May 2, 2024 - With Kyle and Matt Farmer: Insights from Creators & Innovators

Description

Chapters

Transcript

0:29 for for
1:08 trying to figure out how to go live here
1:10 people the live creator has declined it
1:23 uh are we in
1:25 trouble all right did we do it I think
1:28 we did it bamn I think think we did it
1:30 Matt it's like it's like old guys with
1:33 computers we're
1:34 gonna I know
1:36 right I've done about seven streams so
1:39 far of just technical difficulties with
1:41 other creators so I kind of got it
1:42 figured out but we'll see okay good yeah
1:45 this looks good you did it okay I'm
1:47 getting uh I'm getting getting positive
1:49 comments so uh so nice to meet you we've
1:51 never actually talked but uh but I watch
1:53 your stuff all the time so I'm really
1:54 excited you reached out yeah same and I
1:57 had a bunch of my followers be like you
1:59 have to do a live with them and I'm like
2:01 heck yeah any day of the week let's do
2:03 it so I reached out to you um on on that
2:06 side and no it's good it's good to
2:08 finally connect let's do it in a public
2:10 way why not right yeah exactly exactly
2:13 um well so yeah like I think I think
2:16 what might I mean it would be
2:17 interesting to me if nothing else like
2:19 why don't you tell me a little bit about
2:20 your background I'll do yours and then
2:22 maybe we can just sort of navigate
2:24 through like what we're excited about
2:26 what you know like what's going on
2:27 because it's like I feel like we're you
2:29 know we're a year and a half into this
2:32 geni era and it's like it it's starting
2:35 to morph like I feel like there's
2:36 something shifting so I'd love to hear
2:38 where you are with
2:39 it yeah so on my side I so I did
2:43 marketing for about 16 years working
2:45 with very large companies I didn't like
2:47 the marketing world wasn't my place so I
2:50 uh I exited out of my company on that
2:52 side and then conveniently I was trying
2:54 to make an AI version of myself back in
2:56 gpt2 it wasn't quite good enough for it
2:58 cuz I'm like maybe instead of you know
3:00 doing marketing that I don't like I can
3:01 have ai do it for me because I'm like oh
3:03 the decision tree makes sense right so I
3:06 exited and I was kind of looking for my
3:08 next my next thing and then chat GPT
3:10 came out and I'm like okay I didn't
3:12 think we'd have this in our lifetime I
3:14 know but but it's uh it's definitely
3:18 going to be soon at this rate and it's
3:20 getting accelerated every single day so
3:22 I went into the AI influencer side I
3:24 have five kids um so I wanted they I
3:26 wanted them to think I'm cool so I
3:28 decided to jump on Tik Tok and start
3:29 doing videos
3:30 and plus you know I'd always ask myself
3:32 the question Matt if you had a trillion
3:33 dollars after being a degenerate having
3:36 fun and everything what would you do and
3:38 it was really kind of going to teaching
3:39 people on that side I think that's where
3:41 I find you know that's my my power on my
3:44 end so yeah I just basically jumped into
3:47 it I thought you know I was like 90 days
3:49 I'll do 90 videos and 90 days and see
3:52 how I like it and then now we it's like
3:54 a year and three months later and here I
3:56 am that's awesome man that's awesome
3:58 awesome awesome how's the uh have have
4:01 you uh do your kids think you're cool
4:03 yeah they do well the problem is now
4:05 they think they're too cool on that
4:07 because I put them in a couple videos
4:08 and my one son his video got 150k views
4:12 on um this like Story one so he's
4:14 walking he's shreding at school like
4:16 he's the most famous guy out there like
4:19 Dad give me an update how many views do
4:21 I have now and I'm like it's been three
4:23 minutes it's about the
4:26 same that's awesome I've got 23y old
4:29 twin boys and and uh they could not give
4:31 two craps about AI or or think that I'm
4:34 cool although their friends are like oh
4:36 that's he's got a lot of followers
4:38 that's pretty respectable so got
4:40 external validation just not inside the
4:42 house right now um nice so so my
4:45 background is so I I'm a Storyteller by
4:48 training I have a degree in acting I've
4:49 written seven screenplays and then in
4:51 the mid90s I started one of the first um
4:55 uh internet agencies a company called
4:57 agency.com so I was in the in the agency
4:59 world for for a bunch of years um and uh
5:03 I I also right around the time we were
5:06 starting agency.com I I I had also
5:08 started an early online magazine called
5:11 Urban desires and I created a group
5:13 called the worldwide web artist
5:15 Consortium which was essentially anyone
5:17 in New York that was interested in web
5:19 stuff like you know we got together
5:22 every other week and and you know it was
5:25 just a remarkable time in history
5:27 because you know the worldwide web
5:29 everyone was trying to figure it out
5:30 when I started uh we call it whack when
5:33 I started whack the worldwi artist
5:35 Consortium um what my what my fantasy
5:39 was was that you know I'd bring smarter
5:42 people than me into the room which I did
5:44 and that they would know more than me
5:45 and then what what we all learned is
5:47 that everyone was trying to figure it
5:48 out there there were no experts right
5:50 and and it was like that for probably
5:52 two or three or four years where people
5:54 were just trying to figure out what this
5:56 thing was and so I I had a similar
5:59 epiphany when when chat GPT came out I
6:02 had a similar sort of knockup H side the
6:04 head moment of like holy crap the world
6:07 just changed and no one knows about it
6:09 yet so I immediately started the AI
6:10 Salon we we had our first meeting of the
6:13 AI salon on December 7th 2022 so a week
6:16 after chat GPT came out right and I
6:19 started a Tik Tok Channel and it was it
6:21 was really just about you know I think
6:23 sort of um how I'm describing it is
6:25 learning out loud is is going to be
6:27 really important and and just Demis in
6:30 this for as many people as possible as
6:32 I'm learning and so that was kind of the
6:33 mission so we're we're we have a like a
6:36 similar sort of uh uh impetus to to get
6:39 into this but yeah I think it's just
6:42 having been through it with the
6:44 web you know I know that this AI stuff
6:48 is way more profound like way more
6:50 impactful and moving way faster so so
6:53 I'm just like hanging on for dear life
6:55 and just trying to trying to navigate
6:57 you know as as the technology evolves
7:01 yeah that's it and it's evolving fast on
7:04 on like I exponentially every week I
7:07 feel like I missed seven potential
7:09 videos now it's 14 potential now it's
7:12 you know it just keeps going on and on
7:14 like the speed and how fast we're
7:16 getting quality like when mid Journey
7:18 first came out everyone was like oh this
7:20 is garbage and I'm like well this is
7:21 this is version three it's you wait till
7:24 version six and we're on version six now
7:26 and it's like okay that's disrupting now
7:28 let's look at video now let's look at
7:30 the next thing like it's
7:32 [Music]
7:33 unbelievable I didn't see if you did a a
7:36 video on it but if you played with Deco
7:38 here's stations I haven't played with st
7:41 I did Deco here a long time they're a
7:42 great company they basically put
7:44 everything in one and I know you do a
7:46 lot of content those are some of the
7:47 videos that I've watched that you've
7:48 done on that side it seems like Deco
7:50 here seems to be your favorite Well's
7:52 it's well so but I was I was into it for
7:55 a while but then stations came out do do
7:58 you know what stations is have you I
8:00 don't think I've checked it out yet so I
8:02 didn't I didn't understand what the name
8:03 was for it so so basically if you just
8:05 go to The Deco here main site and then
8:08 you click on explore it's you just get a
8:10 feed right like like a normal AI gen
8:12 site but you can click on any one of
8:15 those images and you say create station
8:17 and their metaphor is like a radio
8:19 station so just like a radio station
8:20 just plays music this just generates
8:22 images and it does a an image a second
8:25 just never ending and then any image it
8:28 generates you can turn that into a
8:29 station and and so so for me it's just
8:32 like zero prompting pure intuition just
8:36 like let letting AI sort of do the work
8:39 for you it feels like a shift to me like
8:42 like a technology that represents this
8:45 is the world we're moving into where you
8:47 don't have to be nearly as um
8:49 intentional with how you use it and how
8:51 you deploy it that you know image
8:54 generation as entertainment is is just
8:57 wild to
8:58 experience okay yeah yeah no that's
9:01 that's amazing stable audio did a
9:02 similar thing too they have a live
9:04 YouTube channel where it's just all AI
9:06 generated music 24/7 yeah that you could
9:08 listen to and imagine when it gets tied
9:11 into to like the you know sensors and
9:13 things in in the glasses we're wearing
9:15 where it kind of knows our emotion and
9:16 it's like here let me play You music or
9:18 show you images that you know shift your
9:20 state based on no see I've been thinking
9:23 about that for a while even like the a
9:25 the future is AR so basically you know
9:27 you're you're it's detecting that you're
9:29 depressed and then everyone you look at
9:31 it just deep fakes a smile on their face
9:32 and you're like wow everybody loves me
9:34 the world is a great place even though
9:36 everybody's frowning at you and
9:37 everything else but you know you're
