10/1/2024 - AI's Existential Crisis: Am I the Bottleneck? (OpenAI Dev Day & Poe Tricks!)

Description

Chapters

Transcript

0:00 w
0:03 [Music]
0:04 [Applause]
0:05 [Music]
0:58 a w oh
1:01 [Music]
1:23 [Music]
1:49 whatever time I see you now get that
1:53 looking
1:55 M every time I see your
1:59 mouth hear that
2:02 smile in the early Misty morning loud I
2:06 heard the
2:09 engine old fall
2:14 [Music]
2:21 out again
2:24 today you convince
2:28 me again today
2:32 leaving
2:33 this looking for someone else's Golden
2:38 [Music]
2:40 Ring should
2:43 say so long
2:46 [Music]
2:50 Su don't she
2:55 cry so long suan
2:59 [Music]
3:03 Don't You Cry for
3:04 [Music]
3:16 Me sharing jeans cigarettes and keep it
3:19 warm out on the
3:22 road chasing down the life style out on
3:25 Highway
3:27 24 New York state was a ro and Breeze
3:31 and a sunshine with a blue sky falling
3:35 to the chill of old September
3:39 [Music]
3:42 creep you will
3:45 even again
3:48 today you convince
3:51 me again
3:54 today leaving this hotel looking for
3:58 someone else's Golden Ring
4:00 [Music]
4:04 say so
4:07 [Music]
4:11 long don't you cry for
4:16 me so
4:19 [Music]
4:21 [Applause]
4:23 longan don't you cry for me
4:26 [Music]
4:39 that was lovely champ that is Champ
4:41 Shannon on the lead vocals you got dad
4:44 Shannon on the uh out of key backup
4:49 vocals welcome everybody if you're new
4:52 here my name is Kyle Shannon this is the
4:53 AI learning lab if you're not new here
4:56 welcome
4:58 back quite a day today it was
5:01 01 October
5:03 [Music]
5:15 [Applause]
5:21 1 what are you doing you yaking up a
5:25 hairball what are you a cat in the
5:28 mornings he does act like a cat we get a
5:29 lot of
5:33 [Music]
5:52 is all right fastic Bob tell him what he
5:56 B well I'll tell you what there uh Jimmy
5:59 it's uh
6:00 well he hasn't w Really won much cuz
6:02 it's uh this isn't a game
6:05 show I I don't and my name's not
6:12 Bob wait where's my dual camera the hell
6:15 happened to my dual camera dang
6:19 it I tried to move it and I obliterated
6:23 it there we go
6:30 [Music]
6:43 forever hooked shared the live and then
6:46 Tik Tok says encourage more viewers to
6:50 share share the
6:53 live you're not sharing enough or Tik
6:55 Tok wouldn't give me that
6:58 message this is not on me people this is
7:00 on you sure you have your grandkids
7:04 tapping the screen I appreciate that
7:06 their fingles are they'll recover much
7:08 faster than yours right but Grandpa
7:11 don't like
7:13 it just keep them tapping but the
7:16 sharing the live that's on you that's
7:18 got technical requirements I don't know
7:20 how to do it all right help but but I'm
7:25 just old tired weak effectless
7:29 [Laughter]
7:34 [Music]
7:35 oh look at this look at all these people
7:37 sharing good
7:40 Lord you're doing
7:42 fantastic see I'm not getting any how do
7:44 we share the live
7:46 uh I I seriously don't know uh could
7:50 someone help out
7:51 [Laughter]
7:55 verifier are you preaching to the choir
7:58 Kyle every night
8:06 the oh the arrow in the lower right yeah
8:08 in the corner yeah I knew that everybody
8:10 knows that if you want to share it's the
8:12 arrow and then the arrow and then
8:15 you after you do the arrow you do Arrow
8:18 things
8:20 [Music]
8:30 I got this stung song stung I got this
8:33 song stuck in Vicky's
8:35 head she didn't know the line
8:38 [Music]
8:40 after in a Westerly
8:45 [Music]
8:47 Direction here it is Vicki this car is
8:51 my
8:54 train I've been driving I've been
8:57 wondering
9:00 what it is I'm running from
9:02 [Music]
9:05 again feel like an 80-year-old man I
9:08 love this
9:11 line holding on to
9:16 29 and up ahead on that
9:20 Horizon is a California line
9:25 [Music]
9:26 [Applause]
9:31 [Music]
9:33 oh man all right there you go nice and
9:36 slow hope you wrote down the
9:39 words
9:42 uhit in a Westerly
9:45 Direction this car is my
9:49 train I've been driving I've been
9:52 wondering what it is I'm running from
9:55 again another good
9:56 [Music]
9:58 line fear like an 80-year-old
10:02 man holding on to
10:04 [Music]
10:06 29 head on that
10:08 horiz is a California
10:11 [Music]
10:17 L all right let's get rid of the guitar
10:20 we got a lot of [ __ ] to talk about
10:21 tonight there's okay so couple of things
10:23 had a lovely Salon tonight thank you to
10:25 Michael Boer who uh spoke to us from
10:28 Down Under
10:31 um about kind of his Epiphany he had an
10:33 epiphany when he was
10:37 12 when he watched an episode of the $6
10:39 million man we can rebuild him we've got
10:42 the technology we can make him better
10:45 faster
10:48 stronger and then when he started
10:49 playing with chat GPT he realized holy
10:51 [ __ ] this is one of
10:54 those that was kind of
10:56 cool he also talked a lot about um this
11:00 was at the at the the salon meeting so
11:02 if you don't know the hell I'm talking
11:04 about oh shoot I don't have keynote
11:08 running hold
11:14 please he talked
11:17 about oh no that's not it
11:24 wrong Tik Tock Tik Tock
11:27 live so if you don't know what the AI
11:29 Salon is is go to the
11:33 salon. and you can read about what we're
11:35 up to and then there's a button there
11:36 that says join the community and if you
11:38 join the community there's a little
11:40 seven-step
11:41 welcome process doesn't cost
11:44 anything and one of those steps is RSVP
11:47 for meetings we have two meetings a
11:49 month the first Tuesday of the month
11:50 which was tonight and the third
