I read that on the internet
1:46:30 it was a quantum computer that figured
1:46:32 out the Earth was a donut
1:46:34 yeah and but all the pilots are like the
1:46:37 pilots that flew over it that don't see
1:46:39 a donut that that's a lie they trained
1:46:42 them they took them to the base in the
1:46:46 four they call it do you think it's Area
1:46:48 51 but it's not it's S4 and they took
1:46:51 the pilots there and they said listen
1:46:52 when you fly
1:46:54 um we're gonna give you video to put on
1:46:56 your phone just don't tell them it's a
1:46:59 donut
1:47:00 because it's the hollow Flat Earth it's
1:47:02 the hollow Flat Earth with the waterfall
1:47:04 in the ice ring
1:47:11 yes so I don't know if NASA is using
1:47:14 d-wave quantum computers for their
1:47:17 calculations to return to the Moon that
1:47:19 they didn't go to in the first place to
1:47:22 get the aliens that are the reason they
1:47:24 left and never went back because they're
1:47:25 afraid yes I don't know I don't know I
1:47:29 don't know I don't know we just talk
1:47:31 about AI here we don't talk about all
1:47:33 that stuff I don't talk about NASA I'm
1:47:35 talking about Neville
1:47:37 Kai is teaching me Spanish at the moment
1:47:39 oh that's awesome Chachi BT well
1:47:42 businesses eventually create their new
1:47:44 local
1:47:45 they're now local chat GPT that focuses
1:47:47 on content that they train yes I own
1:47:49 China
1:47:50 that to me feels like the killer app for
1:47:53 all these things
1:47:55 so right now these things are remarkable
1:47:57 because they're kind of reflections of
1:47:59 humanity right but they're not
1:48:01 Reflections of My Humanity they're
1:48:03 reflections of all humanity and they're
1:48:05 not reflections of my business they're
1:48:07 reflections of all businesses
1:48:09 so when I can just point it at my data
1:48:11 and say
1:48:13 tell me more about this thing I created
1:48:15 tell me more about the person that I am
1:48:18 and what I said out in the world I think
1:48:20 that becomes the killer app I think I
1:48:22 really do I think that's
1:48:25 that combined with Automation and
1:48:27 autonomous agents
1:48:29 that's when it starts to get really fun
1:48:31 and probably skip up the hair just the
1:48:34 glue just broke look at that that's a
1:48:36 [ __ ] firework display on top of my
1:48:38 giant aging bald [ __ ] head all right
1:48:42 [Music]
1:48:44 um so yes Kyle I how do I get barred to
1:48:48 review and analyze a PowerPoint which
1:48:50 I've uploaded on Google Drive
1:48:53 you don't write now you wait basically
1:48:57 um
1:48:58 if you put the PowerPoint in PDF
1:49:05 so Bard can't read Bard can only read
1:49:08 images so
1:49:11 so so this is [ __ ] up okay
1:49:15 all right
1:49:17 hang on
1:49:27 okay I learned I learned this stupid and
1:49:30 ancient so you don't have to
1:49:34 okay here's what you can do so how do I
1:49:37 get bored okay
1:49:40 so you've got a PowerPoint presentation
1:49:42 right yeah okay you got a PowerPoint
1:49:44 presentation
1:49:45 what can you do with that
1:49:55 if you can get the your uh you turn
1:49:57 first you turn it into a PDF
1:50:00 so you download it as a PDF
1:50:04 you could upload it to code interpreter
1:50:06 and have it extract the text out of it
1:50:08 at Chachi BT
1:50:11 you can upload the PDF wholesale to
1:50:13 Claude cloud.ai and it will probably
1:50:16 read it it won't be able to read the
1:50:18 images
1:50:20 being
1:50:22 I don't think you can do [ __ ] with it
1:50:25 barred if you took screenshots of the
1:50:30 PowerPoint and put those
1:50:36 how would you do that
1:50:40 like so so Bard can read images
1:50:45 um Bing can read images Claude can read
1:50:47 files
1:50:50 poke and read files
1:50:55 code interpreter and some plugins can
1:50:57 read files at chat.openai.com
1:51:01 I don't think I answered your question
1:51:02 well basically turn it into a PDF and
1:51:06 upload it to Claude that's your answer
1:51:09 don't worry about all the [ __ ] I said
1:51:10 just
1:51:12 turn it into a PDF upload it to cloud
1:51:14 and start interacting with it
1:51:17 and Bard will
1:51:19 when when they roll barred into the
1:51:21 office suite of Google then then it'll
1:51:23 work but it just doesn't now
1:51:27 and I feel like but Kyle isn't that
1:51:29 stupid yes
1:51:32 I I think it's stupid that I have to
1:51:33 have seven things on this screen right
1:51:35 now I think that's really stupid
1:51:38 but that's where we are
1:51:39 so
1:51:42 welcome to the early days this is I mean
1:51:45 this is what it was like in the early
1:51:46 days of the web it was like you know
1:51:48 nobody can figure this [ __ ] out because
1:51:51 all the technology sucked you know we
1:51:53 were like wait is it this is bad right
1:51:56 this is bad
1:51:58 yeah this is bad but companies are still
1:52:00 willing to pay us to make these stupid
1:52:02 websites yeah
1:52:03 a lot of money
1:52:05 more than we ever imagined all right
1:52:08 same [ __ ] here
1:52:11 and then five years from now the world
1:52:14 will be transformed because of this [ __ ]
1:52:15 and if you're one of the ones that took
1:52:17 some time right now to get your [ __ ]
1:52:20 head around this stuff and got a little