9:38 living in it's making stuff more
9:40 colorful you know it's all painted nice
9:42 Etc but yeah the whole realtime
9:45 generation is going to be a game changer
9:47 and we haven't even really I guess we've
9:49 seen it with this station um but we
9:51 haven't really seen what the potential
9:53 is going to be on that we're so early
9:55 it's crazy like I I that when I look at
9:59 um station from Deco here what what
10:02 where where my mind immediately goes is
10:03 I swap out image generation with every
10:06 other kind of multimodal image
10:08 generation and so imagine a system like
10:10 that where you could you know fluidly
10:14 navigate between okay turn that into a
10:16 poem Oh make that a song Oh Let's make
10:18 that a video oh I want to see a bunch of
10:20 images like like where you're just GNA
10:22 uh you know I I want to go into how
10:24 would I start a business around this
10:25 idea and and you know then out will come
10:27 problem solving kind of stuff and and
10:30 just our ability to kind of navigate
10:33 knowledge space you know as as easily as
10:36 you can just watch those images being
10:37 generated that's where I think we're
10:39 heading and it's like what is that skill
10:42 set like like right now like people talk
10:45 about you got to learn prompt
10:46 engineering and structured this and
10:48 structured that well what if you move
10:50 into a world where it's about intuition
10:52 and curation and and I think that's
10:54 going to be an option that people have
10:56 for how they interact with this stuff
10:58 yeah and I think on my side I've been
11:00 doing a lot I'm working with a company
11:02 right now who wants to help people reach
11:04 all their goals and I think another like
11:06 model that we'll see on our side is
11:08 we're dealing with a lot of stuff like
11:09 image generators everything is we're
11:11 doing the input and waiting for an
11:12 output where the proactive side really
11:15 hasn't been cracked yet where you have
11:16 an AI like a C3PO that's basically every
11:19 morning being like hey Matt don't forget
11:21 you wanted to learn Mandarin um and you
11:24 said you know you you really want to
11:25 achieve this goals Etc and having that
11:28 voice as well from the AI when we start
11:30 communicating it with that way it's
11:32 going to be nuts because even using chat
11:33 GPT with voice kind of feels different
11:36 and when the AI starts talking to us and
11:39 then we have the curation on top of it
11:41 so it's like Matt you have this goal so
11:43 I put this lesson together for you yep
11:45 um it you know it gave you that
11:47 inspiration on that side and we it knows
11:49 how to motivate you yeah yeah I'm
11:51 experimenting right now I'm part of a
11:53 group called content Evolution I'm I'm
11:55 the chief generative officer there I
11:57 think I'm the world's first Chief
11:58 generative officer but
12:01 but um what what we built it's so it's a
12:04 group of marketing organizations and so
12:06 you know that the're kind of co-market
12:07 for one another and and it's people that
12:09 are just you know very accomplished have
12:10 written books things like that and so we
12:12 created a group within it an AI kind of
12:15 Skunk Works called uh content Evolution
12:17 collab and we've created digital twins
12:20 of of all of the members of collab and
12:24 so we now have gpts we we created a
12:26 structured interview where we
12:28 interviewed people you know sort of on
12:30 who they are what's their life view what
12:32 do they think about the future how do
12:33 they solve problems what have they done
12:35 for work and and now we have a GPT for
12:37 each one of us and and so where we're
12:40 shifting right now is okay we have these
12:43 digital twins what do we do with them
12:45 how do we interact with them what what
12:47 are other people going to interact with
12:49 them and then as the capabilities come
12:51 along where you can add video to that
12:53 and audio to that and all this other
12:56 stuff you know that idea of kind of
12:58 exploring you know
13:01 designing how you're going to be
13:02 represented whether it's the metaverse
13:04 or in knowledge space or anything like
13:06 that is just Bonkers like it's just
13:12 crazy no it's it's absolutely nuts and
13:15 then you wonder if it's a if you have
13:16 your own GPT like the Matt subconscious
13:19 to some extent you know you have your
13:20 own kind of subconscious where you're
13:22 talking to yourself and making sure
13:24 you're aligning even with yourself too
13:26 at the same time like there's there's so
13:28 many levels to it and it's like there's
13:30 no the craziest part about this industry
13:32 and even just trying to build anything
13:34 alongside of it is there's nothing
13:36 really been done in this space yet so
13:38 it's just like okay we have all these
13:39 gpts now what the what's the use case
13:41 for it how can we apply it like okay we
13:43 could generate perfect images now what
13:45 what can we do with that now with video
13:47 and audio it's nuts I I know and that's
13:52 I don't know like I don't know where
13:53 where you are like I I there's a there's
13:56 a weird malaise right now in the
13:58 industry for me and it might just be for
14:00 me but I feel like a lot of the other
14:04 large language models have caught up to
14:06 gp4 right so we have we have five or six
14:09 now that are sort of hovering around GPT
14:11 4 and then we're all waiting for Sam to
14:13 drop GPT 5 right which we assume is
14:15 going to be way up here and so like I'm
14:18 in this place where I'm like I don't
14:20 really know which large language model
14:22 to put any time into because they're all
14:23 kind of you know it's kind of like
14:25 computers are kind of good enough now I
14:26 feel like large language models are all
14:28 kind of relatively at the same
14:30 capability and you know some are a
14:32 little better than others but they're
14:33 all basically good but I just I feel
14:35 like we're in this this sort of moment
14:38 of pause before shit's about to hit the
14:41 fan no I I definitely agree with you
14:44 because I do um I actually have an
14:46 agency so I do like implementations for
14:48 business and stuff and I think it's
14:50 really like a lot of the the
14:52 capabilities are already there to do so
14:53 much let's say for example it's never it
14:56 never increases we never get anything
14:58 past GPT 4 with what we have with gp4
15:01 you can almost build everything like
15:03 there's not much it's missing like maybe
15:05 on the math side and like the the truly
15:08 planning like planning and math is what
15:09 they're truly trying to crack on this
15:11 end but I feel like it's more going to
15:13 be the tools and applications and
15:15 everything else and you almost see it
15:16 like it's really interesting like
15:18 Microsoft release co-pilot and it's
15:20 pretty much chat GPT but nobody really
15:22 uses it because it's it's you're more
15:24 using the application of it of like chat
15:26 I have both and I never use co-pilot I
15:28 go to chat PT and I use it there where
15:31 it's like the applications are really
15:33 going to push the next part of it of
15:35 just like okay how cuz that's what we
15:37 that's the unknown right now it's like
15:38 how are we going to use the the all this
15:41 application this technology like where
15:43 does it actually fit in and at this even
15:45 at the rate of Technology that's
15:47 increasing some people are almost like
15:49 holding back a little bit because
15:50 they're like I don't want to do that
15:51 because agents are coming out and then
15:53 agents are going to take care of this
15:55 and like and it's the question of what
15:57 are agents going to do and are they
15:58 going to be good is that the way going
16:00 forward do you follow do you follow
16:02 Ethan
16:03 mollik I don't believe so okay you
16:06 should you should follow won business
16:07 school Professor he wrote an article
16:09 like a month ago was I forget the exact
16:12 title I think it was the the tyranny of
16:14 the of the weight calculation or
16:17 something like that and it's basically
16:19 um you know if you were to get in a
16:22 rocket today and and and travel to our
16:24 nearest star it would take you 20,000
16:25 years but if you wait for a 100 Years
16:27 it'll only take take you 50
16:31 and and but in that hundred years of
16:32 waiting you've missed all this
16:33 opportunity and he said it's the exact
16:35 same with AI right now that that you
16:37 shouldn't be waiting because because as
16:40 you wait there's all this opportunity
16:41 for what these tools do right now that's
16:44 that's like truly remarkable and
16:46 mind-bending like I one of the things
16:48 that I keep I just keep having this
16:50 palpable sense of of feeling like you
16:54 know either we're living in a post-
16:55 Sci-Fi world where all the Sci-Fi movies
16:57 of our past are becoming document so
16:59 like oh yeah we can do that now you know
17:02 or or it's just like we're we're living
17:04 in you know some weird sci-fi movie
17:06 where we can just we're starting to have
17:08 these increasing capabilities and it's
17:10 it's just it's totally
17:12 surreal no it's it's crazy and they even
17:14 think with this new AI technology we're
17:16 going to increase lifespans and they
17:18 think by 2030 we could potentially get
17:21 significant like a two to 300 year
17:23 increase on lifespan which will give us
17:25 200 to 300 years to figure out how to
17:27 make it a thousand and like to some
17:29 extent it's like you know we could we
17:31 could be the oldest people in the world