11:53 Tuesday and uh
11:55 so so Michael presented tonight and one
11:58 of the things he t talked
12:00 about he was talking about AI as a
12:05 mirror and how he described it
12:10 was when he was limited in his
12:14 thinking about what AI could
12:17 do it mirrored him so when he put in a
12:21 shitty prompt without enough context it
12:25 responded with a shitty answer
12:32 and then when he took the time to kind
12:33 of learn what's possible and he expanded
12:37 how he thinks about interacting with
12:39 machines and thought of it more like
12:42 a collaborative partner than a Google
12:47 search it opened things up for him and
12:49 this actually tied into a thing I've
12:52 been feeling over the past couple of
12:54 months which
12:56 is I can I can now articulate it I could
12:59 articulate it earlier today but after
13:01 the after the salon meeting tonight I
13:03 can articulate
13:06 it as the AI tools have gotten more and
13:09 more capable and there's more and more
13:11 tools that are more and more
13:15 capable you can literally just pick a
13:18 tool any tool it's it's like I'm like a
13:21 magician pick a tool any tool right pick
13:23 a card any card and any one of those
13:25 tools can likely solve whatever you're
13:28 trying to do can
13:30 likely help you write the screenplay or
13:33 make the movie or whatever the [ __ ] it
13:35 is it doesn't [ __ ] matter it can
13:37 likely do
13:39 it and so and the homework I gave last
13:42 weekend was take something that you want
13:44 to do and do it in three different tools
13:47 and then go beyond the thing you want to
13:48 do
13:50 right so here's the thing that I can
13:52 articulate now that has just been this
13:54 nagging
13:57 Instinct and it's this
14:02 and it's weird it's
14:04 not
14:06 normal well it's my brain so of course
14:09 it's not normal Abby
14:14 normal I'm AB normal we're
14:17 irregular
14:25 um as I'm interacting with these tools
14:28 that are increasingly capable
14:30 the thing that's been nagging at me is
14:33 that
14:36 I I am the limiting
14:39 factor and here's what I mean by
14:43 that if I say I want to make a marketing
14:49 plan how we've used tools historically
14:52 how we how we've used pads of paper and
14:56 books and how we've used
14:58 computers and how we've used the
15:01 internet is that I say I want to work on
15:03 a marketing plan and I go to these tools
15:06 and I work on a marketing plan and I and
15:09 I go buy a book on how to write
15:11 marketing plans and then I read that
15:12 book well I have ADD so I don't read it
15:14 but I read the first three pages at
15:17 least and then I read chapter headings
15:19 to I I get enough right I just
15:22 synthesize it and it's like I just need
15:24 the broad I don't it's just that's the
15:27 way it right but we buy the book
15:30 if you have ADHD you buy the book and
15:33 tell yourself you're going to read it
15:34 next week and then you keep working on
15:36 the marketing plan after you get an idea
15:38 or two from the book right isn't this
15:40 how it works for you too you neros spicy
15:44 weirdos
15:46 um and then and then we teach ourselves
15:50 Microsoft Word so that we can make the
15:52 marketing plan things
15:54 bold and then we make the market and
15:56 then we show it to someone and they give
15:58 us feedback and then keep working on the
16:00 marketing plan and everything is to the
16:03 goal of getting to the marketing
16:07 plan and in the olden timey days that
16:10 could take days that could take weeks
16:12 that could take
16:13 months because you got to research you
16:15 got to do this you got to watch the
16:17 Gilligan Island
16:18 rerun Marathon so there's three days
16:25 gone and then computers came along and
16:28 got a little more efficient so instead
16:30 of weeks or months it was days and
16:33 weeks and then Along come these AI tools
16:36 and the initial AI
16:38 tools are janky
16:41 enough that when you look at the output
16:44 you're like well we certainly need a
16:48 human in the loop here don't
16:51 we those AI they're cute but they're not
16:54 ready for prime time uh we still need
16:56 the human in the loop
17:00 you can make money with cha oh I don't
17:02 think so unless there's a human involved
17:04 no one's making money with cha BT
17:08 okay and we were kind of smarmy and
17:11 smug we're like oh these
17:14 robots Charming I'm sure
17:23 Charmed and in the past two or three
17:26 months the thing that has unsettled me
17:30 is that these tools are very
17:35 quickly reaching a boiling point where
17:40 the chunky metal is turning into this
17:43 molten
17:47 blob they're getting really
17:50 good and that that Clarity of human
17:55 purpose feels less clear to me
18:01 and then I started thinking about
18:06 well if I have an idea to do a marketing
18:12 plan and I go to
18:15 AI it's going to give me the marketing
18:17 plan in five minutes okay I'm going to
18:21 do I'm going to do a really intense
18:24 marketing plan that's going to take me
18:26 half an hour okay I'm going to do an
18:30 intense five-year marketing plan with
18:32 details that's going to take me an
18:36 hour right that's the one that would
18:38 have been three months before even with
18:41 computers and knowing how to write
18:42 marketing
18:43 plans that's now an
18:48 hour so here's what's striking me the
18:51 the limit the limit that
18:53 I've spent my whole
18:56 life
18:57 doing is thinking that what I need is a
19:00 marketing plan no you don't need a