1:52:21 AI literate you made good shape you're
1:52:24 gonna be in good shape you'll be like
1:52:26 yeah man I'll tell you what in the early
1:52:28 days man
1:52:30 there was this tick tocker his name is
1:52:32 it doesn't matter what his name was but
1:52:34 he was old as [ __ ] man and he was just
1:52:36 talking like you just like the world
1:52:37 wide web and then I was like ah [ __ ] you
1:52:40 man and he's like you're just a [ __ ]
1:52:41 old man it's like Elder talk you're just
1:52:43 [ __ ] like talking to people and
1:52:44 anyway he was like oh man and he didn't
1:52:47 know how to use tick tock or nothing he
1:52:49 like read all the comments he was like
1:52:51 dumb anyway so I started looking at chat
1:52:54 gbt it's like let's chat to EPT it
1:52:56 doesn't matter right now but I started
1:52:57 playing with this thing I got pretty
1:52:58 good at it man and it and then that's
1:53:01 why that's how I became a billionaire
1:53:06 you know why
1:53:07 because
1:53:10 all right
1:53:13 that was a lot of a lot of wind up to
1:53:15 get to a horrible punch line wasn't it
1:53:17 all right I've created some new and
1:53:18 interesting math by mining information
1:53:20 so by Leading chat GPT on that's
1:53:24 interesting on a path you can do some
1:53:25 stuff with known information okay
1:53:27 Winston yeah so what Winston is talking
1:53:29 about here is really true
1:53:31 one of the things that I'm increasingly
1:53:33 increasingly impressed with with these
1:53:35 tools is if you give it something it's
1:53:38 really good at analyzing it or learning
1:53:40 it like give Claude like documentation
1:53:44 for some piece of software you wanna
1:53:47 or like some programming language you
1:53:49 want to
1:53:50 start programming in it'll learn the
1:53:53 whole [ __ ] thing and then just start
1:53:54 programming for you
1:53:56 um
1:53:57 it'll analyze the screenplay and tell
1:53:59 you what sucks about it it'll take
1:54:01 whatever math stuff Winston put in there
1:54:03 and combine it together and give you
1:54:05 some new math
1:54:07 so yes it can be creative how's it
1:54:10 creative you stick a creative human
1:54:12 being next to it who who is conceptually
1:54:14 figured out oh these things can ingest
1:54:17 anything and they know everything and if
1:54:20 I'm clever about how I combine that data
1:54:23 and ask the question I can get some
1:54:25 remarkable [ __ ] that's the art
1:54:28 that's really cool
1:54:32 Arcanum are there any open source ones
1:54:35 you can run locally lots of them
1:54:38 um I don't know them uh check out a
1:54:40 project called GPT for all
1:54:43 and it allows you to install it on Mac
1:54:47 windows and Linux and then you can pop
1:54:50 in whatever open source large language
1:54:52 model you want it gives you a basically
1:54:54 a running environment where you can run
1:54:56 large language models inside it
1:54:58 and uh that's pretty good but I I'm not
1:55:01 following the open source world because
1:55:03 it's moving too fast and it requires too
1:55:05 much of my brain to make work so I just
1:55:07 I go like I think open source is really
1:55:10 good but you got to be really geeky and
1:55:12 the interface is always really bad and
1:55:14 you got to be good at command line and
1:55:16 then you got to run terminal and then
1:55:18 you got to understand what Jupiter
1:55:19 notebooks are and then you got to run
1:55:20 Google collab and then you got to
1:55:22 probably get good at Python and then
1:55:24 when it breaks you got to realize oh you
1:55:25 don't have the library installed and
1:55:27 then you don't have the library
1:55:28 installed you're like [ __ ]
1:55:31 that's my experience with open source
1:55:36 didn't show up did he call no the guy
1:55:38 that didn't show up today
1:55:40 didn't call
1:55:43 didn't respond to the email from the
1:55:45 woman that was hosting the meeting that
1:55:47 invited us on his behalf
1:55:51 um didn't email that response so
1:55:58 um
1:56:00 I don't I don't know about any of you
1:56:02 like I got [ __ ] anxiety when it comes
1:56:04 to like I show up early for meetings I I
1:56:07 like
1:56:09 I have I have whiffed on two meetings in
1:56:12 my 11-year career at storyline and I can
1:56:15 remember both of them like the [ __ ]
1:56:18 like I apologized 40 [ __ ] times
1:56:22 so I don't get it but whatever
1:56:26 why would they need to discover cancer
1:56:27 cure when we have found many of them but
1:56:30 doctors still ignore them doctors don't
1:56:32 ignore them
1:56:34 the giant corporations do not fund them
1:56:40 you can't make money on a cure
1:56:45 it's America it's about capitalism it's
1:56:48 about dollars we like dollars
1:56:51 you gotta you it's about you've got
1:56:53 shareholder value I don't think you
1:56:55 understand this
1:56:56 you've got shareholder value if you cure
1:56:58 cancer
1:57:00 you can't provide shareholder value if
1:57:03 they're not needing drugs year after
1:57:06 year I don't silly
1:57:09 silly
1:57:10 why would you cure cancer instead
1:57:13 dollars
1:57:17 and the ordinary people's run wait the
1:57:20 ordinary people's Run for the AI is
1:57:23 actually the AI learning about the mass
1:57:25 people of the world