17:32 for a long time if this technology goes
17:35 the way that everybody thinks it's going
17:36 right now oh that's true yeah like like
17:38 it it redefines OG doesn't it yeah
17:42 exactly we what will be the oldtimers
17:44 you know old G
17:47 guard yeah how many O's do you have in
17:50 front of your
17:52 G no exactly that's interesting um no
17:57 it's good man there's so many tools
17:59 being released all the time it's hard to
18:00 even keep up with them it seems like
18:01 there's and then it's also separating
18:04 who is um who's the Cherry Pickers like
18:06 who's who's releasing technology where
18:08 it's like no it's not actually like that
18:10 and the other ones that are like okay
18:12 this is actually you know they're
18:13 showing us real use case it's kind of
18:15 like how Devon came out I'm not sure if
18:16 you followed Devon and then it turned
18:19 out Devon where well it's you know but
18:21 they just raised a ton of money but yeah
18:22 it's a six-month old company with no
18:24 technology but they had a good demo reel
18:27 I'm I'm feeling similar I I'm I'm in a
18:30 bit of a pet peeve with a lot of a lot
18:32 of companies right now where like you
18:34 know Sor was announced and then it's
18:36 like you know no Clarity on when it's
18:38 going to be out right they just they
18:40 just had to get something out there um
18:43 we just saw the a similar video model
18:45 from China um VD yeah vdu there was emo
18:50 emo from someone else was announced
18:52 Devon was one of those there's a bunch
18:54 of these where they're putting out slick
18:55 videos but then it's like you know
18:57 crickets on when we're going to see it
18:59 and it's like just launch the cuz
19:01 there's plenty for us to play with we we
19:03 don't need to know that yeah something
19:05 Cool's coming cuz it's you know I think
19:07 I think Google was the first one where
19:10 where you know their their Gemini Ultra
19:12 you know thing which they got a lot of
19:14 for but but um like I want to like
19:19 I want to play with some of those tools
19:22 that they showed like the dynamic
19:23 interfaces that were being generated and
19:25 it's like you know they they showed them
19:28 but just you know now it's crickets and
19:30 we're just waiting and it's like just
19:32 just give us access to stuff no that's
19:35 it it's all they're trying to pump up
19:37 stocks or something like that but
19:38 there's so many of those that are coming
19:40 out right now one of the ones that I was
19:41 surprised and I saw you did a video on
19:43 it um LTX Studio I got access to it
19:46 actually really good yeah it's it's
19:48 surprisingly good I mean you know you
19:50 know what's fascinating about it Matt is
19:52 that if you look at what they're doing
19:55 technically there early on um there
19:59 there were a lot of people were talking
20:00 about oh RP wrapper apps are um are kind
20:04 of a waste of time where where you're
20:05 sort of putting a pre- prompted app
20:07 together that's just using you know chat
20:09 GPT as the back end and and that's what
20:12 LTX studio is to a large degree is just
20:14 a whole series of prompts very well
20:17 structured but it's very well structured
20:20 like if like I look at LTX as kind of a
20:23 prototype for you know the kinds of
20:26 sophisticated agents where we're going
20:28 to have where if you swap out
20:29 storytelling and film making for project
20:32 management or for code development or
20:34 for Consulting and you have a similar
20:36 kind of tool as LTX where you have deep
20:39 you know subject matter experts putting
20:40 these systems together you know we're
20:42 going to have incredibly sophisticated
20:44 tools that that can just you know you
20:46 literally give it a prompt and then it
20:48 just goes off and you know it's a it's a
20:50 chain of events one after another uh in
20:53 a specific domain I think that's going
20:54 to be incredibly powerful so yeah I I'm
20:58 impressed with it and and one of the
21:00 things I say on my channel a lot is
21:01 embrace the Jank like it's janky they're
21:04 using a crappy stable diffusion image
21:06 generator right and and you know a
21:08 crappy animation tool or you know video
21:10 tool but it's pretty damn good for for
21:14 the sort of low-level Tech that's in it
21:16 right
21:17 now no I agree and it is I felt like it
21:20 was a mix of all like okay here's this
21:22 face swap technology where you're able
21:24 to put your own face in it the voice
21:26 it's like the voice is okay but yeah
21:28 you're completely right it's just like
21:30 they've and it's almost how it's like
21:31 chained together they're just chaining
21:33 all these Technologies together making
21:35 like some kind of a pipeline and that
21:37 pipeline creates something that's quite
21:38 good in the end and it's also you know I
21:41 I also like to give these apps a lot of
21:43 credit even if it's not the best it's
21:45 not I'm not looking at what it is now
21:46 but I'm like okay what is it in two
21:48 generations from now three generations
21:50 where it may not be perfect at this
21:52 point and I'm fine with that I'll give
21:54 them a good review regardless for that
21:56 I'm I'm so I'm so there I just I don't
21:58 know if you gotten yours yet or if
22:00 you've ordered one but but the rabbit is
22:02 getting all this hate where people like
22:04 it's a grift and it you know it doesn't
22:07 wash my clothes for me and I hate it
22:11 it's like no it's like it's an
22:13 incredibly it's a beautifully
22:15 well-designed product with like cool
22:17 Nostalgia and a and a walkie-talkie
22:19 button and the software is crap right
22:22 now but it's like you're old enough to
22:25 remember remember when the iPhone first
22:27 came out and and they talked about these
22:29 things called apps but they weren't
22:31 really apps they were just shitty
22:32 websites and it's like that's that's at
22:34 the stage this is at right where like
22:36 the software right now is a little weak
22:39 but it's like what I give them huge
22:41 credit for is at least they're trying
22:43 right they're envisioning a future where
22:46 literally a single button can give you
22:49 the you know access to all of the
22:51 technology and knowledge of the world
22:53 right that's that's the sort of Promise
22:54 of AI is it becomes this simplification
22:57 layer and to me this represents kind of
22:59 the first tangible product of someone
23:02 thinking like that and so I give this a
23:04 lot of Grace and it's it's it's way
23:07 better uh constructed and built than I
23:09 thought it would be like it's it's good
23:12 yeah no and I like the price point they
23:13 really went out of reasonable price
23:15 point $200 I didn't get one just because
23:18 I'm in Canada so there's some Nightmares
23:20 of the converting over there you know uh
23:22 so that's the only reason why I'd love
23:24 to try one but yeah I'm the same way
23:26 where it's like you're looking at the
23:28 potential of it not the current state
23:29 because of course even agent agents
23:31 aren't the best right now only because
23:33 they don't have tools so like agents can
23:36 do stuff and all the demos are the same
23:37 look up the financial information of
23:39 Nvidia versus Adobe and charted over the
23:42 last five years blah you know and it's
23:44 like yeah I could do that but the thing
23:45 is it's only because it's got that tools
23:47 so on my side I'm working on just
23:49 creating tools for the agent right now
23:51 because that completely changes how they
23:53 work so on the marketing side I made a
23:56 tool to basically extract like from the
23:58 Facebook ad librar so now I could have
24:00 an agent do a smart competitor research
24:02 for stuff like that so I think it's like
24:04 they're more it's it's so early
24:06 technology and it's same with rabbit
24:08 they don't have the a lot of the apps
24:09 aren't released yet because they don't
24:10 have the tools made for it yet and the
24:12 tools are just basically being like all
24:14 right the llm can now work with this
24:17 because it can't all these agents they
24:19 don't work with Google Docs or any of
24:20 the really important stuff that you
24:22 actually need to get stuff done um but
24:24 once they do that's when it's going to
24:27 be absolutely amazing and where I you
24:29 know I joke I'm like we're about a th000
24:30 tools away from it being good but a
24:32 th000 tools isn't that much yeah exact
24:35 well yeah cuz there you know 10 are
24:36 coming out a day so it's not it's not
24:38 taking us that long um yeah no yeah the
24:42 agent thing is going to be really
24:43 interesting you know when when uh when
24:45 chat GPT has reasoning there there was
24:47 that uh the little Kur fuffle last week
24:50 with gpt2 chatbot that yeah the chat
24:53 gbt2 J chatbot
24:56 um I heard a rumor I don't know if you
24:59 heard this that um open AI is launching
25:02 a search engine next Thursday yes on the
25:04 9th so um that seems to be more legit
25:08 than the gp2 gpt2 chatbot that's just
25:11 kind of unknown and it's amazing how
25:14 fast I didn't report on it just because
25:16 I'm like uh how old this video age
25:18 potentially not the best is it worth
25:21 worth the effort I tried it out so it's
25:23 it's better than other llms at most
25:25 things and it's not and everyone was
25:27 disappointed like oh oh it's it's just a
25:29 little bit better than GPT 4 but I don't
25:31 think GPT 5 that's where we're going to
25:34 see the gains I think it's more the
25:35 softer Suite they're just going to be
25:37 embedding agents inside of it to some
