10/7/24 - LIVE from the Airport: AI Discussion & Q&A with Kyle Shannon

Description

Chapters

Transcript

0:29 just
0:35 this might not
0:39 work there's a lot of announcements hey
0:42 Danielle um you're gonna have to let me
0:45 know if you can hear me I'm going to try
0:47 to talk quiet so I don't annoy my
0:49 [Laughter]
0:51 neighbors can you hear me
1:02 let me know if you can hear
1:05 me sounds good all
1:07 right
1:11 um my flight's delayed for another hour
1:14 and a
1:15 half two
1:19 hours two two hours and
1:23 change loud and clear okay cool
1:26 um see I'm used to loud talking at home
1:29 so I don't know how to talk
1:32 quiet and I'm probably being
1:37 loud what the heck I'm at an airport
1:40 here
1:41 [Music]
1:48 look I have a gimbal
1:53 too I got no one's permission to put
1:56 them on on the internet
2:01 I guess I could get rid of on the
2:06 internet I guess I could get rid
2:08 of although I do have my computer
2:23 here all right
2:30 didn't know if I was tuning into the
2:31 right stream normally don't
2:34 stream yeah yeah normally normally
2:37 there's a dog yeah I got no dog here I
2:39 got no dog I do have access to the
2:41 internet so I can look at [ __ ] on
2:43 Twitter love the shirt thank you yeah
2:45 it's pretty groovy but it's got It's got
2:48 very nice um what do you call it
2:52 inside inside
2:55 fabrics and
2:57 embroidery so we'll we'll get get
3:01 all we'll get all hip we'll get all
3:03 airport hip here so here's so here's
3:07 what we should do just ask me what's
3:09 going on with AI or tell me something
3:12 you're working on and then I will try to
3:17 respond to it or
3:19 [Laughter]
3:26 not what's up talk what's happening
3:29 Dylan where are you going for starters
3:32 I'm headed to
3:34 LaGuardia um spending tonight
3:38 in New York then have a
3:43 meeting in Manhattan then we're driving
3:46 to Philly and then we're working our way
3:48 up New Jersey for three days four
3:52 days Kyle's fleeing the country I I
3:54 gotta get away from the
3:58 Irregulars they every [ __ ] good night I
4:01 just got to go I got to get
4:03 out what do you do for a living I am the
4:06 CEO of a company called Story vine no
4:09 Diddy uh Diddy's not available this trip
4:12 normally normally he and I are hanging
4:13 out but he he I don't know what's going
4:15 on he's not available this week but um
4:18 because normally he gets right back to
4:19 me you know um yeah no Diddy no Diddy
4:24 work stuff yeah yes we are everywhere K
4:27 yeah I'm headed to just a bunch of
4:30 clients just going out and meeting with
4:31 clients and like hey do you still love
4:34 us why don't you give us some some some
4:37 projects why don't you give us some
4:41 projects so yes work
4:44 stuff s so meet me weere
4:47 everywhere I wouldn't announce that Kyle
4:50 that's all right that's the
4:53 reality like we say that to them you
4:55 know we're here cuz we want to you know
4:57 we want to work with you I don't there's
4:59 no shame in that no shame in asking for
5:01 what you
5:03 want I see you cut your hair
5:07 well I didn't get it
5:09 cut I cut it I got a little snippers and
5:14 just did the sides so that's you
5:16 [Laughter]
5:19 know oh
5:22 man I need to change the color of the
5:26 clothes on my avatar have you any
5:28 tips um make them bull cut no it's still
5:33 pretty long um I would make the colors
5:36 of
5:37 your these announcements are really
5:40 annoing can anyone understand
5:43 anything in airport announcements like
5:47 is it not just Charlie Brown
5:54 [Laughter]
5:57 speak I'm being stared at what's a good
6:01 AI
6:02 app to make a video text to video
6:05 there's a bunch of them now um clling
6:08 ai.com is the one I've been using
6:10 recently it's really good it's
6:12 Chinese um Runway ml Gen 3 is really
6:16 good
6:18 um uh Deco here is
6:22 decent
6:25 um Korea I think has video now there
6:29 there just a ton of them pabs pabs just
6:33 came out with a
6:36 new something or
6:38 other the hell did P just come up with
6:41 hang
6:43 on if you go to the AI salon and go to
6:47 the video section there's a tools and
6:49 play section um there's a list of videos
6:53 there oh yeah Pika P Labs has P effects
6:56 here look
7:01 see the thing
7:02 exploding and like the other the other
7:11 thing see the car getting
7:20 squished so that's that's P labs
7:34 Runway cling
7:35 Pika hey
7:39 gen that's all we have there I got to
7:41 update that list that list is
7:44 tired um all right what else what else
7:50 happening that's right the AI Salon yeah
7:52 the AI salon's got some good resources
7:55 in it I'm thirsty
8:00 I suppose I could pack up all my [ __ ]
8:01 and just go get water take you all
8:06 along what' you learn at the a salon
8:08 tonight Kyle likes janii he likes Smart
8:11 Water Kyle got smart water I think he
8:14 thinks he's
8:19 smart oh
8:21 man what else can you put in both earbud
8:25 Buds and turn on noise cancelling the
8:27 backround noise isn't bothering us my
8:30 fear of that I
8:32 [Music]
8:37 can
8:38 yeah see my fear of this that's a little
8:41 better
8:44 iose
8:47 um I can't hear how loud I'm
8:50 being it's not good I don't like
8:54 it I'm too
8:58 paranoid if you're new here check out AI
9:00 Salon yeah if you go to the
9:02 salon. and click on join our
9:05 community that'll take you to a vibrant
9:08 uh community site web-based community
9:11 site um got lots of good resources lots
9:15 of good people and if you're new here my
9:17 name's Kyle Shannon and normally I'm in
9:20 my home office where I can talk loud and
9:21 wave my hands and point at [ __ ] on the
9:24 computer but I'm sitting in an airport
9:27 delayed for two hours so I just figured
9:29 I'd come on live just shoot the [ __ ] um
9:32 so if you have questions about AI pop
9:34 them in the comments um if you're
9:36 working on anything interesting pop them
9:38 in the
9:39 comments um if you want to troll me and
9:41 tell me I look like Rosie O'Donnell on a
9:44 bad day that's fine
9:47 too my favorite my favorite troll was I
9:51 was scroll
9:53 with I was scrolling Tik Tok and I I
9:55 thought Alec baldr got fat
9:58 [Laughter]
10:05 not on a bad day on a good day oh thank
10:07 you I look like R Donald on a good
10:13 day that's not a
10:15 [Laughter]
10:19 troll oh it's pretty
10:22 good oh man that's pretty good all right
10:27 um what are some AI business ideas
10:32 okay so here's here's a general category
10:37 of what I think can be done in business
10:40 in any business right
10:41 now so there used to be this adage you
10:45 can have things good fast or
10:48 cheap kick any
10:51 two I think with AI you can have things
10:54 good fast and
10:56 cheap and so I think that every business
10:59 bus out there is um ripe for competition
11:04 from someone who goes all in on AI and
11:07 does whatever you do in that business
11:10 just better faster
11:11 cheaper like I think that's a really
11:13 easy one like if you knew if you know a
11:15 category learn Ai and just get really
11:18 good at delivering things better faster
11:21 or cheaper than the competition I think
11:23 that's loow hanging fruit with the
11:25 existing tools you don't need to worry
11:27 about much of anything
11:29 just improve your internal processes and
11:32 just kick their
11:39 ass it's now great super fast and Nob
11:42 brainer cheap yeah exactly exactly Jim
11:45 Ross is