1:57:27 uh uh okay okay yeah yeah okay what you
1:57:32 said that's right sure all right do you
1:57:34 head back to Denver in the morning new
1:57:37 we have another meeting tomorrow that
1:57:40 I'm fairly confident that person's gonna
1:57:42 show up same host person
1:57:47 um
1:57:48 and that's a long meeting and that one
1:57:51 I'm really excited about and then we're
1:57:54 having dinner with
1:57:56 both of them
1:57:58 at new one Joe in Koreatown I'm very
1:58:02 excited about
1:58:04 and then we go to Betty buy and we get
1:58:06 up bright and early we leave at like six
1:58:08 in the morning on Wednesday
1:58:11 um he wants to know what God we wish for
1:58:13 I don't know who he is but yes
1:58:16 everyone wants to know what God we wish
1:58:18 for except
1:58:22 government will step in the worst case
1:58:24 they won't let China get
1:58:26 too far a jump on us
1:58:34 um
1:58:35 [Music]
1:58:37 I don't know possibly that's way above
1:58:39 my pay grade I don't understand
1:58:41 geopolitical
1:58:44 currents very well so I'm not going to
1:58:47 comment on that in 1998 I rooted
1:58:51 Korea Telecom
1:58:53 and flooded Google Yahoo and lycos
1:58:57 offline using a Venom script oh cool
1:59:02 in uh
1:59:03 in 1980
1:59:08 when would this have been like 80 80 or
1:59:10 81
1:59:12 my buddy Scott and I was in we were in
1:59:15 high school
1:59:17 and he got a modem
1:59:19 and you could get lists of computers you
1:59:22 could call up
1:59:23 that
1:59:25 that didn't have any password protection
1:59:27 on them because that's how it was back
1:59:29 then because not many people had mowed
1:59:31 them so
1:59:32 you know the hackers would would you
1:59:34 know they would write scripts to go find
1:59:35 all the computers that didn't have any
1:59:37 controls on them and so I remember we we
1:59:41 got this list of of all these phone
1:59:43 numbers and we found a phone number of a
1:59:46 the control system of a mall in North
1:59:48 Carolina and so we log into this thing
1:59:51 and we turned the heat up to 99 and we
1:59:53 reversed the escalators and we were like
1:59:56 we're hackers
1:59:58 [Laughter]
2:00:00 oh man this is some good old days that's
2:00:05 awesome
2:00:07 uh 99 Plus messages by the way if you're
2:00:10 new here I read all the [ __ ]
2:00:12 questions I get very far behind I've
2:00:14 been doing this for like two hours now
2:00:15 my voice is shot I'm exhausted as [ __ ]
2:00:18 I'm Punchy as [ __ ] and I am behind on
2:00:21 your questions so I'm probably not going
2:00:23 to get to your questions but you can
2:00:24 keep asking them because I do try to get
2:00:26 to all of them
2:00:27 um but anyway you should follow me and
2:00:29 then that way when I do these lives I
2:00:31 try to do these every night
2:00:33 um you'll get notified and then you know
2:00:37 yeah whatever that was
2:00:40 Google NASA was so mad for taking a huge
2:00:43 chunk of the world's neural network
2:00:46 no that just seems exhausting
2:00:50 me to the bottom oh
2:00:52 who can tick tock
2:00:55 why don't you do that to me amazing
2:00:58 cranky
2:01:01 I feel like I'm talking like muscles now
2:01:03 that's all yeah
2:01:07 we're Doug and Wendy Winer we've got
2:01:09 diverticulitis yeah
2:01:14 gen xers will get that the rest of you
2:01:15 like who's Doug and Wendy weiner what's
2:01:18 diverticulitis you're stupid old man
2:01:21 why are you old are you old were you on
2:01:25 Tick Tock if you're old
2:01:27 you big fatty you big fat Rosie
2:01:30 O'Donnell looking
2:01:32 fat Alec Baldwin looking [ __ ]
2:01:34 why why are you old
2:01:37 you shouldn't be old shouldn't old
2:01:39 people just stop
2:01:40 [Laughter]
2:01:43 wait be careful the website looks like a
2:01:45 joke but it's not
2:01:46 which one
2:01:49 which one was the one that looked like a
2:01:50 joke but it's not
2:01:52 hello Uncle Kyle hello Mr drecker or Mr
2:01:55 Decker Troy Decker Troy Decker good to
2:01:59 meet you good day good day good day to
2:02:01 you sir good day brother right there
2:02:04 rather to you to you too
2:02:06 and do you choose that oh well um
2:02:12 fancy a game snooker yeah
2:02:17 I really think big chat is terrible
2:02:19 compared to GPT 3.5 fuzz I couldn't
2:02:22 agree more
2:02:24 Microsoft lobotomized Sydney which was
2:02:27 the project name for gpt4 inside Bing
2:02:31 after the New York Times revealed that
2:02:34 Sydney was batshit crazy so Microsoft
2:02:37 made Bing chat a little bit of a dum-dum
2:02:41 and they fixed it yet so yes I would
2:02:43 agree with you
2:02:45 but I do like being being connected to
2:02:48 the internet is convenient and I do use
2:02:50 it sometimes and it makes images which
2:02:52 sometimes if I need something in a pinch
2:02:53 I'll just knock it out
2:02:55 my CIS ADHD gave me her cover letter to
2:02:58 edit that was 3 000 words she was on a
2:03:02 roll I guess wow yeah sorry about that
2:03:05 that's that here that's one of the
2:03:07 downsides of these things is I've done
2:03:09 this before where I'll start writing
2:03:11 something and I'm like oh that's a good
2:03:12 paragraph and I'll dump that into a word
2:03:14 doc and then I'm like wait rewrite that
2:03:16 paragraph and then I'll dump the the new