25:39 extent so instead of like you can make
25:42 gbt 3.5 L perform gbt 4 by using an
25:45 agenic framework for it like just
25:47 because the whole sequence is anytime
25:50 and I tried to do it once I made some
25:51 seating chart um Creator where there's
25:54 an event had 150 people and if you try
25:56 to do it at the beginning you give it a
25:58 list and like make me a a seating chart
26:00 there's too much reasoning built in so
26:02 it's like for every step of reasoning
26:04 you build into a prompt you're getting
26:05 significantly lower results on it so we
26:08 basically created like a lang chain
26:10 where it do it sequentially where it's
26:11 like okay here's the list of people
26:13 create one table so we create one table
26:15 then we'd remove the name from the list
26:17 and then it' be like create table two
26:18 and tell us why um so I think just
26:21 adding that additional techn will be the
26:23 reason why GPT 5 so I think they're just
26:25 going to release like a GPT 4.5 equalent
26:29 it'll be slightly better but when they
26:31 add the agent on top of it I think
26:32 that's when we're going to see that's
26:34 when it's going to blow our mind like
26:35 what it could do that when really
26:37 realistically it's just kind of the you
26:39 know the pipeline that it's similar to
26:41 what LTX created they're just creating a
26:43 better pipeline of prompts on that side
26:45 to to query the model yeah yeah the the
26:49 one the one uh the one chat TMZ uh
26:53 speculation that I heard about that was
26:56 that was that it was chat GP it was GPT
26:59 2 but they had added qar to it and
27:02 that's what got it the gp4 level stuff
27:04 so if if that's actually the case then
27:06 it's really interesting but but it's
27:08 exactly what you said that if they had
27:10 reasoning to even a a less powerful you
27:13 know core model then then all of a
27:15 sudden it gets way better that's
27:16 essentially what you were just talking
27:18 about so that's my fantasy right now is
27:20 that that's what that thing was so but
27:22 we'll see yeah I just at this point I'm
27:24 like I just want them to release I don't
27:26 care what comes with it is it is it
27:28 going to be good bad Etc it's like let's
27:30 see what the new one is and then we
27:31 could get tokens a bit cheaper on gbd4
27:34 that would be nice cuz now llama 3 you
27:36 could use it free through gr like with
27:39 grock and it's super high speed and I'm
27:42 like okay do I move everything over on
27:44 this side like I'm running all my agents
27:46 with Gro because llama 3 is good enough
27:47 for most of it it's just like it's it's
27:50 commoditizing these llms and I can't
27:52 believe meta Zuckerberg going through
27:54 and doing that um they talked about it
27:56 on the all-in podcast of you know it's
27:59 it's basically when you're behind inet
28:01 technology you open source it So Meta
28:03 was number one at Facebook and they lock
28:05 their API and nobody could get access to
28:07 it and now they're behind an AI so
28:09 they're like all right let's open source
28:11 it and they already have it up to a
28:12 million tokens there's a project right
28:14 now that has a million tokens with um
28:17 needle in a hay stack scoring all the
28:19 way up to like 900k tokens perfect um
28:22 attribution on all of that so it's like
28:24 crazy it's just amazing it it really is
28:27 it yeah
28:29 I don't you know it's funny ear early on
28:31 when when chat GPT first came out Alman
28:34 was talking about
28:36 um he believes in you know releasing
28:40 often but but you know kind of delaying
28:43 putting things out there
28:46 because I forget how he he exactly
28:48 phrased it basically people need to prep
28:51 for what's coming like they need to
28:53 understand what's coming and so
28:55 trickling out the features a little bit
28:57 at a time and and maybe this is a
28:59 justification for announcing Sora but
29:01 not really having it out in the world is
29:04 um I don't know if you heard the the uh
29:06 Tyler Perry when he saw Sora he
29:09 cancelled an $800 million Studio
29:11 Expansion Project for his Film Studio in
29:14 Atlanta $800 million construction
29:16 project he looked at that Tech and just
29:18 went nope not going to do it so I think
29:20 there really is something to putting you
29:24 know putting some new thing out there
29:26 that really will change the game let
29:28 people assimilate it a little bit and
29:30 get used to it a little bit and then do
29:31 the next one because I I I it really
29:34 could be completely disruptive if you
29:37 just sort of go from nothing to just it
29:39 can do everything like I don't even like
29:42 people don't how people don't know how
29:43 to even think about basic chat GPT right
29:46 now much less multimodal chat GPT 4
29:49 right like it's like that jump is is
29:52 massive and it's taken us a year to kind
29:54 of get our head around how we even deal
29:56 with that so I don't know I yeah no I I
30:00 agree with that it it takes time it it
30:03 takes time for people to adjust and even
30:05 the tools to be made for it because it's
30:07 like we haven't even if they did GPT 5 6
30:10 months ago and we're on GPT 6 it'd be so
30:13 far beyond what we're even able to use
30:14 it for and it yeah they are they're the
30:17 rap um the rapper Killers too at the
30:19 same time so I think they're being a
30:21 little bit wey of that because I think
30:23 they're just going to try to kill
30:24 everybody at some point you know he
30:25 pretty much said it's like yeah you know
30:28 it mayen happened and it's like yeah
30:30 it's definitely going to happen cuz you
30:31 see they they have an audio product
30:33 that's coming out soon um on that side
30:36 they have they have all this stuff and
30:37 I'm just like okay and then Sora Sora I
30:40 had um a bunch of video companies reach
30:42 out to me and they're like oh yeah can
30:44 we do a video together Etc Sora launched
30:47 and they all canceled at the same time
30:49 they're like well we give up you know
30:51 delete them we're done it's over well
30:55 the thing the thing that was most
30:57 exciting to me about Sora is is the
30:59 approach that the the very first
31:02 sentence of the of the announcement of
31:05 Sora was that you know Sora is a system
31:07 that understands and simulates the real
31:10 world it's not a video tool right it's
31:13 it's a video tool that understands oops
31:16 I just spilled the drink it's a video
31:17 tool that understands physics right and
31:19 you're like oh like that's just a
31:23 completely different
31:24 conversation well it's like UD you know
31:27 udio the the AI song creator that came
31:30 out it turns out Killer song last night
31:32 on my live it was so much fun yeah know
31:35 I love that one but it's not a a music
31:37 generator it's a sound generator they
31:39 were able to make it create standup
31:41 comedy and they're also able to make it
31:43 generate sounds too at the same time so
31:45 it doesn't just understand music it's an
31:48 entire sound generator and even how
31:49 these llms work is amazing just how it's
31:52 like oh we fed up more data and now it
31:54 knows physics and then we fed up more
31:56 data and now it understands reasoning
31:58 and then a bit more data and they never
31:59 know when these points are going to come
32:01 they're just like huh look at that it
32:04 understands that now oh it understands
32:06 something we didn't teach it it's like
32:08 yeah yeah well that's you know with the
32:11 if with ha jens's you know my company
32:14 Story vines a a video storytelling thing
32:16 and I've been playing around a lot with
32:18 hay Jen's translation tools and it's
32:20 like as humans you know as a human who
32:23 speaks English you know I can look at
32:25 Spanish and Spanish is easier for me to
32:28 understand then say Mandarin Chinese
32:30 because it's just a completely different
32:31 thing but to the machine learning models
32:35 it like it doesn't it's just as easy for
32:38 it for it to learn Mandarin as it is to
32:40 learn Spanish and that's just
32:42 mindblowing that like to your point you
32:44 just dump different data in it it learns
32:46 it and now you've got access to that and
32:48 that's just that's just you know I
32:51 that's that's the thing that I think
32:52 changes everything right that that it
32:55 used to be if you wanted to get good at
32:57 something you need needed to go back to
32:59 school or you needed to put your 10,000
33:01 hours in as Malcolm Gladwell would say
33:04 and now you can just have access to it
33:07 or you know that's where we're headed
33:09 and that that to me is just you
33:12 know yeah no it's it's crazy and then
33:14 there's the whole AGI conversation which
33:16 I find it's hard to get the full
33:18 definition of that like that's where
33:19 people really struggle because a lot of
33:21 what people talk about is ASI and that's
33:23 almost what I consider an AGI it's like
33:26 when is AI better than the best at every
33:29 single subject and I feel like we're not
33:32 that far away from it because at that
33:33 point that's when you can use AI to
33:35 train and make AI better and that's when
33:38 we hit the flywheel the flywheel is ABS
33:40 like as soon as we hit that point and
33:42 that's where qar got a little bit crazy
33:44 where it was rumor like oh it made its
33:45 own type of math it's better than ours
33:47 and I'm like well it's that's actually
33:49 believable yeah um but it's really it's
33:52 like how far are we from that and it's
33:53 kind of yeah the last 5 10% may be the
33:56 hardest part and we may 90% but that 10%