11:47 here hey Jim I'm flying around I'm
11:50 flying around to a bunch of different
11:51 cities in the United States looking at
11:53 uh self storage
11:55 units hey do you do you know anyone that
11:57 could help me manage those
12:00 if you could if you could get me someone
12:02 that'd be
12:05 great Jim Jim Jim's a he's an AI Guru he
12:09 used to be in the storage business but
12:11 now he's an AI
12:13 Superstar he used to be like oh Kyle
12:16 like you're how do you do all that
12:18 internet stuff and now he's like
12:20 speaking at conferences de demanding six
12:24 figures for a 30 minute speech it's
12:26 ridiculous he doesn't even do storage
12:28 anymore
12:31 so I can do it remotely D Dylan's going
12:35 to help me out I used to manage storage
12:37 for 17
12:38 years you should you should talk to Jim
12:41 Ross he's really good at that [ __ ] or at
12:44 least at least here's the thing with Jim
12:46 Ross he might be shitty at managing
12:48 storage businesses but his marketing
12:50 materials are really good because he
12:51 uses
12:53 AI who's Jim Ross um Three Mile Storage
12:56 on here the person that said Three Mile
12:59 Storage
13:01 he's also a nice guy he knows a
13:03 [Laughter]
13:06 guy it's fun
13:11 Siri at least Siri's
13:18 consistent all
13:20 right I got to get some water all right
13:23 so we're going to we're going
13:27 mobile you're coming with me
13:33 all right
14:09 I'll be right with you calm
14:26 down all right we're going walk
14:38 all
14:39 right let's see how do you feel about AI
14:42 models being trained on artists work
14:45 without their permission
14:53 um it's not black and white let me put
14:56 it that way
14:58 um if the world were
15:01 perfect what would have
15:03 happened
15:06 is open Ai and everyone like that how's
15:10 it going would have asked permission
15:14 hang on let me buy my water
15:16 here how's it going no it's going all
15:26 right all right
15:30 you
15:31 go you want to say hi to Tik Tok
15:34 [Laughter]
15:38 oh oh I do
15:42 actually thanks a lot
15:47 thanks he had he had no interest in
15:49 talking to you
15:52 none all
15:56 right hang on a sec
16:00 I want to get back to this because it's
16:02 important
16:04 um
16:07 okay in a perfect
16:10 world
16:12 um all of these model companies would
16:15 have asked every single artist's
16:18 permission um we don't live in a perfect
16:23 world none of those copyright holders
16:25 would have given permission they all
16:27 would have said you owe me a bunch of
16:28 money for that and so we wouldn't have
16:31 generative AI we wouldn't have these
16:34 models there's
16:37 also a really good argument
16:40 for
16:41 um all of this stuff Being for the
16:44 Benefit of
16:45 humanity um that there's a tipping point
16:48 at
16:49 which if you if you train models on 10
16:53 artists and don't pay
16:55 them you know that's one in 10 you know
16:58 10% of
17:00 of that content is stolen that's High um
17:04 hang on I'm getting my thing set back
17:08 up um if you have a thousand artists in
17:12 there then it's a smaller percentage if
17:16 you have a million and it's a smaller
17:17 percentage if you have all of them
17:20 across all of
17:22 history then it's you know a grain of
17:26 sand on all the world's beaches in terms
17:28 of
17:29 what's in
17:31 there
17:33 um also a lot of
17:36 the a lot of the people that
17:40 are that make examples they
17:43 prompt the systems to
17:47 explicitly um to
17:52 explicitly create art where they can go
17:54 oh look that's my art you stole from me
17:57 um I don't think that's how most people
17:58 use it it so so that's a part of it so
18:02 this stuff wouldn't exist had that not
18:05 happened
18:07 um the the other thing about it
18:12 is where these model companies are going
18:15 right
18:16 now um I Heard a
18:19 term I think I heard it from Paul rer
18:22 called copyright washing which is really
18:25 interesting where they're starting to
18:28 generate synthetic data so they take a
18:30 model that's been trained on all of our
18:34 stuff and they generate synthetic data
18:36 out of that and then they train a new
18:38 model on the synthetic data so the new
18:41 model hasn't been trained
18:43 on the the direct training of people's
18:47 stuff um
18:50 so I understand why people are pissed
18:53 off but what this channel is about is
19:01 independent of how you feel about AI
19:05 independent
19:07 of all of it it's not going away it's
19:11 accelerating and it's going to change
19:13 everything and what this channel is
19:14 about is assuming that that's true right
19:17 it's possible this is all just hype and
19:19 it all goes
19:21 away I think the chances of that are now
19:24 minuscule
19:30 what this channel is about
19:32 is don't sit on the
19:35 sidelines pissed off you can be pissed
19:38 off but don't be on the sidelines get on
19:40 the field learn how to use these tools
19:43 lion gate just signed to deal with
19:45 Runway ml they're going to train a
19:47 Runway ml model on all of the content
19:49 they've ever produced so they're going
19:51 to have an official model um some of the
19:54 films that are being created with with
19:56 these generative AI movie tools are
19:59 really really good if you're a
20:01 Storyteller and you're not learning
20:02 these tools right now I think that's
20:04 absolutely [ __ ] foolish you could
20:06 still be pissed off about it you could
20:08 still fight for compensation and
20:10 attribution I know adobe's looking at
20:13 ways to do that I know there's projects
20:15 out there you know looking to increase
20:19 attribution and contribution to artists
20:24 um
20:26 but I wouldn't not use this
20:31 [ __ ] I just think that's foolish all
20:34 right my question is how can they keep
20:36 affording to maintain generative AI it's
20:38 so expensive I have no idea hey you're
20:41 on early I know I'm at an airport in
20:43 case you can't
20:47 tell I think I'll put it
20:50 like let's see I'm trying not to get
20:52 people in the background well whatever
20:54 who gives a [ __ ] probably the people
20:57 behind me
21:03 all
21:06 right he
21:09 everybody too much money at stake yeah
21:11 exactly airport lighting is nice I know
21:14 I look all and look how
21:16 creamy creamy is not the right word is
21:20 [Laughter]
21:27 it oh you can put an AI model on your
21:30 home PC way cheaper you can they're just
21:32 not as good yet but they will be they'll
21:34 be pretty good give it a year or two
21:37 dewy yes I have a dewy glow about me
21:39 don't I I'm just I'm just this side of
21:43 moist
21:45 [Laughter]
21:51 dewy all right who who who else has got
21:54 a big provocative
21:56 question when when did you stop beating
21:59 your
22:08 wife why should they benefit off my work
22:12 okay hang
22:14 on who's
22:17 day who's day
22:30 if the they you mean is open AI or mid
22:32 Journey or things like
22:35 that there's a bit of an argument there
22:41 but how do you so so for all of us that
22:44 have freely put our [ __ ] on the
22:48 internet our art portfolios things like
22:53 that how much money were you making on
22:55 the [ __ ] you put on the internet before
22:57 all this stuff
23:01 if someone if
23:03 someone did a search on Google and they
23:05 found one of your images and they
23:06 photoshop that together into some other
23:10 image they're benefiting from your image