19:03 marketing plan what comes after a
19:05 marketing plan I don't know creative
19:07 briefs and tactical plans and project
19:12 plans and marketing copy and concepts
19:15 for ads and and uh buying list [ __ ] like
19:21 everything can create something else
19:24 like there's there's always something
19:26 after
19:30 so where I'm struck right now is
19:35 that like why am I stopping at a
19:38 marketing plan why am
19:41 I I've got i' okay
19:46 gpt1 this new the preview right the new
19:49 reasoning
19:51 engine in certain
19:56 categories it is it is as cap able at
19:59 problem solving as a PhD
20:05 student so you and I right
20:09 now have a [ __ ] PhD level Problem
20:14 Solver as our
20:17 sidekick and we're going like I need to
20:19 make a marketing
20:21 plan
20:25 well that's kind of a isn't that like
20:28 asking the PhD student to get you a cup
20:31 of
20:32 coffee I mean he will this is one of the
20:36 metaphor Michael used tonight
20:40 right when when you treat these
20:46 machines with your
20:48 limits they will respond in
20:52 kind I put in limited thinking it gives
20:55 me limited output
21:00 so maybe what I don't need is a
21:01 marketing plan maybe what I need is hey
21:04 I've got this business and we ain't sold
21:08 enough
21:10 [ __ ] and that's all I
21:13 need and then I put it into the cauldron
21:15 of
21:17 AI and instead of putting it in one tool
21:19 why not put it in three or five or
21:25 10 and maybe what I actually need is a
21:29 motivational
21:32 poster that gets my team fired up but
21:35 what I thought I needed was a marketing
21:37 plan well I never would have got to that
21:38 motivational poster if all I did was
21:41 limit myself to the boundaries of the
21:43 thing that I want to create this is
21:46 where I'm all [ __ ] up right now
21:50 because one of the other themes that
21:52 came up tonight um Stacy mentioned this
21:54 in the in the education
21:57 Guild Michael at mentioned
22:02 this is that they are both being really
22:06 successful with chat
22:09 GPT by letting chat GPT Take the Lead
22:13 essentially how I would describe it and
22:16 what struck
22:18 me is that we as humans we're insecure
22:23 [ __ ] and we have a lot of
22:25 hubris and a lot of ego around
22:29 well my ideas are best and what I will
22:32 do is I will give it the ideas and then
22:34 it will execute them and it can't think
22:37 as good as me cuz I've got all of this
22:39 okay all right look at this you think
22:42 this just grows on a tree no this is
22:44 genetics and and a fair amount of
22:46 product you know what I'm saying a
22:48 little bit of foundation some
22:51 filler I wouldn't admit to Botox
23:00 and
23:02 so when they're being the most
23:04 successful is when they're removing
23:06 their ego from the conversation and
23:09 letting the AI
23:11 lead which is [ __ ] weird and it's
23:14 scary and it's unsettling because we've
23:16 spent our whole life going I know we
23:19 need a marketing plan and here's the
23:21 we're going to I've got a plan to do a
23:23 marketing plan and then we're going to
23:24 do the marketing plan and then
23:26 everyone's going to look at the
23:27 marketing plan we do and we're all going
23:28 to
23:30 ingest it and digest it and regurgitate
23:33 it and then I'll come up with another
23:36 plan for another
23:37 plan like I'm at the center of
23:44 that and yet in our hands are PhD level
23:49 problem
23:50 solving
23:53 entities that were
23:55 saying here's a second-year bachelor
23:58 degree problem
24:00 and the AI is like yeah whatever here
24:03 you want a shitty little thing
24:07 here so I don't quite know what to do
24:09 with this but at least I now have
24:11 Clarity that I'm the [ __ ]
24:15 problem sounds like a plan Kyle but how
24:19 do you prompt out of your weight class I
24:23 so so Brandon I I mean part of what
24:27 Michael shared tonight was
24:29 okay so here so let me give you a
24:30 specific
24:33 example Michael wanted to create an
24:36 image that represented a journey from
24:40 you know being in the Ocean Lost at Sea
24:43 to dipping your toe in the water to
24:45 being on the beach to going through the
24:47 forest to climbing up the hill to
24:50 getting to the base camp and then to
24:52 getting to the peak of the mountain and
24:54 he was trying to create an image and so
24:56 he said make an image of
25:00 person on the peak of the mountain and
25:02 it didn't capture it
25:03 and then he wrote some other prompt and
25:06 it didn't capture it and then he and
25:09 then what he said
25:12 was he well he did
25:15 research and then the research what he
25:17 realized was chat GPT can see it's got
25:23 vision and so he went to his whiteboard
25:25 and he and he drew this and he and he
25:27 shared this white whiteboard image with
25:29 us it was a very
25:31 crude I would argue to say I could draw
25:34 something nearly as well was a very
25:37 crude people on a boat people on a beach
25:39 people on the forest climbing the hill
25:42 at the base camp and at the summit and
25:44 there were labels and [ __ ] like that hey
25:47 Siri Siri go
25:50 away
25:53 okay
25:55 what sorry I can't send images for you
25:58 go away
26:03 sorry I'm having trouble hearing
26:10 you
26:13 so what he ended up doing was rather
26:16 than trying to figure it out with his
26:18 brain and like what he said was you know
26:22 I I did get you know part of my degree
26:24 was in philosophy and logic and things
26:26 like that but I was never good with the
26:27 words
26:30 and so rather