33:59 is going to take us years but as soon as
34:01 we get to that point where AI is able to
34:04 train AI better than we are that's when
34:06 it's able to do it 3,000 times an hour
34:08 and then that's when we go into Skynet
34:11 or AGI or whatever whatever the next
34:13 step is the unknown there's there's also
34:16 a lot of conversation about semantics
34:19 and and they kind of drive me crazy like
34:22 first thing I thought when I when I when
34:24 I played with chat GPT was oh the Turing
34:27 test has been soled like I immediately
34:29 went to the test Sol and then there were
34:31 like months of people going well the
34:33 real touring test you for 5 hours and
34:36 like they're they're they're sort of
34:37 moving the goalpost right and I feel
34:39 like you know AGI is similar where you
34:41 know I guess AGI is you know better than
34:44 the median human at most jobs or
34:46 whatever it is right and I'm like Well
34:48 Chad GPT already knows significantly
34:51 more than I ever will like buy thousands
34:54 like it's it's already way past where I
34:56 am in my opinion now now can it actually
34:58 execute my job and have judgment and
35:00 reasoning okay it doesn't have that
35:02 piece right but but just in terms of
35:04 access to knowledge it is significantly
35:07 Beyond me so now you add in a reasoning
35:09 and logic layer to that and yeah it's
35:12 absolutely you know miles beyond what
35:14 I'm ever going to be capable of doing
35:17 but we all have access to it so that's
35:20 that's again the thing I get really
35:21 excited about is yes it can do that but
35:24 we get to direct those tools right and
35:26 that's that's where you know that's for
35:29 now at
35:30 least yeah
35:32 exactly no it's um no I completely agree
35:36 where it's like because that's what they
35:37 say it's like the 50% median and I'm
35:39 kind of like well I feel like it's
35:40 beyond that and that's the whole like
35:42 and you know one downside of humans is
35:44 we'll never agree on anything um to any
35:47 extent and that's why it's like we don't
35:48 have these definitions and we Pro we'll
35:50 be long into ASI and like Singularity
35:53 before even humans can can figure it out
35:56 I'm I'm a pretty big optimist in general
35:58 for all this technology and everything
36:00 so you know everybody's always saying
36:02 like oh AI is going to create all these
36:04 problems and I kind of think if we make
36:06 a super intelligence that it has a good
36:08 shot at being smart enough to solve the
36:10 own Pro its own problems that it creates
36:13 it kind of beyond that and that's one
36:14 thing that humans have never been the
36:16 best at like we got intelligence locked
36:18 down pretty good uh solving our problems
36:20 of course we have solved quite a bit of
36:22 them but for it to just do like a a
36:24 clean sweep and be like okay all the
36:26 problems are solved
36:28 yeah yeah let's go fix it yeah no I you
36:32 know I'm an optimist as well I I feel
36:34 like
36:35 it's you know if if if you look at
36:38 history every single technology that's
36:41 been significant has been met with
36:44 resistance and riots and violence you
36:48 know the printing press the steam engine
36:50 the flipping Loom you know uh there were
36:54 RI the loom right you know yeah so every
36:57 single Tech technology so it's not
36:58 surprising to me we also have you know
37:01 50 years of Hollywood telling us the
37:03 robots are going to kill us right so so
37:05 I think there's a natural fear plus
37:07 there's just that fear of a new
37:09 technology but what I'm experiencing in
37:12 the AI salon and in the AI learning lab
37:16 is the the the fear that people have
37:20 almost always like with very few
37:23 exceptions is people that haven't tried
37:26 it and what I am experiencing is that
37:28 the minute people try it they're kind of
37:30 reintroduced to their humanity and like
37:33 what what they can what they can do now
37:36 with these new tools and I'm just
37:38 watching people reinvent themselves and
37:40 ReDiscover their passions and and you
37:43 know start businesses and it's it's
37:45 absolutely flipping inspiring right uh
37:48 for people that actually try it and so
37:50 that's like my mission is just similar
37:52 to yours like just you know learn about
37:54 this stuff try this stuff get curious
37:56 about it because the you get in touch
37:58 with this thing it's just like boom this
38:00 world opens up and it's
38:03 remarkable no it is and I think there's
38:05 a lot of fear and I talk to a lot of
38:07 artists about it and everything else
38:08 it's I truly view it as a tool of
38:11 Leverage and like someone who's very
38:13 artistic will always beat me in AI image
38:15 generation if we were to go head-to-head
38:18 because they know what good art is and I
38:20 don't and same with music generation and
38:22 even from the marketing side copywriting
38:24 you can't write copy with the best large
38:27 language model if you don't know what
38:29 good marketing copy is you have to have
38:31 the base knowledge so just you have to
38:33 view it as a tool of okay I've got
38:35 knowledgeable in this field and now I
38:37 have this tool that can make me a
38:38 hundred times more productive on that
38:40 side and create a hundred times more
38:42 pieces of art or even just as an artist
38:44 train to model on all of their art and
38:46 then in the future everybody's going to
38:48 be able to do use your generator or your
38:51 image model to create more versions of
38:53 your art that are customized to them
38:55 like it is it's truly a leverage thing
38:57 and not just something where it's like
38:58 okay well my career is dead it's like no
39:01 you just got some more time on the beach
39:04 now you don't have to sit there in Ms
39:05 paint just going away being like well is
39:07 this good Etc it's like and I have a
39:10 rule if any task takes me over five
39:12 minutes I try to find a way to have ai
39:14 replace that or automated or or do do
39:17 anything on that end so I'm yeah I'm
39:19 extremely optimistic on that side where
39:21 and it's going to be a long time before
39:23 someone with no knowledge of art is able
39:26 to go in there and beat a true artist at
39:28 AI image generation like they do have
39:30 the advantage I look at some of my
39:32 Generations versus like the Community
39:33 Feed and I'm like wow these guys are
39:36 amazing like this is
39:38 unbelievable that's awesome that's
39:40 awesome yeah yeah I don't know man I
39:45 um yeah I'm I'm really hopeful about it
39:48 what what are the things that you think
39:50 what you know what are the things coming
39:51 up that you're excited about is there
39:53 anything that you're on pins and needles
39:55 about is it all GPT 5 for you like
39:58 what's what's the stuff that you're you
40:01 know you're impatient for I think the
40:04 video stuff um especially even just
40:07 helping I'm not a v I learned video
40:09 editing myself and I'm not a good video
40:11 editor so just more tools that are going
40:14 to actually help with that adobe has
40:15 their new stuff where you can put
40:16 generative inside of it I think that's
40:19 going to be pretty cool um definitely
40:22 just the AI video generation getting
40:24 that to the point of say mid journey and
40:26 everything else once we truly figure out
40:29 text and AI generators like idiogram is
40:31 a massive leap for it but it's still for
40:33 any kind of complex text you're looking
40:34 at maybe two of the four Images are
40:37 correct for it and there's limitations
40:39 um on that side so that's pretty
40:42 interesting um the song stuff is really
40:45 cool too I know a company they're called
40:47 son AO they're YC combinator I found
40:50 twers team they made it most um of these
40:53 these song generators are um they're
40:56 actually using language models and but
40:59 he's using a diffusion model which
41:01 actually is going to let you go in and
41:02 prompt edit the entire song after it's
41:05 been created so just because like
41:07 instead of you know udio you have very
41:09 limited edits that you could do of
41:11 course you do the lyrics and stuff but
41:12 you're like hey I want to trombone here
41:14 and like drum solo drum solo let's
41:18 go but just like expansion of these
41:21 Technologies on top of it too because I
41:23 also you know I see it for my kids my
41:25 oldest just turned 10 um and you know I
41:28 got five of them so it's just even
41:29 seeing how they're using these tools and
41:31 giving them the capabilities to do stuff
41:33 like create their own stories and videos
41:35 and seeing them do so just more and more
41:37 that coming to the consumer level at
41:39 reasonable prices I still think some of
41:42 it's a little bit expensive my AI bills
41:44 can get a bit expensive but yeah other
41:46 than like gp5 it's good to see all that
41:48 stuff come out but I think really I'm
41:51 excited about just the different models
41:54 and like the like I said the proactive
41:56 chatbot I think think when somebody
41:58 nails that inflection tried to do it and
42:00 like it's still good and you're still
42:01 kind of talking to it but you know where
42:03 we have our true either second brain or
42:06 a robot assistant or anything like that
42:08 and then of course I'm all in humanoid
42:10 robots I just started doing videos on
42:12 those because I want one I'm like
42:14 sponsor me send me a robot I don't want
42:16 to ever do the dishes again cook clean
42:19 well I can give that to robots that's
42:21 fine that's awesome we on my on my Live
42:24 a month or two ago U we were talking