23:13 but not making money on it I
23:15 mean I can understand
23:19 that you can say they they shouldn't be
23:21 able to benefit from that
23:23 but you can make a similar argument that
23:26 a library shouldn't be able to you know
23:29 just let people look at your [ __ ] for
23:33 free but there's an argument there for
23:35 the greater good of humanity I think
23:37 there's there really there really is a
23:39 legitimate argument here if you haven't
23:41 used these tools you need to use these
23:43 tools and understand what they make
23:45 possible because not only are they
23:47 benefiting from your work they're
23:50 benefiting from all of our
23:52 work and it's not really them that are
23:55 benefiting we are benefiting like we get
23:58 to tap into to the collective output of
24:03 humanity it's it's in a lot of ways it's
24:07 no different that if you go to Art
24:10 School let's let's say you're an
24:13 accomplished artist and I go to art
24:15 school and in the
24:18 textbook or or my my art Professor my
24:21 painting Professor loves your work and
24:25 and puts up for a whole week we study
24:27 your work and we paint on your work and
24:29 we learn your
24:30 techniques you don't get paid for that I
24:33 then go out in the world and I'm so
24:35 inspired by your work all of my work
24:37 looks like your worth I'm benefiting
24:39 from your
24:41 work that's kind of how these tools work
24:44 right your work ceases to exist when
24:46 they embed them into these
24:49 models it goes into this bizarre
24:52 mathematical Laton space and images turn
24:54 into these
24:55 noisy amalgamations like your wor work
24:58 ceases to exist written works cease to
25:00 exist they go into
25:02 this thousand dimensional latent
25:05 space and then everything is
25:07 reconstructed from the piece
25:10 Parts in this really Dynamic way
25:13 so I don't know I I mean listen it's not
25:16 black and
25:18 white in the perfect world they wouldn't
25:21 have done
25:22 that but if they' asked for for
25:24 permission none of this would exist so
25:31 no you're plagiarizing me it's not
25:33 plagiarism it's not how it
25:35 works L pazi it's not how it works learn
25:38 how the tools work they're not
25:39 plagiarizing they're not copying and
25:41 pasting it's not
25:43 plagiarism exactly what I've been saying
25:45 from the
25:46 beginning it's also no longer your work
25:50 than if it's been altered libraries are
25:52 different why are libraries different
25:54 why are museums
25:56 different why
26:06 the people that walk in get to
26:08 be get to interact with your work how
26:11 did that
26:15 happen how did you pop in here I didn't
26:18 let you in
26:20 here I have no idea I try to get out
26:26 myself that's bizarre I got to tell you
26:29 what though man your laugh is infectious
26:32 and this has been phenomenal so I follow
26:35 you I join AI group I go all it's all
26:38 good so I will figure out how to get the
26:40 heck out of here have a lovely flight oh
26:42 my goodness I don't even know what the
26:45 heck I didn't even do anything I didn't
26:47 either that's pretty cool well nice
26:49 talking to
26:50 you be safe love okay thank you
26:55 byebye was awesome
27:09 lucky oh I like her not Dar I thought I
27:12 totally thought that was a Dar
27:16 moment but I let Dar in that's the first
27:19 time I've ever had
27:21 no I'm getting real
27:24 stuttery I wonder if my app's about to
27:26 crash
27:33 all right what else should we talk
27:35 about Google's a World Encyclopedia and
27:38 it gave generative abilities
27:41 yep indeed indeed
27:44 indeed yeah
27:57 [Music]
28:04 I'm a Charlie
28:07 64 all right where are you going from
28:12 I'm leaving Denver heading to laguadia
28:17 I generated a guitar for a book cover
28:20 copyright office said nope not covering
28:23 the art yeah I know so one of the things
28:27 you're going to need to do the there's a
28:28 bunch
28:31 of there are a lot of miles to go with
28:34 being able to copyright this
28:36 stuff
28:39 um you've got to you've got to
28:41 demonstrate right now to the copyright
28:43 office significant human
28:45 contribution um so if you want to
28:48 copyright stuff through them um you're
28:52 going to have to document all of your
28:54 contributions
28:59 and even then they'll probably still not
29:00 do it but at some point they will I mean
29:04 if if you look at their arguments for
29:06 why they're not allowing you to
29:07 copyright generative AI
29:09 work you shouldn't be able to copyright
29:12 digital photography because all you're
29:15 doing is pushing a
29:20 button you push a button and [ __ ] comes
29:22 out right you shouldn't be able to
29:24 copyright that but you can
29:29 and yet if I go in and I prompt a gener
29:32 generative image model 400
29:36 times and then I take that image and I
29:38 manipulate it and then I put that on my
29:40 book cover why is that not
29:44 copyrightable because they say it's
29:49 not
29:51 whatever okay here's here's my other
29:53 point of view on that stuff don't get
29:56 your panties in a bunch about whether
29:58 you can or cannot copyright [ __ ] the
30:01 copyright stuff is about to go through a
30:08 nuclear
30:09 meltdown it's
30:12 just it's a bunch of laws that were
30:15 essentially created when it was really
30:18 hard to copy
30:21 stuff and now not only is it not hard to
30:24 copy stuff you can generate stuff out of
30:26 thin air like that
30:28 we're we're moving into a world of
30:31 infinite content generation infinite
30:33 personalized content
30:35 generation
30:38 so
30:39 all just assume your shit's gonna be
30:42 ripped
30:43 off move on make the next
30:50 thing um then you got to get some 2.5 so
30:53 you don't have to look up so high and we
30:56 can't see up your nose here wait
31:00 is that better wait do you want me oh
31:02 you need to be
31:09 higher how's that that good is that non
31:14 nose look
31:19 upy
31:21 manifestation what's your background um
31:25 I'm a Storyteller by training um have a
31:29 degree in acting I moved to New York
31:30 City out of school um started r a
31:33 theater company wrote seven screenplays
31:37 um in the
31:39 mid90s I created one of the first online
31:42 magazines and then co-founded agency.com
31:45 which was one of their first digital
31:47 agencies and so built a lot of the early
31:51 websites for the Fortune 500 we built
31:53 Coca-Cola's first website we built
31:55 British Airways first big website and
31:57 ticketing system system we broke just
31:59 just tons of stuff like that
32:03 um sold that in
32:06 2002 and then um my current company is
32:10 Story vine which is an automated video
32:12 storytelling
32:14 platform and then about three years ago
32:16 I got enamored with generative image
32:18 making with stable diffusion and dream
32:21 booth and things like that and then chat
32:23 GPT came out and I lost my [ __ ] mind
32:27 because um
32:30 what what struck me about chat GPT is
32:35 that it does or it did it
32:40 does for machine learning and AI what
32:43 the worldwide web did for the
32:46 internet so when the internet first came
32:49 out well it it' been out for decades but
32:51 it it was command line you had to you
32:54 had to type in IP addresses and you had
32:56 to do command line
32:59 um prompts to get it to do anything and
33:03 then Tim burners Le invents the