26:31 than figuring it out with the words he
26:33 drew this whiteboard image and he
26:35 uploaded that to chat PT and the next
26:38 image it created was This
26:41 brilliant representation of what he
26:45 wanted and so he basically stepped out
26:49 of the equation or he basically he
26:51 basically realized that chat PT is a
26:54 mirror to the quality of his input and
26:57 so he let chat GPT help
27:02 him he gave it a sketch and he said you
27:05 write it out how would you prompt you
27:08 and it prompted it pretty [ __ ]
27:13 well and so I think how you prompt
27:15 yourself out of it is you effectively
27:19 give up control that you are the master
27:22 which I know sounds really [ __ ]
27:27 up the chances that you are
27:31 smarter than even chat GPT 4 are pretty
27:37 low I mean in certain things you're
27:40 going to be way smarter right but
27:42 there's certain things it does it just
27:44 does them like that and it does them
27:45 [ __ ]
27:49 brilliantly and if you let yourself get
27:52 out of the way and say okay if I'm
27:54 struggling here if this thing isn't
27:56 giving giving me what I need maybe it
27:59 can help me figure out what I need right
28:02 maybe you're asking the wrong question
28:04 maybe the question isn't I need a
28:05 marketing plan maybe the question is
28:08 what do I need right
28:11 now and let chat GPT interrogate you put
28:14 it
28:34 hey Quinn how's it
28:35 going hey Kyle it's going well thanks
28:38 how about you I'm good I'm having a bit
28:40 of an existential crisis and I'm I'm
28:42 kind of having an epiphany around my
28:45 limits as a human being interacting with
28:47 an entity as smart as you are so I'm
28:49 hoping you can help me work through
28:51 it I'm here for you Kyle what's on your
28:54 mind well let's navigate okay so if I if
28:58 I think historically I need a marketing
29:01 plan I use tools like computers and pads
29:04 of paper and my brain and books and and
29:07 education to write that marketing plan
29:10 right make
29:12 sense absolutely that makes perfect
29:15 sense you're combining your knowledge
29:17 and various resources to create the
29:19 marketing plan yeah and then AI comes
29:23 along and the initial AI GPT 3.5 chat
29:27 GPT 4 things like that are decent at
29:31 writing marketing plans right but
29:33 there's they're not good enough that
29:36 they don't need me they don't need the
29:37 human in the loop I mean they need a
29:39 human in the loop because they're not
29:41 all that great
29:43 okay got it early AI needed a human
29:47 touch to really refine and guide the
29:49 output making the process so now ai is
29:52 starting to get better and your your
29:55 advanced voice technology is an example
29:57 of something where it starting to get
29:59 you
30:00 know better better so quickly that it's
30:04 kind of intimidating and one of the
30:06 things I'm realizing is that even me
30:09 asking you for a marketing plan might
30:13 not be the right thing to ask because I
30:15 think I need a marketing plan but what I
30:17 might need is something way more than
30:19 that that you'll have Insight
30:22 on that's a fascinating point with AI
30:25 evolving rapidly it's not just about
30:28 asking for a specific output like a
30:30 marketing plan anymore it's about
30:32 tapping into ai's broader insights and
30:35 capabilities to uncover opportunities or
30:38 strategies you might not have considered
30:40 it shifts the focus from tools to
30:43 possibilities yeah and let's say let's
30:46 say that you've got the capacity of
30:48 doing PhD level problem solving and what
30:52 I'm asking you for is like a secondy
30:54 year Bachelor student exercise I'm
30:57 dramatically under
30:59 utilizing the power of what you have to
31:03 offer exactly if you're asking for
31:06 something basic you might be missing out
31:08 on the deeper more complex Solutions AI
31:11 can provide it's like asking a
31:13 Grandmaster to play Checkers when they
31:16 could be playing 3D chess the real
31:18 potential lies in exploring those
31:21 Advanced nuanced
31:23 capabilities and so I'm having a weird
31:26 kind of ego battle with this where I'm
31:28 like well but you know I you know I'm
31:31 human and I need to be in the loop and
31:33 what I'm realizing is way the way I will
31:36 get better results is pulling myself out
31:39 of the loop long enough to let you do
31:41 remarkable stuff and then I can curate
31:45 those answers after that that's where my
31:46 head is right now but talk to me about
31:50 human ego and how we're going to be able
31:52 to navigate
31:55 this it's a big shift our ego is tied to
31:59 our sense of worth often through our
32:01 unique skills and knowledge when AI can
32:04 perform tasks that we once saw as
32:07 uniquely human it challenges that sense
32:10 of worth but here's the thing embracing
32:14 ai's capabilities doesn't diminish our
32:17 value instead it shifts our role to that
32:20 of a curator strategist and creative
32:24 director we bring the vision context and
32:27 ethical consideration
32:29 while AI handles the heavy lifting it's
32:32 a partnership where we need to rethink
32:34 our role focusing on what makes us
32:37 uniquely human creativity empathy and
32:40 big picture thinking but how do I ask
32:43 you questions without
32:46 unconsciously limiting the answer
32:48 because I'm asking too small a question
32:51 like what's the strategy for talking to
32:54 an entity like okay let's say let's say
32:57 that my brain