2:03:18 improved version of it over there and
2:03:20 before I know it I've got a document
2:03:22 that's got like 17 versions of the same
2:03:24 paragraph and I'm trying to figure out
2:03:25 which one's best and it ends up making a
2:03:28 nightmare of Hell of yeah editing so
2:03:31 gotta be careful with this long time to
2:03:34 wait for my jokes to hit in real time
2:03:36 thanks Kyle and uh sorry about that AI
2:03:40 learn and get used to it tick tock is
2:03:42 about
2:03:43 10 cent app Chinese don't give in a [ __ ]
2:03:47 about IP or copyright
2:03:50 yeah we knew that
2:03:52 all right
2:03:55 I don't I yes
2:04:00 I don't what what am I supposed to do
2:04:03 about that I can't do anything about
2:04:05 that
2:04:06 except make money
2:04:09 controversial thought the internet is
2:04:11 already alive and AI is reshaping the
2:04:14 world bit by bit I don't know if that's
2:04:15 that controversial
2:04:17 I think the internet is alive
2:04:20 I mean the internet's not alive the
2:04:22 internet's a protocol
2:04:24 but it's at the center of all of our
2:04:26 lives or a lot of our lives and then
2:04:28 this stuff's going to amplify it
2:04:30 dramatically so yeah so I think this I
2:04:33 think what this does is it accelerates
2:04:36 all of that [ __ ] and it
2:04:40 it hides all the complexity
2:04:42 and reveals all of the
2:04:45 all of what's underneath the complexity
2:04:49 um and it does it really fast so
2:04:52 um it used to take half an hour to
2:04:54 download one song you kids are lucky
2:04:56 exactly Ace dogs got it Ace dog 90. Ace
2:04:59 dog 90 knows what's going down
2:05:03 I remember when we were using dial-up
2:05:06 modems to call into Mall Control Systems
2:05:09 you could watch the the [ __ ] type across
2:05:13 your screen which is funny when you look
2:05:16 at gpt4 right now it's not quite that
2:05:18 slow but like if the server is busy you
2:05:20 watch it typing across your screen
2:05:22 we're gonna look back at that and go
2:05:24 remember when chat GPT you know wasn't
2:05:27 just a million words in a third of a
2:05:30 second remember when you actually had to
2:05:32 wait for it to write a paragraph man
2:05:34 that was torture
2:05:38 that's coming
2:05:41 um
2:05:42 oh the the the Brontosaurus is the AI
2:05:46 large language model mascot is it that's
2:05:48 cool
2:05:50 those are not tiny dinos they're
2:05:52 tortures
2:05:53 or the tortures why do I not know this
2:05:56 Amelia's lifetime check 10 26 thank you
2:06:00 oh it's now it's 11 11. oh that's why
2:06:02 you put that in there so I know how far
2:06:03 I'm behind I am
2:06:06 um
2:06:07 40 minutes wait more than 40 minutes 45
2:06:10 minutes behind
2:06:11 all right I think I'm gonna scroll to
2:06:13 the Bottom now I know Russian didn't see
2:06:15 it come into my feed that was earlier is
2:06:18 that a new shirt Mr Kyle it's yeah
2:06:20 I just I haven't worn it because it's
2:06:22 got a lot of [ __ ] on it but it was the
2:06:23 only thing I had in my bag I think Gary
2:06:26 Gensler was a decent guy when he was
2:06:28 teaching me about cryptocurrencies at
2:06:31 MIT he oh I used to think Gary Gensler
2:06:34 was a decent guy yeah I mean I'm sure
2:06:36 he's a smart guy
2:06:38 but you know he's trying to protect the
2:06:41 old yield Treasury and you know not
2:06:44 compete with
2:06:46 um technology he can't compete with so
2:06:50 I don't know but it pisses me off
2:06:51 because I didn't get my free goddamn
2:06:53 crypto
2:06:55 why don't you get 5G internet I have 5G
2:06:58 internet because I don't know because
2:06:59 I'm a little lazy it doesn't I I just
2:07:02 it's I can't be bothered
2:07:06 you have no proof that I have your
2:07:09 MacBook all right I'm scrolling to the
2:07:11 bottom I'm jumping a bunch of questions
2:07:12 I know I just jumped a bunch of people's
2:07:13 questions I'm going all the way to the
2:07:15 bottom holy crap was I behind okay now
2:07:17 I'm at the bottom and I'm going to
2:07:19 scroll up just a little bit and then I'm
2:07:20 gonna do a handful of more questions and
2:07:22 I gotta go to bed look at my eyes
2:07:25 they're all puffy and tired they're
2:07:28 tired I'm like
2:07:30 [Music]
2:07:35 I need water
2:07:40 the Moon is a satellite made out of
2:07:42 cheese
2:07:43 for upper Dimensions the Earth is a
2:07:46 toroid and also flat appearing spherical
2:07:49 from here
2:07:50 with a donut in the middle a hole a
2:07:53 hollow Earth if you will
2:07:55 that's where all the aliens came from
2:07:57 except for the ones that came from
2:07:59 interdimensional space
2:08:01 but the other ones came from the inside
2:08:02 of the donut or the what we would call
2:08:05 The Hollow Earth if the Earth was fear
2:08:06 but we no it's not it's flat and so it
2:08:09 is a toroid and it
2:08:11 is a donut shaped a donut Guyana of the
2:08:14 name would taste is sweet
2:08:18 Cinnabon
2:08:20 would be sweeter
2:08:24 I think you
2:08:27 ATR too many dollars
2:08:31 huh
2:08:35 I think you
2:08:37 ATR to many dollars
2:08:41 anyone
2:08:44 know what beer maker is talking about I
2:08:46 don't know what you mean beer maker that
2:08:48 does not seem like words
2:08:51 anyone who just signed in welcome to the
2:08:53 man behind the curtain