33:04 worldwide web and makes it so you can
33:06 click on the internet click click click
33:11 click um and chat GPT does that for
33:15 these incredibly complicated
33:16 sophisticated
33:18 tools and oh thank you very much whoever
33:21 just sent a gift thank
33:22 you
33:26 um so so chat GPT is
33:31 democratizing access to
33:33 these insanely powerful tools like if
33:37 you think about how simple the HTML
33:40 technology is
33:41 hyperlinks or you think about how
33:43 sophisticated AI is
33:46 um you know the worldwide web changed
33:49 the world changed the world changed how
33:51 we do business Chang how we communicate
33:54 changes how we travel
33:57 everything and is going to do that times
34:00 10,000 million Bonkers are you wearing
34:04 smart pants on the plane I am not I'm
34:07 wearing I am probably wearing the same
34:09 jeans I wore when I got yelled at by the
34:13 British Airways look
34:15 [Laughter]
34:20 lady oh
34:22 man what time is it oh I've got to watch
34:26 okay
34:28 I've got
34:30 time no offense I do the same thing I
34:34 don't know what same thing I don't know
34:35 what I was doing what was I
34:38 doing
34:40 [Laughter]
34:46 OMG the difference then was I was flying
34:49 British A's first class for free because
34:51 they were a
34:53 client that's why I got yelled
34:56 at it's gener if it work for British
34:59 Airways we wear smart
35:07 pants [ __ ]
35:11 Brits oh yeah yeah she yeah she yelled
35:14 at me because I was on a so so on your
35:17 boarding pass it basically says you know
35:19 he's a freebie he works for us he's a
35:22 vender and so you know there's an
35:25 agreement there's a gentleman's
35:27 agreement that when flying first class
35:29 for free we we don't look like a shitty
35:35 [Laughter]
35:49 American we missed you today in
35:52 prototypes oh I
35:53 know yeah I was in the middle of I I had
35:57 my my wife had has an art show opening
35:59 Friday so I had to hang her show this
36:01 morning and run her home and then get to
36:03 the airport and then as I'm pulling into
36:06 the airport they're like flight's
36:07 delayed for two hours I'm like
36:10 ah what are you going to
36:16 do all right who else has got questions
36:19 thoughts opinions on how awful AI is I
36:23 mean again here's the deal if you're
36:27 still here and you're pissed off at AI
36:29 if you haven't used
36:33 it get out of the stands get off your
36:35 soap box be pissed off about it but
36:38 learn it go use it go play with it
37:02 Dr J it's changed my life have you tried
37:05 the real time voice on co-pilot I've not
37:09 I didn't know it was available
37:14 yet
37:17 has has AI improved even faster than you
37:20 predicted last
37:23 year sort of yes and sort of no okay
37:26 yeah so okay okay all right that's a
37:29 really good question
37:33 okay in the past three
37:39 months can you hear that
37:43 [Music]
37:51 kid two things are happening
37:54 simultaneously and the third thing is
37:55 that kid screaming um
37:59 in the past three
38:02 months all of
38:05 the all of the major model companies
38:08 have caught up to where open AI was with
38:11 at least gp4 like they're all in the
38:14 neighborhood are really really
38:17 good the image oh my God the image
38:20 generation models are really really good
38:24 the video generation models are really
38:26 really good the music generation models
38:28 are really really good so all of them
38:30 have gotten
38:35 good um
39:05 the the1 model the 01 reasoning
39:09 model is insane it's Bonkers um the
39:13 realtime voice that open AI just
39:15 launched is insane it's Bonkers we're
39:18 about to get um claud's 3.5 Opus right
39:22 right now we've got 3.5 Sonet which is
39:24 their mid middle model that you know got
39:26 as good as GPT four we're about to get
39:29 Opus um at some point here we're going
39:31 to get GPT 5 at some point here we're
39:33 going to get grock 3 which is being
39:35 trained on 100,000 models right so the
39:39 next three months the tools are going to
39:41 get
39:41 insane I'm in the middle of kind of this
39:44 exis existential crisis around AI where
39:47 it is clear to me now that these tools
39:50 are
39:53 significantly more capable than my
40:00 they're more capable than my
40:02 ability to use them well I feel like I'm
40:07 just like an
40:09 adult you know using these things that
40:12 are just these insane insanely powerful
40:15 thing so so so in that
40:20 sense I I think I think things have
40:23 evolved about as I expected what what is
40:26 not evolving is fast as I thought it
40:29 would is
40:31 adoption nobody seems to give a [ __ ]
40:34 nobody seems to be really paying
40:36 attention now the people that use these
40:38 tools are like holy [ __ ] these things
40:40 keep getting better and better and
40:43 better but for the most part nobody
40:47 cares and I I just I find that I because
40:51 what's happening is as the as the tools
40:54 are getting better so it used to be if
40:56 you were here if you hadn't started yet
40:59 there was only a handful of tools that
41:00 were really good and you could sort of
41:02 get on and learn prompting and you could
41:04 sort of figure it out because it was
41:05 small well now there's like 20 or 30
41:09 major tools that you could use catching
41:12 up to that is not easy you can do it but
41:15 not easy but as the tools get more
41:20 sophisticated the distance between
41:23 people that are AI literate and the
41:25 people who haven't even thought about it
41:27 yet is like it's accelerating away from
41:30 people and
41:32 businesses and
41:34 and that lack of adoption has surprised
41:37 me I mean it shouldn't like it took 25
41:40 years for ad agencies to admit that
41:43 digital advertising was as important as
41:45 TV
41:49 advertising so I've seen it
41:51 before um I find that really frustrating
41:55 that that people are just letting this
41:57 pass
41:59 by but it's Bonkers it's absolutely
42:02 Bonkers agreed 17 months in and I still
42:05 feel very inadequate to use these tools
42:07 yeah I feel less and less adequate to to
42:11 to deal with them you work for Claude AI
42:14 I do not I am just a like what would you
42:18 call me a um a well quaffed pundit am I
42:25 a pundit I don't think I'm an in
42:29 I don't have pink
42:31 hair I have yet to twerk I don't wear
42:34 maninis on the
42:36 [Laughter]
42:44 channel I'm just very curious yeah I
42:47 like ttles exactly that's what are your
42:50 qualifications I like
42:53 twiddles and yet well that's the thing I
42:57 mean
42:59 this is
43:01 why AI is a really tough term because it
43:05 it's way too
43:09 broad what this channel is about is not
43:12 really about AI it's about generative AI
43:15 right it's about post chat
43:17 GPT use of the tools by people that are
43:20 not
43:22 scientists
43:26 um and what's amazing about where we are
43:30 right now is you don't have to have a
43:32 math degree to take advantage of these
43:34 tools like that's what chat GPT did
43:36 that's the gift that that one piece of
43:40 software that by the way open a I didn't
43:44 think anyone would care
43:46 about it was the fastest adoption of
43:49 software in history 100 million users in
43:52 six
43:53 weeks
43:55 um because it's as remarkable as it is
43:59 but it it just blows me away that
44:01 businesses are not pouncing on this AI
44:05 stuff but they're