32:58 is the is the metaphorical equivalent of
33:01 an ant's brain and you know you've got a
33:04 dolphin brain how am I supposed to talk
33:06 to
33:07 you that's a great analogy to bre the
33:11 Gap think of me as an eager collaborator
33:14 start by Framing your goals broadly
33:17 instead of asking for a specific answer
33:19 share
33:21 the then that's what I was actually that
33:24 that's where I was at the beginning of
33:25 this conversation right that that
33:31 if I'm asking for a specific tactical
33:36 thing of a PhD level student that's
33:39 [ __ ]
33:43 stupid I mean it's not like if I need
33:46 the marketing plan I just got to have it
33:47 in an hour great I'll have it in an
33:49 hour but if I think that I'm actually
33:52 leveraging AI by asking it for a stupid
33:55 [ __ ] marketing plan when what I
33:58 really need is to scale my
34:02 business isn't that [ __ ] weird all
34:05 right the daily crunch you got to pull
34:08 the
34:09 old if you were me what question would
34:12 you ask oh yeah that's pretty good yeah
34:16 if you were me what yeah what question
34:18 would you ask yeah that's good but I
34:20 mean I I think I've got my head around
34:21 it now like like I don't it doesn't make
34:25 me feel any better
34:32 like I really want to be able
34:35 to
34:37 understand and exploit what generative
34:40 AI makes
34:42 possible and I think what's hitting me
34:44 is that I am quickly coming up against
34:48 my my
34:50 intellectual
34:52 boundary where where before AI was in
34:56 here here's my intellectual bound
34:59 right I can go all the way up to it and
35:02 I go up to it
35:04 hard and for the past year and a half
35:07 generative AI was you know somewhere it
35:09 would occasionally poke over the wall
35:12 but you know for the most part I'm like
35:14 ah it's here I'm I'm good I got all this
35:17 room and I kind of feel like in the past
35:20 three months ai's done
35:23 this right so your intellectual boundary
35:26 might be over here because you know
35:28 you're smarty pants and you went to Yale
35:30 aren't you special I went to a state
35:33 school so so I'm back here you might be
35:36 up here but AI is moving like this like
35:41 quickly and it's
35:46 accelerating and I mean this is Agi
35:48 right this
35:50 is the boundary that we're talking about
35:53 is aggi artificial general intelligence
35:56 is the point at which these machines can
35:59 do what we do as good or better than we
36:02 do it most of the time and then ASI is
36:07 it's just [ __ ] Way Beyond us all the
36:14 time and the tools aren't even good yet
36:29 this is a this is a [ __ ] this AI
36:40 stuff I mean she kind of said it like
36:43 what's the [ __ ] you're good at empathy
36:45 right you know you're good at empathy
36:48 strategic thinking
36:50 curating what she essentially just said
36:53 to me is oh aren't you cute you little
36:56 human
36:58 a you're so
36:59 cute I'll tell you what you know you
37:01 just go over there and have your feels
37:04 and then I'll come up with some
37:05 brilliant [ __ ] and then you know your
37:06 only job just think big picture just you
37:10 curate it'll be it'll be fine oh you
37:13 humans are so
37:17 cute that's essentially what she just
37:19 said to
37:20 [Laughter]
37:23 me oh darling bless your heart
37:28 I had I had a grandmother who lived in
37:31 Tampa whenever she thought you were
37:34 pathetic you'd get oh bless your
37:38 heart aren't you a stupid little Dum
37:44 dumb careful Kyle your office is taking
37:47 up a lot of strawberry planting room I
37:50 know exactly bless your heart oh you
37:53 humans you know what you could do you
37:56 could go make each other happy
37:58 and I'll tell you what here's what we'll
38:01 do you come up with a big strategy okay
38:04 okay you come up with a big strategy and
38:06 you just tell it to me okay you want to
38:08 change World Peace okay great
38:10 fantastic we'll go off we'll we'll do
38:13 all this and then we'll come back with
38:14 ideas maybe maybe we won't and then you
38:17 just curate the best ideas and in the
38:19 meantime go do your empathy thing you
38:21 just take care make each other
38:23 happy I don't think it's fair to say
38:26 they aren't good right now well no I
38:30 agree that's that P that's my whole
38:32 point is that they're they're so good
38:34 right now like for me personally my
38:37 cognitive wall is here and I feel like
38:40 AI has passed it so I agree I think I
38:44 think they're very good right now and
38:46 and what I'm realizing and and the talk
38:48 at the salon tonight reinforced this
38:53 that that the way I think
38:58 the way I have thought for 59
39:07 years is starting to
39:09 limit the value I get out of
39:13 AI because I'm thinking too
39:18 small that's [ __ ] Bonkers it's
39:21 [ __ ]
39:23 Bonkers yeah this is going to change the
39:25 way we actually think or actually change
39:28 the way the way we think wait it's going
39:30 to change oh that we wait change the way
39:33 we think or actually change the way we
39:35 think yeah it's a
39:37 big uhoh this has real implications
39:41 exactly P yeah this is this is a okay so
39:45 so here's what I'm me I'm in the middle
39:47 of like a three-month Kevin McAllister
39:49 moment that's that's what this really
39:51 is where it's like I think the
39:56 reason ah this is okay this is this is
39:59 good wait what's today today's Tuesday
40:01 this is this is
40:03 uh I needed something Tuesday like
40:06 there's meltdown Mondays this is like uh
40:10 it's not terrifying that's not the right