42:27 about the the robots and we actually
42:29 came up with you know I've written
42:30 screenplays so I'm always sort of
42:32 thinking stories and uh we we came up
42:35 with with a a movie idea um that um four
42:40 different people are given early
42:42 versions of humanoid robots they win it
42:44 like in a grocery store contest that's
42:46 in the town of the robot manufacturer
42:49 and then and then it's all the wacky hij
42:50 Jinks with with the 1.0 version of the
42:53 inhome robot right where you know the
42:56 minute the people leave the robots kind
42:58 of go
42:59 nuts you know no I think like that's a
43:02 whole another and even on the the
43:04 medical side I'm really excited to see
43:06 just more medical breakthroughs and
43:08 everything I think that's going to help
43:09 shine a good light on AI and what
43:11 they're doing with even stuff like
43:13 because we have ASI and stuff like Alpha
43:15 F fold or you know the chess and Etc
43:18 like we're at that top Beyond humans in
43:20 a bunch of those fields but they've been
43:22 very closed doors very specific like
43:24 Google deepmind basically takes care of
43:26 all of that stuff and I'm really excited
43:29 for some of that even just the fact that
43:31 they're able to generate all these new
43:32 materials um on on their end where it's
43:35 like oh we had 13 before but we found 87
43:38 new ones that could potentially change
43:40 all of humanity um within a short period
43:43 of time it was three 380,000 new stable
43:47 crystals they found and humans up to
43:51 that point had discovered 22,000 and
43:54 then using computers had found another
43:56 44,000 so so like humans in computers up
43:59 to that point it was like 66,000 right
44:02 and then overnight here's 380,000 stable
44:06 crystals like I I mean you know like the
44:09 the the implications and impact of that
44:12 we probably won't know for five or 10
44:14 years but holy crap right like CU in
44:18 there is a battery that charges in 30
44:20 seconds that lasts for a day right you
44:22 know in that is you know you know easy
44:25 electri you know electricity trans
44:27 Mission all you know all just all the
44:29 sort of stuff is probably sitting in one
44:31 of those crystals that we have yet to
44:34 sort of bring to the world it's just
44:36 staggering yeah no that's that's it like
44:38 those are really going to be some of the
44:40 big ones of course on the tool side I'll
44:41 take anything I could get give me even a
44:43 better version of mid Journey I'm happy
44:45 you know get I'm excited for the the
44:47 release of that it's it's interesting to
44:49 see large companies like Adobe fall kind
44:52 of behind in it just because of how
44:53 they're trying to do it with Premiere
44:55 was kind of their first admit of defeat
44:58 so with premere how they're like we're
44:59 going to integrate Runway ML and Pika
45:01 and Sora to it it's like okay it's good
45:04 to see on that side because the new
45:06 Firefly I tested it out um hopefully
45:09 adobe's not listening but it wasn't good
45:11 yeah you know but they have the users so
45:14 I'll tell you I'll tell you a funny
45:15 thing about Firefly that you can try so
45:18 I I you have a go-to prompt you're
45:20 Spaceman you know my my go-to prompt is
45:24 a a 70s muscle car in an abandoned
45:26 Factory with light streaming Through the
45:28 Windows Firefly because it was trained
45:30 on public domain images it doesn't know
45:33 what a 70s muscle car is so when you do
45:35 it it's like a
45:37 1962 like stand they're horrible so my
45:42 test for Adobe is you know when it
45:44 recognizes 70s muscle cars then I'll
45:46 know it's made a leap but even version
45:48 three can't do 70s muscle cars it's
45:51 really funny no it's so it I think
45:53 they're all using the Nvidia one IO had
45:56 the same thing so Nvidia has one where
45:58 it uses like a layering technology and
46:00 you could tell they're just cutting out
46:01 one version like they're they're getting
46:03 all these like say Lions they cut them
46:06 all out and they're almost just like
46:08 putting like glue stick out and
46:10 squishing them all they're squishing
46:12 them all together on that side so
46:14 they're not they're trying to be like oh
46:15 we have all this training data and
46:17 they're just training it in a really
46:19 weird way and just using stuff that's
46:20 not competitive but even just seeing the
46:23 the open market significantly outperform
46:25 these corporations yeah that's uh that's
46:28 that's a good thing to see that's a good
46:30 thing to see well you know going back to
46:31 the rabbit R1 it's like when I when I
46:34 really took in what that device was I'm
46:36 like that should have been something
46:38 that Apple disrupted the industry with
46:41 right but it wasn't you know Apple's
46:44 kind of fat and happy and just sitting
46:47 on their Laurels and like where's Apple
46:49 by the way I'm that I'm waiting for I
46:51 guess worldwide developer conference
46:53 will see something but it's it's uh they
46:55 they are they are
46:58 visibly late to the party no they're
47:00 gonna try so Apple's always late on
47:02 technology even smartphones like we had
47:05 the because I remember back in the day I
47:07 wanted one of those windows phones or
47:09 the pal phones or everything else where
47:11 they're probably going to come in late
47:12 but I could see it they're trying to
47:13 train their own models and they released
47:15 a paper saying that oh it's this 30
47:17 billion model is better than gbd4 and I
47:19 was like yeah right maybe maybe from
47:21 your testing but they're they're
47:23 definitely behind in it even like
47:25 Samsung's not doing the best job they're
47:27 just I don't think it makes sense
47:29 they're Hardware companies you know
47:31 there they software they're not really
47:32 made for this big gen AI boom on this
47:35 side but yeah they're significantly late
47:37 at this point like you could some people
47:39 tried to say that oh they had AI first
47:41 it was Siri and I'm like ah I don't I
47:44 don't know about that fliping use Siri
47:46 it's hard yeah like I think we should
47:49 flush that brand down the toilet and
47:50 start over quite honestly but whatever
47:53 yeah yeah we'll we'll see with it and
47:55 same with like Alexa and Google and and
47:57 everything else but no I I honestly
47:59 think they'll probably partner with open
48:01 AI they the thing is they have enough
48:03 Capital where they could even like
48:04 acquire well not after Microsoft did
48:07 what they did Microsoft took the most
48:08 interesting out of all the tech
48:10 companies like Google's really just
48:11 sitting on it they're not sitting on
48:13 their hands they are releasing some good
48:14 stuff but considering they made the
48:16 Transformer model like they are like
48:19 they basically started it to some extent
48:21 and to see them so far behind it's like
48:23 and it it's an interesting thing I see
48:25 on large platforms where they're having
48:27 issues where um so I've talked to a lot
48:30 of just software companies um learning
48:33 companies Etc so this one was telling me
48:35 about it and it made total sense because
48:37 we made a custom tool to go into it to
48:39 be like a coach and co-pilot and you
48:41 know help people complete the course and
48:42 everything else and they're like yeah
48:44 Matt the the biggest problem with it is
48:47 we don't just have like coach like
48:49 coaches on here we have tutorials on how
48:51 to install a washer and dryer so it's
48:53 like any AI solution they create has to
48:56 go across their entire product Suite so
48:58 that's why you see with these big
48:59 companies they're basically hands tied
49:01 behind their back because they can't
49:03 just do something Niche really well they
49:05 have to make a broad AI that goes across
49:08 everybody and it usually just doesn't
49:10 work that well it's not trained right
49:12 for it yep well Google also had
49:16 the the you know the the the
49:19 disadvantage of effectively having a
49:21 monopoly they didn't have to innovate
49:23 and and unlike Microsoft Google doesn't
49:26 have a software business to fall back on
49:29 right they're they're all advertising
49:31 and and search advertising right that's
49:34 what is it 70% of their revenue or
49:36 higher 77 something like that so so okay
49:40 you invent a thing that if you take it
49:42 to its logical conclusion is going to
49:45 just answer someone's question rather
49:47 than give them a list of links well
49:49 their whole business is based on giving
49:51 you 40 links and getting you to click on
49:53 30 of them right they're not they're not
49:55 in the business to actually get you the
49:57 answer they're in the business of
49:59 getting you to engage with other
50:01 people's content and and that's one th%
50:06 at odds with what chat GPT does right
50:09 like it's it's like so so that's a
50:11 that's a rough thing you just
50:12 disappeared yeah my camera overheated
50:14 one sec oh that's really funny it's it's
50:17 only good for uh it's only good for like
50:19 50 minutes and then it that's hilarious
50:21 I just got to turn my fan around it'll
50:23 come back wait for it yeah I I'll just
50:26 have to be smaller
50:28 for I'm AI you caught me yeah right
50:32 that's
50:34 good um yeah but but what you
50:38 know here's the other thing about Google
50:40 that that the the thing that they
50:43 absolutely have is they've got the
50:44 smarts right they've got the engineers
50:46 that can do um can do remarkable
50:50 technology and can can produce um
50:54 remarkable technical