44:11 not and I think I mean I kind of
44:13 understand it because it because it
44:16 requires you to actually think it
44:19 requires you to actually like really
44:21 think about well wait if these tools can
44:24 do
44:25 that doesn't that mean that our business
44:28 model is totally [ __ ] right that's one
44:30 of the possible implications of AI and
44:32 then you got to deal with that that's
44:34 not fun to deal
44:36 with uh new Category 6 created for
44:39 Milton that's
44:41 terrifying yeah Milton went from like a
44:43 category one to a category five in in 24
44:47 hours I saw some scientist that was
44:50 talking about how quick the mbars
44:52 dropped and he started
44:53 crying like the scientists haven't seen
44:56 anything like this terrifying I got
44:58 family in Tampa it's just heading right
45:01 toward Tampa right through Tampa to
45:05 Orlando it's
45:08 wild can you make
45:13 money can you make money with chat gbt
45:16 caught you without your toy I can still
45:18 say it that toy exists because you can
45:20 make money with chat
45:22 [Laughter]
45:26 BT oh you can make money with ch
45:29 [Music]
45:39 GP all things all things language is Deb
45:43 my kids in the first year of computer
45:45 science and I wonder how much her proess
45:48 professors are up to date on it what
45:50 what I would strongly now listen if if
45:54 your if your daughter is like my boys
45:56 you can't tell her [ __ ]
45:58 she won't listen to
45:59 you I would have her spend every waking
46:04 moment teaching herself AI every waking
46:08 moment have her join the AI Salon have
46:11 her join projects if she's into computer
46:13 science have join open source
46:15 projects every waking moment because I I
46:18 would my instinct is that the professors
46:21 are not going to be up to speed on it so
46:24 I would not piss around with that like
46:30 if she's not actively learning to
46:31 program using
46:33 AI like you know what you should do
46:37 Deb if you haven't played with Claude
46:40 artifacts yet go to
46:43 claw. and just say to claw. make me a
46:48 playable clone of the video game Frogger
46:52 or Space Invaders or just some Whatever
46:55 video
46:56 game and watch it code it and then play
46:59 that game and figure it out enough so
47:03 that you can you can replicate it right
47:05 so so you can do
47:06 it and then and then if she doesn't know
47:10 about this stuff yet um next time you
47:13 see her just say oh you know I know
47:15 you're learning coding at school you
47:17 want to see something cool and then go
47:19 to Claude and just have a code a video
47:22 game for you in front of her
47:32 how do you do that
47:34 Mom that's
47:39 annoying P's in what's happening
47:43 Pate Oh Lordy all right what else should
47:46 we talk
47:50 about
47:52 um it looks like a bunch of stuff
47:55 there's there's a robot but announcement
47:59 on the 10th October
48:04 10th I have a feeling something else is
48:06 going to drop oh by the way oh this is
48:08 something I wanted to say so it's
48:10 becoming it's becoming
48:13 clear from documents and emails and
48:18 things that mea moradi the CTO of open
48:23 AI that just
48:24 quit um she's been the
48:28 one slowing things down or she was the
48:31 one slowing things
48:33 down she didn't want 01 preview launched
48:37 um
48:38 apparently this is fun apparently the
48:41 testing team had nine days so I'm I'm
48:45 working on a musical right now we
48:47 actually just wrote in a
48:48 scene yester over the weekend Sunday
48:53 yesterday we wrote in a scene into the
48:56 musical
48:57 where the CEO of Microtech uses chat GPT
49:02 ping chat we call it he uses ping chat
49:05 and he says this is so good we got to
49:07 launch it and the testing team the the
49:10 lead tech guy says you know well we can
49:13 have it tested in a year he goes a year
49:15 we're going to launch it in a week and
49:17 then they negotiate back and forth and
49:20 basically they have seven days to test
49:22 it well apparently that that's actually
49:26 essentially true that 01 preview they
49:28 had nine days to test it because someone
49:31 was making a big announcement meta was
49:33 making a big
49:36 announcement and Sam mman wanted that
49:38 thing launched um so yeah so anyway so I
49:43 have a feeling we're going to start to
49:44 see a lot more stuff coming out of open
49:47 AI um now that miror is not
49:50 there we can hear you uh oh did did
49:54 my something go away
49:57 everyone is
50:02 here who's been using the free version
50:04 of chat
50:11 [Music]
50:17 GPT you're
50:19 fine he's been doing really cool
50:24 stuff you're a b
50:27 you're good okay
50:29 cool oh that's a nice dog
50:35 look nice
50:38 doggy nice
50:40 doggy I like doggies doggies are
50:46 nice oh you're blind client blind he's
50:49 blind oh that's cool what's with the
50:50 early life do you notice I'm not in my
50:52 shitty home office I'm in a rather
50:55 expansive home office C in airport my
50:59 flight's delayed for two hours
51:02 so and because of that whenever when I
51:05 was supposed to go live like I would
51:06 have gone live at my hotel except I'm
51:09 not going to be there when we normally
51:11 go live so I just was like [ __ ] it so
51:14 here we
51:21 are you told him about the advanced
51:23 voice
51:24 version yeah if if he's if if he's um if
51:28 he's blind the The Voice version I mean
51:31 The Voice version right now is just
51:32 absolutely Bonkers but it doesn't even
51:34 have it doesn't have Vision yet wait
51:36 till it gets vision and you can just
51:38 sort of take a picture in front of him
51:40 and have it describe describe what's in
51:42 front of
51:43 him it's amazing like the the the
51:46 accessibility that AI is GNA deliver to
51:49 people is just [ __ ] crazy so crazy no
51:53 worries no champ of your guitar hey Kyle
51:55 what's up what's happening spin B3
51:58 call them about the AI Salon yeah if
52:00 you're new here um my name's
52:03 Kyle um we talk about AI stuff here if
52:06 you want to be part of a really cool
52:07 community that a lot of them hang out
52:10 here go to the salon. and that's
52:14 uh that's our website and click on join
52:17 the community and it's an online
52:19 community of AI
52:21 optimist so especially if you're someone
52:24 that's curious about AI if you're
52:26 optimistic about Ai and you're the only
52:28 one in your life that is and everyone
52:29 else is like you know it's theft right
52:32 you know it's the world's greatest
52:33 plagiarism machine don't you um if you
52:36 get a lot of that uh join the AI Salon
52:38 because then you can just talk about Ai
52:39 and not feel like you're looking over
52:40 your shoulder um I put my I put my new
52:44 musical Sydney on Facebook it got it got
52:47 [ __ ] crickets um Eric Eric saner from
52:51 the salon posted um yeah this is my
52:54 friend Kyle and you may hate AI but uh
52:57 he's kind of cool and if you don't hate
52:59 it too much you should look at this
53:01 thing so Facebook is a whole another I'm
53:05 at least used to LinkedIn and Twitter
53:07 where there's some AI positivity
53:10 Facebook's Facebook's where all the
53:12 people that are not trying it are
53:18 apparently my Corp is leaning on my gen
53:21 skills but they aren't going to pay me
53:24 more don't be generous
53:31 Corporate America so you know what I
53:32 would do