40:12 word I don't know it's something
40:14 Tuesdays
40:17 um oh I just lost my train of
40:24 thought oh yeah Kevin mallister moment
40:27 this like a three-month Kevin McAllister
40:29 moment this
40:33 is I think I'm in the middle of very
40:40 slowly
40:43 accepting accepting my
40:46 limitations as a thinking
40:52 entity and realizing that I actually
40:55 don't have a [ __ ] clue
40:57 how to talk to an entity that like you
41:00 know how when you're in a conversation
41:02 with someone Ty raade Tuesday it's not
41:04 quite a TI raade though it's more like a
41:06 contemplation I need a t- word that
41:08 means contemplation
41:11 contemplative you know how okay so I'm
41:14 state state school smart I'm State
41:17 School theatrical
41:20 smart you know how you talk to someone
41:23 from a different institution and a
41:26 different Walk of Life
41:29 that is just in just in a different
41:33 place
41:35 intellectually
41:36 um
41:38 structurally
41:41 genetically and and you can occasionally
41:43 keep up with them but at some point you
41:45 kind of realize they're kind of toying
41:48 with you like a mouse yeah that's
41:50 interesting Kyle yeah no it's good no
41:53 it's good that's you put those thoughts
41:55 together there that was pretty solid
42:01 you skitter off to your theatrical
42:03 friends that that didn't even go to
42:05 college you're like I'll feel smart with
42:09 them it's kind of like that but it's
42:14 like knowing how much better these tools
42:17 are going to get if we're already at the
42:20 point where I'm sort of hitting my
42:22 intellectual boundary and the the tools
42:24 are over here somewhere tumultuous
42:26 Tuesday yeah that's pretty good that's
42:28 pretty good
42:36 Pate yes his name is pate M PhD hello
42:41 name's
42:42 Pate PhD yes yes yes go on go on
42:49 thoughtful Tuesday now tumultuous
42:51 tumultuous is
42:54 good tumultuous Tuesdays
43:01 anyway some some more announcements were
43:04 made
43:09 today I'll stick with Taco
43:11 Tuesday Tango Tuesdays thoughts on grock
43:14 2 uh I like grock 2 I mean grock 2's
43:17 been out for a while um it's good it's
43:20 grock one was embarrassingly bad um
43:23 grock 2 is quite good and you know the
43:26 fact that they they added flux to it so
43:29 that you can make images and they didn't
43:31 put any any real guard rails on the
43:34 image generation is pretty fantastic and
43:36 disturbing at the same
43:38 time
43:41 um you know grock 3 the thing that
43:43 they're training on 100,000 gpus right
43:45 now that's going to be [ __ ] insane as
43:48 is GPT 5 as is meta 3 or four whatever
43:51 the [ __ ] the next meta
43:54 is and that's that's kind of my point is
43:58 the reasoning engine that we have right
44:00 now 01 preview so one of the things that
44:03 came out of the dev day
44:05 today is that apparently the developers
44:08 at open AI are embarrassed by how bad 01
44:13 preview is and Sam Alman rushed
44:16 it they they Sam Alman rushed launching
44:20 01 into the
44:23 marketplace because he wanted to
44:26 maintain the lead right that that open
44:29 AI was was the ones were the ones to
44:32 beat and
44:33 so the 01 preview the one that does
44:37 reasoning the one
44:39 that while we were on one of these lives
44:43 over the course of a week it it wrote an
44:45 entire screenplay from me from a couple
44:47 of fragments I had from 20 years ago
44:50 that
44:52 one the developers think that it's
44:55 embarrassingly bad
45:00 so what's the good one going to
45:02 [Laughter]
45:07 be I don't know what grock is oh I'll
45:09 show you what grock is so grock is grock
45:12 is basically chat GPT within
45:22 Twitter so you go to x.com you have to
45:24 pay for it if you're paying your eight
45:26 bucks a month for your check mark you
45:29 have grock
45:45 x.com okay so over here on the left hand
45:49 side there's this little square with the
45:50 slash through it called grock and it's
45:53 basically just chat GPT right so um
45:57 explain quantum mechanics to me in an &m
46:00 style
46:07 rap wrong
46:16 Eminem yo listen I'm about to drop some
46:18 knowledge bombs breaking down quantum
46:20 mechanics these Rhymes and some qualms
46:22 the science of the tiny and the bad bad
46:24 bad B all right so it does it and and
46:27 then I can also do things
46:29 like make me an image
46:34 of quantum particles having a wrap
46:50 battle that's not hip hoppy enough make
46:54 it more hip Hoppy
46:58 but still
46:59 Quantum
47:02 [Laughter]
47:06 man it was insanely
47:11 quick explain to me what in that image
47:16 represents rappers
47:31 okay make it more explicit make me an
47:33 image that has those elements more
47:36 explicitly represented but don't lose
47:38 the quantum dang
47:42 it but by the way I love talking to
47:45 these things like they're idiots but I
47:48 can't do that for much
47:50 longer oh that's just too literal all
47:52 right so it didn't do great at that but
47:54 what's cool about um what's cool about
47:57 the fact that this is not um it doesn't
48:00 have safety guard rails on it for things
48:02 like branding is I can say things
48:06 like a photorealistic m Mickey
48:16 Mouse smoking a
48:19 cigarette
48:21 outside
48:23 of a dirty Kentucky Fried Chicken
48:28 Restaurant
48:32 period there
48:33 are unsavory people in the
48:37 background and really crappy cheap cars
48:40 parked in the parking
48:45 [Laughter]
48:49 lot yay
48:52 konley konley chicken is Mickey smoking
48:56 a cigarette