50:56 results unlike Apple they don't have the
50:59 humanity side of the house right Apple's
51:01 always been a balance of the art and the
51:04 science right Google went all in on it's
51:07 it's all engineers in fact they took it
51:09 to to the extent that they fired all
51:11 their managers and said you know it's
51:12 just it's all Engineers it's all about
51:14 engineering and what we're interacting
51:17 with are large language models right
51:19 we're we're interacting with Humanities
51:22 um and um and so I think they've got a
51:25 really distinct disadvantage to how do
51:27 you create tools that really require a
51:31 lot of that you know human component
51:33 where I think apple apple should be
51:35 really good at this once they once they
51:37 you know get it rolling they should be
51:39 pretty good at this I don't know that I
51:41 I think Google's going to struggle with
51:43 that for for some time no I I
51:46 agree um it's they they have some
51:49 Advantage where they have like a
51:51 business Suite that they could really
51:52 help with it but
51:54 they're it's it's just interesting to
51:56 see how they're kind of going about it
51:58 cuz I think they got a little bit cocky
52:00 too with Gemini where they stopped
52:01 worrying so much cuz they released the
52:03 whole that internal memo that released
52:05 we have no moat and then after that they
52:07 almost they were just they got cocky
52:09 they're like okay you know we we got
52:11 this figured out we got a million tokens
52:14 you know this is our mode it's like we
52:16 have better tpus and everything else and
52:19 I think that's where that's where it
52:21 kind of dropped off on on that side and
52:23 they're dealing with the same thing as
52:25 all these other big companies is um it's
52:28 like how do we roll out something that
52:30 works for everybody and that's always
52:32 the hardest thing and I feel like that's
52:34 why even co-pilot Microsoft did co-pilot
52:37 and like they're supposed to leverage
52:38 open AI for it but if you compare open
52:40 AI users to co-pilot users even though
52:43 co-pilot's free um it's it's not the
52:45 best on that end they are winning on a
52:47 couple ones like with their GitHub their
52:49 GitHub is actually getting really good
52:50 when it comes to that but there's also
52:52 other competitors too uh at the same
52:54 time but it'll be interesting to see how
52:57 all goes and then all these companies
52:58 have a lot of like I think if Google
53:00 gets really far behind they have enough
53:02 money where they may try to fix it with
53:03 acquisition which is the Microsoft play
53:05 like how Microsoft they didn't buy
53:07 inflection they're like we'll just hire
53:08 all their
53:09 employees and then by and that's because
53:12 everybody was I how do you how do you
53:15 how do you hire away the two co-founders
53:18 and most of the employees and not get
53:20 sued into Oblivion or you know at least
53:22 taken to court there there's got to be a
53:24 side deal there I I I have never seen
53:27 that kind of you know sort of pilfering
53:32 of the the intellectual Firepower of a
53:35 company before cuz inflection raised
53:37 $1.3 billion do it's not like you know
53:40 they don't have they're not small yeah
53:42 you know vested investors so I I don't
53:45 know how they pulled that off but that
53:47 that one's that was pretty impressive
53:49 yeah that'd be an awkward conversation
53:51 with the investors like H yeah about
53:53 that we're going to go with Microsoft
53:54 but don't worry we'll give you a couple
53:56 employees um I think a lot of the reason
53:58 was just uh the whole antitrust and
54:01 everything where companies are having
54:03 issues buying companies right now so all
54:05 these big mergers are like how the figa
54:07 and Adobe deal got basically pushed
54:10 under and I think that was their way of
54:12 getting around it where they're like hey
54:14 if we try to do it we're going to go
54:15 under Federal scrutiny let's just let's
54:18 just go around the let let's just hire
54:20 everybody from there and see what
54:22 happens yeah that's that was my instinct
54:24 is that it's some sort of you know 3D
54:26 chest acquisition that you know a year
54:29 from now the acquisition will be
54:30 complete right you know yeah that they
54:33 just about did it with open AI like
54:35 during that whole weekend that was the
54:36 craziest weekend ever Sachin was just
54:39 ready to take on 300 new employees on
54:42 that side that would have worked out
54:43 crazy for them that was nuts well that's
54:46 that's what the inflection guys that you
54:48 know they started that that division
54:50 that Alman and Brockman were going to
54:52 start so yeah that was that was a
54:56 fascinating
54:58 like that's that's one of the things I'm
55:01 excited to learn about is once all the
55:03 all the new technologies from open AI
55:06 release where they can actually talk
55:07 about what happened back in November
55:10 that's going to be fascinating because
55:13 there they talk about a movie right oh
55:16 no that's what I was thinking in Netflix
55:17 that's a five parter on Netflix for sure
55:20 movie for sure like what the hell went
55:23 down there and and just just the drama
55:26 of you know they're gone they're back
55:28 they're gone they're back now they're
55:29 working for Microsoft now they're not
55:31 working for Microsoft like wow that was
55:34 that was crazy and then we haven't heard
55:36 from Ilia since then oh overheated again
55:39 IL Ilia you know what the problem is
55:41 Ila's got your camera yeah I know I
55:44 guess right I got to just put my fan on
55:47 it it's a downside you get set up with a
55:50 DSLR and you think it's going to be
55:51 great but then you get all the problem
55:54 of it yeah see I'm I'm still on the
55:59 um just turning on my phone my my my
56:04 studio is a phone and a you know a $10
56:06 tripod from
56:07 Amazon you have decent lighting though
56:10 yeah I that's a my lighting is a little
56:13 LED strip from Ikea it's a $10 light a
56:16 $10
56:18 tripod and my
56:22 phone oh man you don't you honestly you
56:25 don't need much no I just overkilled it
56:28 I had the problem is with me is when I
56:30 get into stuff I get like super I have
56:32 to be careful what stuff I get into
56:34 because next thing I know you know it's
56:36 like even I'll go out for a bike ride
56:38 next thing I know I'm calling my wife
56:39 like yeah I'll be home at like 3:00 a.m.
56:42 you
56:42 know I'm having to Uber back because I
56:45 just got a little bit too little bit too
56:47 cocky with how far I go that's right you
56:50 wrote into Wisconsin over the niag falls
56:54 passing yeah 100%
56:58 that's
56:59 hilarious nice so yeah tell me more
57:01 about the whole AI Salon AI Labs I
57:05 haven't got too much into it on on that
57:07 side but I just want to know what you're
57:09 creating on your side it's everybody
57:11 that talks about it says amazing things
57:13 so congratulations for one on that thank
57:16 you thank you very much yeah the AI
57:17 Salon is really a remarkable group so so
57:19 that's so we have meetings every other
57:22 week and then we've got a mighty Network
57:24 so if you go to AIS salon. mn.com Mighty
57:27 networks.
57:29 co.com um that takes you to the AI salon
57:31 so you should definitely come join and
57:33 then you know within it um there's all
57:35 sorts of different areas think of the
57:37 mighty networks as kind of like a cross
57:39 between it's like Facebook and Discord
57:41 had a baby right you know so so there's
57:43 conversation areas we have a showand
57:45 tell area where people can show off the
57:47 work they're doing we've got guilds in
57:49 four different categories creative
57:51 business tech and and ethics um and and
57:56 there's just all these sub communities
57:57 within it we really encourage people to
57:59 use their real names you know one of the
58:02 things about Discord culture is it's
58:04 it's because it came from gaming it's a
58:06 lot of anonymous Tech Bros and things
58:08 like that one of the things I feel
58:11 really passionately about is that we are
58:14 in the middle of a transition from from
58:16 a future of work perspective from people
58:19 hiring um because of the skills you had
58:22 to people hiring for your aptitude and
58:24 because they know and trust you because
58:27 the AI is going to do a lot of the
58:29 skills part and so you're not just
58:31 hiring people for their skills anymore
58:33 so I think being in community and and
58:35 building trusted relationships um in a
58:38 community like the AI Salon is going to
58:40 be the future of work I think that's how
58:42 I think that's where hiring goes um and
58:45 so so so that's it there's also a thing
58:48 similar to the early days of the
58:49 worldwide web
58:50 where this stuff is moving so fast that
58:54 literally no one on the planet has their
58:57 head around this so everyone's trying to
58:59 figure it out so if you're in a
59:01 community of people that are all trying
59:03 to figure it out you'll get all these
59:04 different perspectives and you'll learn
59:06 a lot faster right so there's a there's
59:08 a sort of speed of learning thing that
59:10 happens and then the relationship
59:11 building thing really becomes
59:13 foundational for for where we're headed
59:15 so that's the salon and it's it's just
59:17 it's amazingly generous people they
59:20 share what they learn they're they
59:21 they're collaborating a lot they're
59:23 starting projects together uh it's
59:25 really exciting and then the AI learning
59:27 lab is really just my Tik Tok Channel it