53:33 cuno
53:37 is I don't know your
53:39 situation but that's
53:43 horeshit if they're leaning on you for
53:45 that stuff I mean there's two things you
53:48 could do you could just explicitly ask
53:50 for it say hey pay me more or like put
53:54 yourself out there start start having
53:55 interviews I saw thing I saw a thing yes
53:59 or no this morning A Guy created um an
54:03 autonomous agent that went and applied
54:06 for a thousand jobs in 24 hours and he
54:09 got 50
54:11 interviews but but [ __ ] it go find a
54:13 place that's curious about AI like like
54:15 I just I I feel like working for any
54:18 company any especially a big Corporation
54:20 right now that is not actively like
54:23 actively seeking out their employees who
54:25 know anything about AI and promoting
54:27 them and saying hey teach us how to use
54:29 this create trainings for us they're
54:32 going to get [ __ ]
54:35 so I've been so generous in teaching
54:37 people and ask for a raise nope yeah I
54:40 [ __ ] them I I would I
54:43 would I would open your
54:49 possibilities that's shitty that's just
54:51 shitty yeah because you've been you're
54:53 you're very generous you're really smart
54:55 you're
55:08 yeah you essentially brought them into
55:11 the AI AG that's sad yeah it's crazy
55:14 yeah it's just wrong I yeah that does
55:16 not make me
55:19 happy but that's exactly you know when
55:21 someone said before like am I am I
55:23 surprised by how fast things have
55:25 progressed
55:29 I'm surprised by
55:32 how how little Corporate America has
55:37 progressed it's
55:39 weird but it's every time it's every
55:42 time it's every new
55:46 technology I don't know I mean I think
55:48 it's just [ __ ] corporate hubris right
55:52 it's like well we do what we do that's
55:55 what the the advertising holding
55:57 companies right now their clients are
55:59 explicitly saying to
56:01 them we want faster
56:05 turnaround higher quality for less
56:09 money and the agencies are like
56:12 wow this is how we do things we're us
56:16 and we're going to do it like
56:20 this
56:22 staggering I'm finding it difficult to
56:24 keep interacting with the folks that
56:26 still insist on hating on AI well
56:29 exactly archetypal architect
56:35 I I mean that's what this channel is
56:38 about I don't mind I actually like
56:42 talking to people that hate on
56:44 AI but not to hate on
56:48 AI in an ongoing way like if if you're
56:52 hating on AI if you're afraid of AI and
56:54 you've never tried it
56:57 and then you try it and then you're
57:00 still hating on it to that
57:04 degree I'm down with that you know
57:07 you've got your reasons you understand
57:08 what it is but to a person like I don't
57:12 think I've seen an exception to this the
57:14 people that are the most angry about it
57:16 and the most afraid of it have never
57:19 tried it they don't understand how it
57:21 works they're not curious about it
57:23 they're not curious about what it can do
57:25 for them they're not curious about how
57:27 it could reinvent them they're
57:29 not they they have zero interest in it
57:33 and I just I have no time for that like
57:35 I I'm sorry like like you know we've got
57:39 this Milton hurricane sort of blasting
57:41 across the gulf right now heading for
57:44 Tampa like I I feel like it's the
57:47 equivalent of saying well I don't agree
57:48 with
57:51 hurricanes okay isn't it still coming
57:55 well yeah but I don't agree with it
57:59 okay might want to buy some plywood but
58:02 didn't you hear me I don't agree with it
58:08 okay it's bizarre to
58:14 me unfortunately you know I think that
58:18 the I think that the crypto Bros the nft
58:21 Bros and the blockchain The Blockheads
58:24 the blockchain idiots
58:30 that you know hyped up all that stuff
58:32 and said it was going to change
58:33 everything and then it all kind of
58:34 flamed out and deservedly
58:40 so I think people look at that and they
58:43 they they look at the AI enthusiasm like
58:45 oh it's just like another blockchain
58:47 it's just like another crypto I yeah if
58:49 it gets real just call me when it gets
58:52 real and like it's
58:54 getting really [ __ ] serious like it's
58:57 getting really [ __ ] powerful and I
58:59 and I just there's just not a lot of
59:01 interest in it right
59:06 now it will make things better open more
59:10 doors that's basically goes what goes
59:12 through my head like do you think you're
59:15 GNA stop it they're not they're not
59:37 oh
59:38 who I just about dropped my
59:41 gimbal
59:43 Jesus I [ __ ] lost
59:47 it oh man all
59:53 right all right what else what else what
59:55 else what else
59:57 exactly the core I find the intention of
59:58 the AI Salon play curiosity generosity
1:00:01 Etc
1:00:05 yeah like if nothing else play with
1:00:08 it you know have have you lost your
1:00:11 sense of Joy the answer is yes of course
1:00:13 all the people that're like
1:00:17 AI they have that face what is with that
1:00:20 face AI stupid ai's plagiarism ai's a
1:00:24 threat ai's gonna kill us
1:00:27 you know what I say to them if you keep
1:00:29 saying that your face is going to stay
1:00:31 like that you
1:00:35 ninny have you used the new canva
1:00:38 feature the new canvas feature on uh
1:00:42 chpt or the new
1:00:44 canva canva like some AI feature in
1:00:47 canva have a good trip Kyle gotta go see
1:00:50 you later Drew thank
1:00:54 you um the canvas feature on um on open
1:00:59 AI is crazy
1:01:02 good yeah on chat GPT that's crazy good
1:01:05 AI is going to kill us have you heard of
1:01:06 global
1:01:08 warming we're witness it sounds like
1:01:10 they just invented a category six uh
1:01:14 thanks to the uh thanks to the couple of
1:01:17 degrees
1:01:21 Centigrade oh man people who are anti-
1:01:25 AI are just bed because they know they
1:01:27 need to learn about it well you know
1:01:29 what's funny what's your name groom dude
1:01:32 dirty waffle okay first of all awesome
1:01:35 name 10 out of 10 room dude dirty waffle
1:01:39 okay you're you're the you're the name
1:01:42 winner today I like it
1:01:47 um you're absolutely right that
1:01:51 that I think part of the fear and and
1:01:53 listen I've I experienced this when I
1:01:56 first started getting into
1:01:59 this I don't know if you can tell I'm an
1:02:01 older Feller I know you're like but
1:02:04 you're on Tik Tok aren't you like 22 no
1:02:06 no I'm older than that and when I first
1:02:09 got into this stuff I was like I had
1:02:12 like AA and anxiety and I'm like I don't
1:02:14 know if I can learn this again I don't
1:02:16 know have the energy to learn it again I
1:02:17 don't know if I want to learn it again
1:02:20 oh looks like our plane is deep
1:02:22 planing um and and then I just was like
1:02:26 [ __ ] it I'm going to try it I'm going to
1:02:28 do it anyway and what you learn when you
1:02:30 play with these tools is that they're
1:02:32 not as hard to learn as you assume
1:02:34 they're going to be
1:02:37 because what we've learned about AI
1:02:40 historically is that you need to be good
1:02:43 at math you need to be good at science
1:02:45 you need to be good at programming so
1:02:47 when people hear AI they think like it
1:02:50 departments you know doing predictive
1:02:53 modeling for for business Analytics and
1:02:56 the reality is no no you can just make a
1:02:59 recipe book for your