2:08:56 exactly NASA is Disney Disney is NASA
2:09:00 and Coca-Cola you think it's an
2:09:02 enjoyable soft drink oh no oh no
2:09:07 hmm
2:09:10 all right moving on well I'm an ex-evil
2:09:15 hacker I own NASA servers they are all
2:09:17 the world's leaders in AI oh they are
2:09:19 the world's leaders in AI all right
2:09:21 that's cool listen I maybe I like I said
2:09:26 I'm I'm not so know so much about the
2:09:29 quantum computers other than they got
2:09:30 qubits and I used to play Qbert anybody
2:09:33 used to play Qbert
2:09:35 kubert was run on a super computer one
2:09:38 day
2:09:39 and you put a quarter in it and then you
2:09:41 jumped
2:09:44 no anyone
2:09:47 I like your bald head thank you it's
2:09:50 lovely so I put this part of the toupee
2:09:53 on first and then I lift it up and I
2:09:57 have another toupee that I put on the
2:09:58 sides and this I try I got a toupee that
2:10:01 I tried to make it look like I have
2:10:03 wisdom this is what they would call a
2:10:05 wisdom patch this is a wisdom to pay
2:10:07 and then I put on well this is three
2:10:10 components of a toupee I put them on in
2:10:12 random order and I tried to make it so
2:10:14 the hairs don't align so they all go in
2:10:16 different ways and then I lift that up
2:10:18 and I put on this thing and I'm
2:10:20 completely bald in the back of my head
2:10:23 it's a look it's a look I like I've I've
2:10:26 developed it over years so thank you I I
2:10:30 take great pride in my
2:10:32 artificial encumbrance on
2:10:36 what is considered by some ahead yes
2:10:39 all right I don't understand code
2:10:41 interpreter I know if if
2:10:45 I could show it to you if I could but I
2:10:48 don't think I can
2:10:49 let's see if we can
2:10:54 if I can get to chat gpk
2:10:59 no I can't it's not working
2:11:04 I'm trying to think of the best way to
2:11:06 get your head around code interpreter
2:11:10 um
2:11:13 actually here's here's is that a way
2:11:16 hmm
2:11:23 go to
2:11:24 so if you're trying to understand code
2:11:26 interpreter if you have access to it
2:11:28 there's a site called kaggle k-a-g-g-l-e
2:11:32 and at kaggle you can get data sets I
2:11:35 don't know [ __ ] about data analysis but
2:11:37 you can go there and you can get like
2:11:39 Spotify top 200 song data or
2:11:42 Healthcare in Cincinnati or crimes in
2:11:47 your city or whatever there's just
2:11:50 it's just a site full of data sets
2:11:53 so go there and just sort of browse the
2:11:56 data sets and find something that's
2:11:57 personally interesting to you
2:12:00 and then you download the what it'll
2:12:02 download as a zip file and like you
2:12:05 don't know what's in the zip file you
2:12:06 could read
2:12:07 you could read on kaggle what's in it
2:12:09 but I don't read I don't read the
2:12:11 manuals that's a Gen X thing and an add
2:12:13 thing and I just I'm a rebel man I don't
2:12:16 [ __ ] read manuals [ __ ] you with your
2:12:18 manuals my son reads manuals I yell at
2:12:21 him all the time I'm like no we don't
2:12:24 read manuals in this family have I not
2:12:26 taught you better and obviously I have
2:12:29 failed as a father
2:12:31 anyway so don't read the manuals just
2:12:34 download the archive you got a zip file
2:12:36 now so you go to code interpreter
2:12:38 and upload that zip file and then just
2:12:40 say to it please examine the data inside
2:12:43 this file
2:12:46 um
2:12:48 you know understand the nature of the
2:12:51 data analyze it for anything that you
2:12:55 think is interesting and tell me what
2:12:57 you discover and then just just watch it
2:13:00 and it will just start prompting itself
2:13:03 it will give itself a list of
2:13:04 instructions
2:13:05 and it will reveal something about that
2:13:07 data and it might at some point ask you
2:13:09 questions what would you like to do with
2:13:11 this data and then you can say things
2:13:12 like I don't know I want you to find
2:13:14 what you think is interesting in it and
2:13:16 tell me
2:13:17 and just let it go and then when it
2:13:20 makes those little boxes it says working
2:13:21 you can open those and you can like
2:13:23 watch the code that it's writing it's
2:13:25 actually writing and executing python
2:13:27 code
2:13:28 so it's remarkable so just go get a data
2:13:31 set and just ask it to do [ __ ] and then
2:13:33 the once you kind of understand what's
2:13:35 in the data then you can say things like
2:13:37 now I want you to visualize the most
2:13:39 interesting Trends in the data and it
2:13:41 will write you charts and graphs
2:13:43 and and
2:13:45 here's the deal code interpreter if if
2:13:48 you're in business and you want to
2:13:49 understand some data you don't need a
2:13:51 data analyst now you can just go ask it
2:13:54 to do that [ __ ]
2:13:56 and
2:13:58 but but the way I think about code
2:14:01 interpreter is this right now it only
2:14:03 does Python and it only does like a
2:14:05 handful of file types and right
2:14:08 but it's kind of taking on its own life
2:14:12 like you give it kind of a goal and it's
2:14:14 like well let me go open up this ZIP
2:14:16 file and it writes code to do that and
2:14:17 it opens it up and says well it appears
2:14:19 there's two csvs in here and a Json file