59:29 just started out I was doing a lot of
59:30 the similar stuff you were doing where
59:32 focusing on products and things like
59:34 that and then on on April 30th a year
59:38 ago I did my first Tik Tok live and
59:41 here's what I did I um I used to come
59:44 home every night I have a SIM racing rig
59:46 and similar to you something I go into
59:49 it right yeah yeah so I would come home
59:51 and every night I'd have dinner you know
59:53 i' get home I'd have dinner and then I'd
59:54 Sim race for two or three hours and as I
59:56 got into this AI stuff and and just
59:58 really realized listen I got to put time
1:00:01 and focus into this I thought what if
1:00:03 for the next year instead of sim racing
1:00:05 two or three hours a night I went live
1:00:07 two or three hours a night and so that's
1:00:09 what the AI learning lab has really
1:00:10 become is just this nightly live jam
1:00:14 session and the biggest surprise I I
1:00:16 don't know if if you do a lot of of
1:00:18 lives but the biggest surprise for me in
1:00:21 that is I kind of had a preconception of
1:00:24 media as you know you produce this
1:00:25 content people consume your media and
1:00:28 what the AI learning lab has become is
1:00:31 like this interactive jam session where
1:00:33 you know I'll sort of come in I have no
1:00:35 agenda sometimes I'm demoing software
1:00:38 sometimes we're just you know
1:00:39 pontificating about what's going on
1:00:41 sometimes there's a lot of input
1:00:42 sometimes there's not um but because a
1:00:45 lot of the members that show up to the
1:00:47 live all the time are also in the salon
1:00:50 I can be talking about stuff or trying
1:00:52 stuff and they're like actually making
1:00:54 images and posting them to the salon and
1:00:56 they're like oh go check out the
1:00:57 Irregulars Channel by the way the the uh
1:01:00 people that show up to my live regularly
1:01:02 they nickname themselves the Irregulars
1:01:04 a bunch of them are here hello um but
1:01:08 the Irregulars are just this amazing
1:01:10 Bunch that they just show up night after
1:01:12 night after night
1:01:13 because they recognize that this AI
1:01:18 stuff in order to be literate with
1:01:21 what's going on you have to actually be
1:01:23 active in the conversation and so that's
1:01:25 what that Channel's become is just like
1:01:27 a just keep the conversation going keep
1:01:30 playing keep exploring and and that's
1:01:32 that's the AI learning lab so they they
1:01:34 really compliment one another nicely
1:01:37 nice oh that's crazy and great uh yeah
1:01:40 giving up Sim Racing for that you got to
1:01:42 blow the dust off of it at this point I
1:01:44 guess um anything I could do to help on
1:01:46 my side you know if you ever want me to
1:01:48 jump on a training or anything like that
1:01:51 I'm open for it couple things I would
1:01:53 love to have you join the salon and then
1:01:55 I think I haven't explored doing this
1:01:57 kind of thing like doing these Joes I
1:02:00 like this I think this was fun so maybe
1:02:02 we maybe just on a you know regular
1:02:04 basis every you know week or two we just
1:02:06 grab an hour and and just play I think
1:02:09 that could be good for both of us and
1:02:10 good for the audience no 100% I'd love
1:02:13 to do more of these lives I've been
1:02:15 that's one thing I find even about the
1:02:17 AI creators is it's actually a really
1:02:19 good community on our side just of
1:02:21 creators everybody's willing to
1:02:23 collaborate so not sure if you know who
1:02:25 Tech friend is but I've done live with
1:02:26 him um AI Savvy Victor I've tried to do
1:02:31 so we've had technical difficulties
1:02:33 about five times you know but I'm
1:02:34 feeling feeling optimistic that we're
1:02:36 going to get uh get it going eventually
1:02:39 yeah it's it should at some point um but
1:02:42 even just like the connections I've made
1:02:43 on this side everybody's been great and
1:02:45 you know it's like really on our side
1:02:47 it's just trying to to educate people
1:02:49 while learning ourselves that's the
1:02:50 whole thing it's like I don't claim to
1:02:52 be an expert in any in any of this stuff
1:02:55 I'm just I'm learning and teaching
1:02:57 everything I learn and I find even when
1:02:59 I teach I learn stuff better too at the
1:03:01 same time every night I one of my
1:03:03 mantras is there are no experts and if
1:03:05 and if anyone tells you they are an
1:03:07 expert run um because yeah no no one can
1:03:11 know what's going on with this stuff
1:03:13 right now um yeah it's it's I every
1:03:16 single night someone will share
1:03:17 something or I'll see something that's
1:03:19 actually the uh the the decco here
1:03:22 stations thing I didn't know about those
1:03:24 and and Danielle one of the Irregulars
1:03:26 came in and goes you know about these
1:03:27 stations I was like No And then I went
1:03:29 and looked at it and I was like holy
1:03:31 crap so like a lot of I'll discover
1:03:33 stuff in real time and that's really
1:03:36 exciting yeah I've had a bunch of them
1:03:38 they just come in my chat and they're
1:03:39 like tengra AI I'd never seen it before
1:03:41 I just did a video on it one of the one
1:03:43 of my chatter because I've been
1:03:44 exploring with lives a bit more lately I
1:03:46 hadn't I did lives for a bit after I
1:03:48 sold my agency because I was teaching
1:03:49 people how to do affiliate marketing and
1:03:51 I didn't want to teach like 87y old
1:03:54 teren who doesn't know how to use his
1:03:55 laptop how to do compete in the most
1:03:57 competitive thing you can in marketing
1:03:59 and I I feel bad I I'm not one of those
1:04:01 I can't push some Ponzi course and be
1:04:03 like you'll make a m I didn't want to be
1:04:05 that person so I did lives for I did non
1:04:08 I did about 200 lives over 200 days and
1:04:11 I was just like ah I don't really enjoy
1:04:13 this but I do like teaching on on my
1:04:15 side and I just would give everybody all
1:04:17 the keys for it I'm like okay here's
1:04:18 what you and no one did anything for it
1:04:20 too at the same time so that's where I
1:04:22 was like really the Tik Tok medium is
1:04:24 and I like the challenge for me it's
1:04:25 like can I review a tool in one minute
1:04:28 or even when I do something like a
1:04:29 tutorial on like how to use a stable
1:04:31 diffusion API in a minute and 30 seconds
1:04:35 you know you're just like yelling at the
1:04:37 mic being really quick and then you're
1:04:38 cutting it too at the same time like
1:04:41 okay sorry guys you may have to watch
1:04:42 this five times to get all the
1:04:44 information but you know it's a it's a
1:04:46 personal Challenge on that side yeah
1:04:48 that's funny um um kruno Chris Chris
1:04:52 Bruno uh just just uh posted you can
1:04:55 make money with chat gbt that's I've got
1:04:56 a little gold button I'm I'm at work now
1:04:58 but I have it at home that that one of
1:05:00 the things that used to drive me nuts is
1:05:02 all these Tik tokers like you can make
1:05:03 money with
1:05:05 chbt it's just like no no don't buy
1:05:09 those courses just just come in here and
1:05:11 play so I'm right there with you like
1:05:13 just learn just you know give away the
1:05:15 knowledge because it's moving so fast
1:05:17 that the the thing that's worth $1,000
1:05:19 do today because you've got the
1:05:21 proprietary knowledge it's going to be
1:05:22 useless 3 months from now so might as
1:05:25 well just give it away now so that's
1:05:26 what my head is well that's it and a lot
1:05:28 of it is just like here's how you can be
1:05:30 an automated spammer you know just
1:05:33 exactly spam you can now make, 1600
1:05:36 phone calls per minute unsolicited good
1:05:38 luck have fun it's so exhausting yeah
1:05:41 exactly no it's it's too much and or
1:05:43 even just the oh we won't tell you the
1:05:46 tool but you have to comment for the
1:05:47 tool name comment for and I'll DM you
1:05:49 the tool name or it'll be something and
1:05:52 they'll overblow the capabilities like
1:05:54 this new AI tool is going to replace
1:05:56 Hollywood and you know then you go to
1:05:58 get to the end of it it's the most basic
1:06:00 tool in the end it doesn't come near any
1:06:02 of that you're like come on that I call
1:06:04 it the AI you know the AI hype Lords
1:06:07 they they they have great videos for
1:06:09 that kind of retention you really even I
1:06:10 get motiv I'm like yeah let me check
1:06:12 that this sounds amazing then I find out
1:06:14 something like oh it's mid Journey okay
1:06:16 that makes sense so anyway uh listen I
1:06:19 gotta run I got another meeting I got to
1:06:21 go to but but this was awesome mat thank
1:06:23 you for inviting me and let's let's
1:06:24 definitely do this again this was
1:06:25 awesome aome yeah no I had a great time
1:06:28 let's do it again for sure it was great
1:06:30 to meet you talk in a public form if you
1:06:32 ever want to connect let me know hit me
1:06:34 up via DM we're we're connected now but
1:06:36 yeah maybe even next week we'll do
1:06:37 another one okay great yeah let's let's
1:06:39 figure something out maybe we'll do a
1:06:40 theme or whatever okay cool yeah nice
1:06:43 all right yeah cheers all right byee now
1:06:46 I gotta figure out how to hang this
1:06:47 thing up yeah I know me too that's
1:06:49 always an awkward part at the end and
1:06:50 then I do it too
1:06:51 [Laughter]
1:06:54 early all right cheers
1:06:56 see you and task let's
1:07:03 go all right there you go people peace
1:07:07 out good to see you I'll see you tonight
1:07:09 at 8 that was fun gotta go bye