1:03:01 kids it's like you know go make a pretty
1:03:05 picture you want to make a song that
1:03:07 doesn't suck you can do that
1:03:09 now and so I it's it's why we ultimately
1:03:13 landed on the first step in the journey
1:03:16 to AI literacy in the AI Salon is play
1:03:18 first play just
1:03:21 play don't don't put so much pressure on
1:03:24 yourself that you have to learn it
1:03:25 because first of all you can't there's
1:03:27 too much to learn it's changing too fast
1:03:30 it's evolving too
1:03:33 fast here you know be cool well going to
1:03:37 turn
1:03:38 my thing so all the people walking away
1:03:41 those are the people coming off my
1:03:47 plane
1:03:50 whoops here see oh man
1:03:59 um yeah just learn this
1:04:03 [ __ ] be the product of AI playfulness
1:04:07 yeah exactly such a great way to say
1:04:08 that thank
1:04:10 you have you used open AI canvas so fast
1:04:13 well you know what it is jeie it's it's
1:04:16 um I think all of us that have been
1:04:18 playing with with chat
1:04:21 GPT we've all mastered some version of
1:04:24 this
1:04:25 technique you have chat GPT vomit out a
1:04:28 bunch of stuff and then you get it to
1:04:31 like 80% right in getting it to 80%
1:04:35 you've got like 17 different versions of
1:04:38 it and then it always will give you one
1:04:41 good thing and then a bunch of crap
1:04:44 right so you have 17 versions in chat
1:04:46 GPT so you take the latest one where
1:04:48 you're like it's close enough you copy
1:04:50 that into word or docs or whatever and
1:04:53 then you go copy and paste a bunch of
1:04:54 paragraphs that you thought were good to
1:04:56 and you paste them over there and now
1:04:57 you have this Frankenstein document
1:05:00 that's just [ __ ]
1:05:01 horrible and editing it is a
1:05:04 disaster and so what canvas does is it
1:05:07 just it gives you the document and you
1:05:09 just do no shitty title shitty paragraph
1:05:12 shitty section add a section and it just
1:05:15 does
1:05:15 it
1:05:19 so it's pretty
1:05:22 amazing
1:05:24 uh all right right how long have you
1:05:26 been live I don't know 45 minutes
1:05:31 maybe I asked the same question to two
1:05:33 models got different statistics yeah
1:05:36 well welcome to
1:05:39 hallucinations loving the canvas for
1:05:41 writing
1:05:44 yeah at some point I might dick around
1:05:46 with it for programming I think I'm G to
1:05:48 wait until they they put a web renderer
1:05:51 or yeah web renderer in there so you can
1:05:53 actually see the stuff is there another
1:05:57 good AI video to video other than Runway
1:06:00 everything is image to
1:06:03 video I don't think so the only one I'm
1:06:06 hearing people talking about right now
1:06:08 is is runways video to video that's the
1:06:10 only one I'm seeing so I don't
1:06:20 know how do I do that how do you do what
1:06:23 video to
1:06:24 video I tried code
1:06:27 today I was damn where those agents to
1:06:30 do this for me Jason if you haven't
1:06:32 played with it yet play with um play
1:06:35 with
1:06:36 repet because repet repet not only
1:06:39 writes the code but it generates the
1:06:41 development environment the database and
1:06:44 the web server and just spins up your
1:06:46 application it's pretty
1:06:50 remarkable isn't Adobe doing video to
1:06:53 video they might be I haven't played
1:06:54 with it yet
1:06:58 use AI to write my book
1:07:01 yes yes you should what what you should
1:07:05 do Sunrise is if you don't have um the
1:07:09 paid version of chat GPT you should get
1:07:13 that you should
1:07:16 um use the 01 preview model which is the
1:07:21 reasoning engine have it write your
1:07:24 outline so if you've got no not if
1:07:26 you've got ideas just dump literally
1:07:29 just copy and paste everything into a
1:07:31 chat and then say make sense of this and
1:07:34 write me an outline and let it write you
1:07:37 an outline out of that and it'll think
1:07:40 about it and it'll do it it'll write you
1:07:41 a really good outline and then you can
1:07:43 even in there have it
1:07:45 um have it
1:07:49 uh you can have expand the
1:07:52 outline and then you flip over to gp4
1:07:55 with
1:07:56 canvas and pop that outline into the
1:08:00 canvas and just say okay now write
1:08:02 chapter one right chapter two right
1:08:05 chapter three it's pretty it's pretty
1:08:08 Stunning Pretty
1:08:12 stunning make sure it doesn't sound like
1:08:14 chat gbt yeah use use your existing
1:08:17 writing and things like that for for
1:08:20 tone but I you don't listen you don't
1:08:23 have to have chat GPT write the chapters
1:08:25 but have it do the outlining because
1:08:26 it's staggeringly good at that and then
1:08:28 if you want to handw write them you can
1:08:30 handwrite them how I've been doing it is
1:08:32 I let
1:08:33 them write the first draft of the
1:08:35 chapter and now with canvas because you
1:08:37 can go in and you can edit the piece
1:08:39 Parts I can get it really really close
1:08:41 and then I'll hand edited after
1:08:47 that cam kin what's
1:08:50 happening I wish it would let you
1:08:52 customize GPT canvas the way we want it
1:08:55 they will it it'll get there it's
1:08:57 remember all these tools are being
1:08:59 released early and
1:09:05 janky it sounded like chat GPT 3.5
1:09:09 before now it writes better yeah it
1:09:11 writes much better I think the 01
1:09:13 preview writes really really well
1:09:15 actually what did I miss um just me
1:09:18 shooting the [ __ ] I'm at an airport my
1:09:20 flight was delayed by 2 hours so I'm
1:09:22 just going live cuz I like I'm heading
1:09:25 to New York so I couldn't go live
1:09:27 tonight I'd be too
1:09:46 late he busy here
1:09:50 now I made so very realistic old looking
1:09:53 family type photos with flux Pro .1 and
1:09:56 posted them on the salon oh that's cool
1:09:58 that shirt color that shirt
1:10:00 color oh cover the keyboard I know this
1:10:04 is this is I didn't I didn't wear this
1:10:06 for maximal
1:10:12 readability name a branch of it that is
1:10:15 enhanced but not replaced by Ai and
1:10:18 go a branch of it
1:10:26 that is enhanced but does not replace it
1:10:28 I mean shortterm health
1:10:32 desk but long term I think health desk
1:10:35 gets fully
1:10:40 replaced I think what you're going to
1:10:42 end up with is you're going to end up
1:10:43 with managers you're going to end up
1:10:46 with managers and coordinators that
1:10:48 coordinate all the AI tools so I think
1:10:51 that you'll have it groups
1:10:54 of 5 to 10 people instead
1:10:57 of 50 to
1:11:01 500 so I think I think you'll still have
1:11:03 you know you still got to walk down to
1:11:06 Jim's office and you know take his
1:11:09 laptop off his desk when he is laid
1:11:23 off yeah I love you taking us on this
1:11:25 trip to the airport for a while lots of
1:11:28 places but yeah eventually full
1:11:30 replacement yeah my Help Desk position
1:11:33 was replaced yeah I I
1:11:36 know vom
1:11:39 dude so much stuff like I feel like
1:11:42 customer support in
1:11:44 general goes away essentially entirely
1:11:48 like call
1:11:50 centers I can't see a future where they
1:11:52 don't go away entirely because right now
1:11:55 realtime voice the real-time voice API
1:11:58 is about $18 an
1:12:02 hour a call center employee in India is
1:12:05 about $2 an
1:12:06 hour well the API calls for from open AI