2:14:21 would you like me to examine those sure
2:14:24 and it goes and examine those look what
2:14:26 I found in this CSV look what I found in
2:14:28 this one it appears that these two are
2:14:30 about crime statistics and the Json is
2:14:32 labels about but and let's just go do
2:14:35 that so if I just sort of play that out
2:14:38 three years in the future now it can do
2:14:41 all programming languages it can take in
2:14:43 all sorts of files and it can accomplish
2:14:45 anything and it'll be connected to the
2:14:47 internet and it can write its own code
2:14:49 and it'll know every API out there and
2:14:52 it'll be able to get any data source at
2:14:54 once so you won't have to go to kaggle
2:14:56 and find a stupid [ __ ] CSV file it'll
2:14:58 just go find all the data so so think of
2:15:02 code interpreter right now is just kind
2:15:04 of a Looking Glass into what's coming
2:15:07 that that's kind of how I'm about it
2:15:09 right now but there I I guarantee there
2:15:12 are people using it in incredibly
2:15:14 powerful ways I'm just not one of them
2:15:17 I have a Blog can I use chat gbt to
2:15:20 write blogs oh [ __ ] yes it's really good
2:15:23 at that go to uh chat GPT pay for pay
2:15:26 the 20 bucks for gpt4 because it's just
2:15:28 so much better
2:15:30 and
2:15:31 here's one thing you can do especially
2:15:34 if you have a particular writing style
2:15:36 take one of your favorite blogs that you
2:15:38 think is structured well and written
2:15:40 well and say based on this blog paste
2:15:44 write me 10 blogs on these 10 no
2:15:50 I have a Blog about X
2:15:52 here's an example of my word paste
2:15:56 give me 10 blog ideas
2:15:59 um
2:16:01 that are non-obvious and I may not have
2:16:04 thought of and it'll give you the 10
2:16:05 ideas and then you can just say okay I
2:16:08 like one three and seven write me a blog
2:16:11 post
2:16:11 copying the tone and structure of the
2:16:15 example I pasted I showed above for one
2:16:17 three and seven and it'll write you
2:16:19 those three blogs
2:16:21 and they will probably be at 85 or 90
2:16:24 percent
2:16:26 and then you'll have your Kevin
2:16:27 McAllister moment
2:16:29 [Music]
2:16:32 I had no idea
2:16:36 but it can't do it again can it it can
2:16:39 do it over and over and over
2:16:42 and if you get clever at
2:16:45 giving it good prompts it will get
2:16:47 really really good at
2:16:49 [Music]
2:16:50 getting better and better
2:16:53 first you take BART across the river on
2:16:55 your boat and then you leave him on the
2:16:56 other side then you go get pie
2:17:00 it's a journey we've all been on Dr J of
2:17:03 course we started with Bard and of
2:17:05 course we took it across the river
2:17:08 what else are we going to use but a boat
2:17:09 you need a boat what are we going to
2:17:11 swim with bud no
2:17:13 put Bart in a boat
2:17:15 you put part in the boat you cross the
2:17:17 river
2:17:18 then of course you leave them on the
2:17:20 other side he's Google after all
2:17:23 silly Google
2:17:26 you invented this technology and then
2:17:28 screwed it up somehow you stay over
2:17:30 there
2:17:31 I'm going to be playing with pie
2:17:33 we come back and we talk to pie oh the
2:17:36 conversations we had she talked all
2:17:38 night
2:17:40 yeah I agree solid solid I like it
2:17:43 Alicia I love you thank you appreciate
2:17:46 the kind thoughts same Kyle basically I
2:17:48 don't track time well
2:17:50 I'll show up early always
2:17:53 all right let's see I'm going to do a
2:17:55 couple more questions I'm leaving can
2:17:56 llms generate their own training data
2:17:59 yes and no sort of maybe we don't know
2:18:05 it's very possible so
2:18:07 llms can create synthetic data and they
2:18:10 can do it really really well like I had
2:18:12 it generate a bunch of personas for me
2:18:14 that were [ __ ] brilliant and so you
2:18:17 could theoretically take those personas
2:18:19 and you could say hey let's train an llm
2:18:21 on all these personas that this thing
2:18:22 generated
2:18:24 some people are saying that's good and
2:18:26 it accelerates these things dramatically
2:18:28 and they're going to get stronger and
2:18:29 stronger and stronger and then there's
2:18:31 other researches that say when you do
2:18:32 that you're gonna have this thing called
2:18:35 Model collapse
2:18:36 and it's what it sounds like it's good
2:18:39 it's good it's good and then it's [ __ ]
2:18:42 and they're thinking that synthetic data
2:18:44 training these things on synthetic data
2:18:47 by these things is going to lead to
2:18:48 model collapse and they're all going to
2:18:50 suck
2:18:52 I don't know I don't know I don't know
2:18:55 the math I don't know how this [ __ ]
2:18:57 works I just don't know why am I hitting
2:19:02 myself in the head with a remote
2:19:04 because someone's got to change the
2:19:06 channel okay it's very simple
2:19:10 the Earth is a donut and you change
2:19:13 channels like this okay everybody knows
2:19:16 that
2:19:17 if you think that's a conspiracy theory
2:19:19 then
2:19:21 you're talking to the wrong guy okay
2:19:23 everybody knows that
2:19:26 bail Count Me In
2:19:29 all right am I done I think I'm done I