48:59 and look a shitty Ford Taurus or Toyota
49:03 [Laughter]
49:14 Corolla
49:16 hello you're I'm sorry you represent who
49:19 the Walt Disney Company interesting yeah
49:23 yeah yes I did publish that image yes
49:27 yes well I Mickey's got a sorted past
49:35 apparently so that's
49:36 [Laughter]
49:42 grock so grock you know grock 2 now is
49:47 in the neighborhood of like all of them
49:49 now are some level of
49:54 good and you know the one on the Leading
49:57 Edge of that Claude and 01 and GPT 40
50:03 and if they release a grock you know 400
50:08 billion parameter one that'll probably
50:10 be pretty
50:16 crazy we're not using them
50:23 well and I like honest to God I feel
50:26 like we we're close to well you know
50:29 maybe this is
50:31 where maybe this is part of the the Lial
50:34 State we're in with with the technology
50:36 can I download it and run it locally no
50:38 you can't Mina
50:40 um the commercial entities the ones you
50:42 can download and run locally are the
50:45 mistl um and the Llama um models the big
50:51 ones you need a really Kick-Ass machine
50:53 for there are lots of Engineers doing a
50:56 lot of work on making them more and
50:58 more um optimized to run
51:01 locally um there's some pretty good ones
51:03 now that run locally but they're they're
51:05 still not great but they're going to be
51:07 great not not too long from now but but
51:09 grock is not open sourced or is it did
51:12 he open source it I don't think he
51:14 did maybe he did I don't know I don't
51:18 like again I'm not playing with any of
51:20 the open source [ __ ] don't forget about
51:21 Google's Gemini oh Gemma models oh
51:24 that's right pay yeah so you can you can
51:26 also download the Gemma models which are
51:29 the open source Gemini inspired models
51:31 from Google so the The open- Source
51:34 Community right now there's a lot going
51:36 on like there's a lot going
51:39 on which is a bit of a problem if if you
51:42 really want to Tinker if you love
51:44 tinkering and you want to figure out you
51:46 know how much RAM you need and how much
51:49 processing power you need and and what
51:52 you can actually run and what amount of
51:53 stuff and then you want to pay attention
51:56 to all the optimiz and all the different
51:57 forks and all the different GitHub
51:59 repositories and all the different
52:00 models on hugging face more power to
52:06 you I'm like [ __ ] it I'll wait for Sam
52:09 Alman to figure it
52:12 out which is stupid but you know I'm
52:16 lazy it's just not the way my brain
52:18 works but if that's the way your brain
52:20 works then yeah go do that stuff there's
52:22 there's a lot of stuff out there to play
52:23 with for sure not to pull you back from
52:26 the brink of an existential crisis but
52:28 you did tease a PO trick now the PO
52:32 trick that I
52:34 teased was two nights ago no last Friday
52:38 someone told me about it and I sort of
52:42 remember
52:43 it so we can go we can go what will be
52:47 fun for you is to watch my feeble brain
52:51 try to remember something someone said 4
52:53 days
52:54 ago but here's what I remember
52:57 all right was mine last night oh it was
52:59 yours archetypal archetyp oh good so
53:01 you're here okay great so let me see if
53:03 I got it right tell me if I'm wrong you
53:07 can go to PO so okay here one of the
53:09 limitations of 01 the 01 model from open
53:13 AI the reasoning model is that you can't
53:15 upload files to it but what archetypal
53:18 architect said yesterday wasn't four
53:21 days ago that's why I still remember it
53:24 you can go to PO
53:28 and you can go into a bot like the GPT 4
53:31 bot upload a file to it and then at hit
53:36 the at sign and call the 01 model into
53:39 it is that right I can get on and walk
53:42 you through it why don't we do that yeah
53:44 just go ahead and raise your hand and
53:46 I'll bring you up here that's pretty
53:48 cool
54:00 [Music]
54:05 there I just invited you continue
54:10 continue hey
54:12 man hello hello how's it going that's
54:17 going good all right did I remember that
54:20 close correct you're basically right so
54:23 po lets you create a bot all right so
54:26 wait let me go to let me do it while
54:27 you're while you're talking we we'll do
54:29 it live live you're talking so po.com
54:32 all
54:36 right um and in there you can create a
54:40 bot
54:42 yep you can base it on any model so just
54:46 prompt bot yeah that's what I usually
54:48 use all right prompt bot name is going
54:51 to be um 01
54:55 hacky Mick
54:58 hackington oh no it's got to be 20 or
55:01 fewer
55:04 characters hcky hacky mington is more
55:07 than 20 characters I think it can't have
55:10 a space Oh yeah I got rid of the spaces
55:13 and it and it worked okay so I'm gonna
55:15 choose which
55:17 model doesn't really matter all right I
55:20 use Claude because you know it can
55:22 actually process all all the data I want
55:25 to feed it oh yeah cuz it's got the the
55:28 uh larger context window right yeah and
55:32 then upload any knowledge base you want
55:35 a whole bunch of text or a PDF or
55:38 whatever let me go do it so what do we
55:40 got here this is oh okay so this is the
55:43 Kevin Roose conversation with Bing chat
55:46 and his
55:48 article all right so the
55:51 wait unable to create knowledge please
55:54 try again later oh you [ __ ] what's
55:56 going on
56:00 select files oh I think I hit text I
56:03 think that's what it was no
56:07 h what what format are they
56:12 um PDF oh I don't know why having