1:12:11 in the past year went
1:12:13 down
1:12:16 100x so if it goes down 100x from $18
1:12:19 that's 18 cents so probably within a
1:12:22 year you're going to have
1:12:26 realtime
1:12:27 voice that knows everything about the
1:12:30 company can answer all the questions
1:12:32 you'll probably have a you know
1:12:34 autonomous agents powering
1:12:38 that probably the ability to do rag
1:12:41 retrieval augmented generation so it's
1:12:44 accurate and it'll probably be less than
1:12:46 $2 an
1:12:48 hour I don't see how that function and
1:12:52 it's going to be a better customer
1:12:53 experience by a lot
1:12:58 what's your customer service you know
1:13:00 your customer call experience these days
1:13:03 set on hold for 45 minutes get someone
1:13:06 who doesn't know anything can't do
1:13:08 anything is maybe calling from like some
1:13:10 internet cafe somewhere in the
1:13:16 world and then you get to talk to
1:13:18 someone who if they do it right will
1:13:21 default to sounding like your ZIP code
1:13:29 and they'll know everything and they'll
1:13:30 be infinitely patient and infinitely
1:13:34 helpful so yeah that goes away entirely
1:13:38 I
1:13:39 think now that might be five years but I
1:13:42 think it's going to be sooner than
1:13:47 that I think the physical part of it
1:13:50 stays until the Bots get good so that's
1:13:54 probably five years
1:14:00 H don't
1:14:01 know cam K hey Kyle I forgot to tell you
1:14:05 my school's implementing an AI student
1:14:07 info desk with an avatar oh that's super
1:14:11 cool that's really cool love
1:14:13 that I've seen a video of Mark CEO of
1:14:16 Facebook said he believes that we have
1:14:18 more agents than people on earth I think
1:14:21 that's right Jason I mean level five AGI
1:14:25 so so open AI has the five levels of AGI
1:14:29 chat Bots agents um or no sorry chatbots
1:14:33 reasoning agents innovators so combine
1:14:38 these to get true Innovation and then
1:14:41 organizations and how Altman and some
1:14:43 others are describing it is
1:14:46 imagine having a corporation of a
1:14:49 thousand employees working for you
1:14:52 that's what we'll all have so yeah I
1:14:55 think I think there will be way more
1:14:56 agents than
1:15:11 people I was thinking instead of 10
1:15:13 admins they only need two for the it
1:15:15 group yeah that seems right to
1:15:16 me you need kind of coordinators right
1:15:19 you need at least two because someone's
1:15:21 going to call in six so you just need
1:15:22 them to be able to cover for each other
1:15:25 and basically what they're going to do
1:15:26 is they're going
1:15:30 to it's going to be their job to keep up
1:15:34 with what the company needs and get the
1:15:38 inputs into the machines but even that's
1:15:39 going to get automated yeah I don't
1:15:43 it's I just I seriously don't think
1:15:45 people are prepared for what's
1:15:49 coming all right what kind of Agents
1:15:51 should we be thinking about making I
1:15:54 think an important one to start playing
1:15:58 with right now is start
1:16:02 making digital twins of yourself start
1:16:06 playing with avatars so you know visual
1:16:09 likenesses visual animations video
1:16:11 avatars things like that and then also
1:16:13 start working with
1:16:17 um creating digital twins of your
1:16:20 personality oops I think
1:16:22 my my gimbal battery is about to die
1:16:39 should I use chat1 mini or preview
1:16:43 um well Mini's faster I mean here's the
1:16:49 thing in a given
1:16:53 day I don't have that many times where I
1:16:56 really need to use 01 at all but I would
1:17:00 say anywhere where you've got like a
1:17:01 multi-step something you need to do I
1:17:04 would start with 01
1:17:06 preview um just because it's bigger and
1:17:09 better but it's slower um if you just
1:17:12 need to knock something out really quick
1:17:13 I'd use mini I think 01 preview you get
1:17:17 30 chats a week and mini you get 50 so
1:17:21 maybe you just use mini as a backup or
1:17:23 if you want it fast use Mini
1:17:26 but I wouldn't I wouldn't think of using
1:17:29 01 as your daily chatbot like if you
1:17:31 just want to if you just want to do
1:17:33 basic chats use
1:17:37 40 for
1:17:39 Omni I'm so pissed off at them for
1:17:42 naming these things so overlapping it's
1:17:45 just stupid it should be called G
1:17:49 GPT GPT reasoning GPT
1:17:54 right R1 would be better than
1:18:00 o1 all
1:18:02 right um what were those five things
1:18:06 again oh not separate okay um the five
1:18:10 levels of AGI this is from open AI the
1:18:12 first level is chat Bots the second
1:18:15 level is reasoning so 01 preview is
1:18:18 reasoning the third level is autonomous
1:18:21 agents I think we see those probably
1:18:23 within a month but but I think
1:18:25 confidently by the end of the year the
1:18:28 fourth level level four AGI is
1:18:30 innovators or
1:18:32 inventors so creating new novel Concepts
1:18:35 and ideas that level is really just
1:18:39 combinations of these applied to
1:18:41 something and then level five is
1:18:44 organizations where you take agents and
1:18:47 innovators and you put them in
1:18:48 organizations that go do something on
1:18:51 your behalf so I think I want to start a
1:18:55 consulting
1:18:57 company and it just goes and does all
1:19:00 the research does all the sales does all
1:19:02 the customer support does it
1:19:12 [Music]
1:19:17 all oh my God what did I miss Mary Mary
1:19:20 you didn't miss much you missed me
1:19:22 waiting for I think my flight board
1:19:27 and
1:19:31 um excuse me do you know what time the
1:19:33 new boarding is supp 15 minutes okay
1:19:37 yeah I got like 15 minutes
1:19:40 so I'll go for another 10 minutes or so
1:19:43 what did you miss you didn't miss much
1:19:44 I'm just shooting the [ __ ]
1:19:56 headed to New York City I am I'm headed
1:19:59 to New York City
1:20:01 tonight meeting in the morning let we
1:20:04 drive to
1:20:23 S only come down
1:20:27 my
1:20:43 not no
1:20:51 booze all right I'm gonna get out of
1:20:53 here
1:20:57 pouring through the comments anytime
1:21:00 you're on at a different time I stop and
1:21:02 drop that's that's sweet Mary Mary no
1:21:04 I'm I'm just doing this because my
1:21:06 flight was delayed and so I figured I'd
1:21:08 just go on spontaneously and just shoot
1:21:10 the [ __ ] so we've been just shooting
1:21:11 this [ __ ] by the way my
1:21:15 name should bring a cartoon character to
1:21:19 mine the one with the red
1:21:23 hat the red
1:21:34 hat I can't remember
1:21:36 it dirty
1:21:39 wa what era is it '90s not Ren and
1:21:42 Stimpy is it Ren and Stimpy it's it's
1:21:44 Ren and Stimpy isn't
1:21:48 it yeah thank you Red Hat red mustache
1:21:52 maybe yeah exactly
1:21:55 are we talking Warner Brothers or
1:21:57 running
1:21:59 s all right I'm gonna
1:22:07 walk all right everybody I'm gonna get
1:22:09 out of here um hope you had fun uh let
1:22:13 me pack up my [ __ ] go grab a snack for
1:22:16 the plane and then I will see you
1:22:19 tomorrow try co-pilots real time voice
1:22:23 when you get a chance okay cool all
1:22:24 right
1:22:25 peace
1:22:27 out early 90s Red Hat red trench
1:22:30 coat I don't
1:22:33 remember [ __ ] now it's going to bug me
1:22:35 all right well keep coming back to this
1:22:36 and you can tell me who it was all right
1:22:37 peace out everybody