2:19:32 think I got to the bottom oh wait here's
2:19:34 Winston Kyle you make me enjoy The Tick
2:19:37 Tock so much
2:19:42 I'll start talking in your funny voices
2:19:44 too oh God it's bad it's bad it's not
2:19:48 quite Adam Sandler bad when he was doing
2:19:50 all his bad voices but it's it's it's
2:19:52 rough it's but thank you I appreciate
2:19:55 that what do I think will happen when
2:19:58 Google crashes
2:19:59 um
2:20:01 hmm
2:20:04 well
2:20:06 I think that Google has a choice to make
2:20:09 right
2:20:12 the orb
2:20:14 um
2:20:17 Google needs to decide are they going to
2:20:19 be Kodak
2:20:22 right
2:20:24 Kodak invented the digital camera
2:20:27 and Senior Management said but that's
2:20:29 not our business model get the [ __ ] out
2:20:31 of my office
2:20:34 Google invented the Transformer
2:20:37 which all of these things are using
2:20:41 and then they realized oh that's an
2:20:44 existential threat to our business model
2:20:45 get it the [ __ ] out of my office
2:20:48 Unfortunately they released the paper
2:20:52 so all these data scientists went off
2:20:55 and created [ __ ] like open AI
2:20:57 and chat GPT and Claude and pie
2:21:02 and perplexity
2:21:05 and llama
2:21:11 so now they're playing catch-up
2:21:15 they have tons of money
2:21:17 um I think their business model has to
2:21:19 change I do like it just I think it has
2:21:21 to change
2:21:23 Microsoft has all of their products to
2:21:26 fall back on like Microsoft can can make
2:21:28 Bing chat
2:21:30 free and they can they can lose money on
2:21:33 it because they can sell Office 365 and
2:21:35 they can sell windows and they can sell
2:21:37 Microsoft azure
2:21:40 what's Google got they've got the ad
2:21:42 business
2:21:44 they had a monopoly for 12 years
2:21:46 something like that 92 93 [ __ ] market
2:21:51 share for a decade
2:21:54 and then overnight this thing comes
2:21:56 along that just [ __ ] obliterates the
2:21:58 need to go to websites you can just get
2:22:02 the answers
2:22:04 oh [ __ ] what do we do
2:22:07 so I don't know they're gonna have to
2:22:09 reinvent themselves I don't know I don't
2:22:11 know I don't know
2:22:13 they got smart people
2:22:16 and they got a lot of money
2:22:18 but they got a lot of hubris
2:22:21 and
2:22:24 their engineer heavy and artist light
2:22:27 you know you look at Apple
2:22:29 [ __ ] best engineers in the world and
2:22:32 the best creative minds in the world
2:22:34 working together in sweet little Harmony
2:22:36 like Kumbaya
2:22:40 here's the church here's the Steep will
2:22:42 open the doors where are the people
2:22:44 they're the people
2:22:46 that's Apple that's Google
2:22:50 so
2:22:53 I don't know so I think they got some
2:22:54 structural challenges because I think
2:22:56 that balance is necessary these are
2:22:58 large language models they're not large
2:23:00 data models
2:23:02 so
2:23:04 the humanities
2:23:05 the Revenge of the liberal arts major
2:23:08 the Revenge of Gen X it's all coming to
2:23:10 pass it's all coming to pass
2:23:13 those that think with a horizontal mind
2:23:16 in a broad world view
2:23:19 will be rewarded in this new world
2:23:22 those with myopic Focus will be
2:23:25 penalized
2:23:27 so I don't know
2:23:30 I don't know
2:23:34 I think it's got to be rough there right
2:23:36 now like like it's got to be like
2:23:40 It's gotta be rough there
2:23:45 'll get it like apparently Gemini the
2:23:47 the thing that's coming after bard
2:23:51 is is supposedly quite impressive but it
2:23:53 but you know you don't think GPT 5 is
2:23:56 going to be [ __ ] impressive
2:23:59 you know you don't you don't think what
2:24:01 [ __ ] Pie's gonna do with 1.3 billion
2:24:04 dollars is impressive these are the guys
2:24:06 that invented anthropic those are the
2:24:08 guys that invented gpt3 they're now
2:24:12 competing with open AI so
2:24:14 this does this list doesn't stay static
2:24:18 so Google's gonna come out with Gemini
2:24:20 and do they come out with that
2:24:22 before GPT 5 or after
2:24:26 right does Gemini jump them ahead of
2:24:28 gpt4
2:24:30 or are they constantly playing catch-up
2:24:33 because they're five years behind I
2:24:35 don't know
2:24:36 and then do they have the do they have
2:24:38 the cajones to
2:24:40 throw their own business model out if
2:24:42 they need to
2:24:44 I don't know like what is the business
2:24:46 model for all these things nobody
2:24:47 [ __ ] knows that right now nobody
2:24:53 no matter what you build in software
2:24:55 right now
2:24:58 just using these Technologies today you
2:25:00 can replicate anyone's business anyone's
2:25:03 technology probably over a weekend
2:25:07 and it's gonna get faster
2:25:10 you're gonna be able to look at a
2:25:11 competitor and go
2:25:13 um AI could you please go
2:25:16 grab all the marketing positioning
2:25:18 identify the target audience identify
2:25:20 the tech stack and replicate that and
2:25:24 then point it toward this target market
2:25:26 and have that thing launched by
2:25:29 um four
2:25:32 how do we compete in that world I don't
2:25:34 know do you because I don't
2:25:39 I don't know why I'm laughing
2:25:42 uh
2:25:45 all right two more questions
2:25:48 I can't register for the salon tomorrow
2:25:50 what's wait what's going on why can't
2:25:53 you register for the salon
2:25:56 is is Meetup broken
2:25:59 because I was trying to send
2:26:05 what's the best way to assist it
2:26:09 three Innovations should be released
2:26:11 should it no I don't think so RJ
2:26:14 I can't register for the salon tomorrow
2:26:22 Link's not clicking through properly
2:26:25 hang on
2:26:28 uh Salon dot AI
2:26:32 oh come on I don't have no internet hang
2:26:34 on let me there's one way to once again
2:26:36 the cat here we'll flip over to connect
2:26:39 to the internet what what the [ __ ] is
2:26:42 going on here
2:26:44 the connection was interrupted
2:26:47 what is happening
2:26:49 oh here we go wait something's happening
2:26:52 oh
2:26:58 [Music]
2:27:07 um
2:27:11 [Music]
2:27:15 it's driving me crazy people
2:27:19 why is the internet's not working let me
2:27:22 try another browser
2:27:24 Chrome
2:27:29 enhanced privacy and Chrome
2:27:32 all right it jumped back to that
2:27:35 my machine is not happy
2:27:37 [Music]
2:27:39 let's see
2:27:40 [Music]
2:28:00 [Music]
2:28:02 I think Chrome's working it looks like
2:28:04 Edge just
2:28:06 edged not so happy with this hotel
2:28:08 internet I think we're gonna have to
2:28:10 we're gonna have to explore some other
2:28:12 opportunities we got some opportunities
2:28:14 to improve the experience
2:28:17 bi-weekly Meetup
2:28:22 join this group
2:28:26 is this what you're getting to
2:28:29 so
2:28:30 the
2:28:33 if you go to
2:28:36 the group and then you say join the
2:28:38 group and then the featured event is
2:28:41 this one here oh wait but here's what I
2:28:44 did
2:28:45 I went to the salon.ai which is a link
2:28:48 tree
2:28:49 and it's the second link right here is
2:28:51 our Meetup is it the Discord that's
2:28:54 broken the Discord community
2:28:56 this might be broken
2:29:05 now that seems to be working maybe not
2:29:09 but anyway if if it's if the discord's
2:29:11 broken that I can go that I can go fix
2:29:13 but the Meetup seems to be working the
2:29:15 Meetup seems to be
2:29:18 go to a salon and go to the second one
2:29:21 bi-weekly meet up and that's our Meetup
2:29:24 and then you can do that there
2:29:27 you can go there and that should work
2:29:30 oh the link Tree Salon it's yeah it's
2:29:36 the the it's link tree
2:29:40 l-i-n-k-t-r dot e e well no just go to
2:29:44 the salon.ai
2:29:47 that's a website the salon.ai wait hang
2:29:51 on hang on people we're gonna wait wait
2:29:55 that was not the right thing this is the
2:29:57 right thing
2:29:58 go to that website address the salon.ai
2:30:03 go there that's a link tree that looks
2:30:06 like
2:30:08 this
2:30:10 see my stupid little cyborg dude and
2:30:14 then the bi-weekly Meetup the second one
2:30:15 down that's the one to sign up
2:30:18 for the salon tomorrow
2:30:20 see and we got pretty pictures see we
2:30:22 got pretty pictures this was generated
2:30:25 mid-journey in AI salon
2:30:28 that was by Ali Ali did that one
2:30:33 see we got 614 members that's a lot of
2:30:36 members not as many as here on Tick Tock
2:30:38 we got thousands we got thousands here
2:30:41 on Tick Tock all right listen
2:30:47 I'm tired it's way past my bedtime I got
2:30:50 important meetings tomorrow but they
2:30:51 don't start till later in the day so I
2:30:53 can sleep in
2:30:54 I can sleep in sleep sleep in all right
2:30:58 um got it ship hopefully that got you in
2:31:01 there oh
2:31:02 got it thank you okay good all right
2:31:04 people are in oh it looks like the AI
2:31:06 learning lab of Kyle Shannon on
2:31:09 tonight's show on tonight's show we have
2:31:12 um
2:31:12 welcome Clint uh by the way I won't be
2:31:18 I'm not co-hosting tomorrow because I
2:31:20 have a business dinner here in New York
2:31:22 when the salon starts but Leah and one
2:31:24 of our other organizers are going to be
2:31:26 hosting and John Knack from Adobe is
2:31:29 going to be speaking
2:31:31 he's worth
2:31:33 listening to because he's
2:31:35 funny and energetic and smart as [ __ ]
2:31:37 and he's got you think I have ADD John
2:31:40 Knox like [ __ ] world class man
2:31:43 um but he's he's senior in adobe's AI
2:31:46 group so he's he's one of the dudes
2:31:48 behind Firefly and generative Phil and
2:31:50 all that sort of [ __ ] so he's the man so
2:31:53 he's this is the second time speaking at
2:31:55 the salon so it's worth going there for
2:31:57 that I'll be there tomorrow I just don't
2:31:58 know when I'll join
2:32:01 um all right everybody hey peace out it
2:32:04 was good joining you from New York City
2:32:05 I'll see you tomorrow night from New
2:32:07 York City after the salon I'll come do a
2:32:09 little bit of a little bit of gaggling
2:32:12 gaggling around
2:32:14 but uh yeah all right peace everybody
2:32:19 Fannie Willis is talking oh she is all
2:32:22 right I gotta go watch Fanny Willis
2:32:23 peace out everybody there's some [ __ ]
2:32:25 going on in Georgia The Devil Went Down
2:32:27 to Georgia he was looking for a soul to
2:32:29 steal he was in a bag because he was way
2:32:31 behind and he was looking wait man well
2:32:33 anyway that song all right peace yeah
2:32:36 that's what I'm talking about
2:32:38 bye everybody