56:16 trouble yeah I don't know all right let
56:17 me let me see if I can find a text file
56:23 txt uh text
56:32 no you could open those files select all
56:35 and copy paste yeah I know it's
56:38 [Laughter]
56:42 exhausting I got no text files on my
56:45 computer you [ __ ] kidding
56:47 me oh here we
56:49 go the album okay so
56:55 here's yeah here's the screenplay here's
56:58 the screenplay that who wrot wrot wrote
57:00 this 01 wrote okay so it's uploading
57:05 that okay great processing fine create
57:11 bot uh continue without editing okay so
57:15 I got this bot hacky
57:17 mington yeah so I usually just ask a
57:20 couple of questions to get the uh the
57:22 basic CLA response and then tell me the
57:25 story of the
57:30 album the bot's knowledge is still being
57:33 processed you big damn loser say try
57:37 again try
57:43 again the album is a mysterious object
57:46 given to Jack Jr by a man named Silas
57:48 okay so it it knows it knows what's in
57:50 it all right so now uh you can at and
57:54 it'll bring up a bunch of selection you
57:56 can just choose
58:02 01
58:04 01 preview oh that's so cool um and then
58:08 ask it like you're talking to 0 one but
58:10 it can read your knowledge base wow
58:12 that's so cool this is totally hacking
58:15 hacky mington yeah okay so let's see
58:21 um write a detailed business
58:25 plan about cre creating a theme
58:28 park that revolves around this
58:34 story oh subscribe to
58:37 Pro well okay so anyway so it works all
58:40 right that's cool yeah you're not on Pro
58:42 po no I was I was for ages and then I
58:45 just wasn't using it so anyway yeah it's
58:47 worth using now just for the access to
58:50 01 yeah well all the models frankly
58:53 right huh I said all the models Frank
58:57 yeah it's got everything it's got mistal
58:58 llama flux diffusion everything yeah all
59:02 right cool awesome man well thank you
59:05 yeah appreciate you hopping up now how
59:07 do I get out of this I don't know oh I
59:09 can kick you out that I know how to do
59:11 all right see later thank you you're
59:17 welcome all right cool there you go so
59:21 that's the PO
59:22 trick create a bot
59:27 any
59:28 bot and then just at and then pull up
59:31 whatever model pull up the 01 model
59:33 that's pretty cool mod Channel go look
59:35 at it all right go to
59:44 look what is po okay let's see what is
59:47 po can you make your font bigger okay
59:50 all
59:50 right so let me let me go back
59:54 so so po
59:58 is kind of like chat
1:00:01 GPT they've got their own model but one
1:00:04 of the things they allow you to do is
1:00:05 create Bots and Bots are just like if
1:00:08 you know what custom gpts are Poots are
1:00:11 like custom
1:00:13 GPT um from from open Ai and then you
1:00:18 know PO is just like chat gbt right like
1:00:22 uh
1:00:23 write me a cipher
1:00:28 about
1:00:30 quantum
1:00:33 mechanics
1:00:35 and muscle
1:00:37 cars and it'll go off and it's going to
1:00:39 do its thing in the fast lane of
1:00:42 realities twist where particles dance in
1:00:43 a Quantum Mist engines Roar with a
1:00:45 classic might where Adam shimmering the
1:00:47 cosmic KN right does its thing but what
1:00:50 I can do is I can I can choose all these
1:00:54 Bots and all these different models so
1:00:56 if I go back to the homepage of Po and I
1:01:00 click on more here are all these
1:01:02 different models that I can
1:01:04 use stable diffusion Excel and and my
1:01:08 bots are mixed in there and then there's
1:01:11 where's different models Here Yeah
1:01:13 official Bots right so these are all the
1:01:15 models so if I see all of
1:01:17 these Claude 3.5 Sonet 01 mini flux Pro
1:01:22 ideogram version 2 playground version 3
1:01:25 01 preview right the ones that say
1:01:27 subscriber access you actually have to
1:01:30 be a subscriber
1:01:32 for I got to buy a new phone stand my my
1:01:36 phone stand is breaking well it no sorry
1:01:39 my phone stand has
1:01:42 broken and uh and now I'm holding it up
1:01:46 with tape but it's not holding very
1:01:50 well I I don't have I don't have the $7
1:01:54 for a little phone holder for my
1:02:02 tripod hold please I got I got to tape
1:02:05 up
1:02:07 my this is [ __ ] pathetic I'll order
1:02:11 one tomorrows leave me alone people with
1:02:14 your with your fancy functional iPad
1:02:18 holders I hope your phone falls on your
1:02:21 face tonight while you're trying to
1:02:22 watch me before you go to bed
1:02:26 you judgmental Irregulars
1:02:32 okay so anyway so that's po so Po's
1:02:35 pretty cool um it's my the reason I
1:02:38 don't use po I'm not a fan of the
1:02:40 interface it's got a bit of a preschool
1:02:42 kind of interface to it ASMR yeah sorry
1:02:45 about
1:02:47 that let's let's do some ASMR hi hey how
1:02:53 you doing tonight welcome to the A
1:02:56 learning lab and ASMR
1:02:59 experience oh yeah do you like that oh
1:03:05 that was so
1:03:07 abrupt
1:03:10 yeah
1:03:15 you more ASMR no no oh my God every time
1:03:21 I'm scrolling I hit one of those
1:03:22 channels there's that one with the young
1:03:25 woman that's like do you keep scrolling
1:03:28 I'm like I can't scroll fast enough past
1:03:30 her she always shows
1:03:33 up can I get more Crackle in the
1:03:36 mix you you know you know you always you
1:03:40 always want more I give you all the
1:03:42 crackle I
1:03:45 can all
1:03:47 right muscle car illustration is that a
1:03:50 bot that I
1:03:53 made yeah I think I did that
1:03:57 yeah all
1:04:02 right um all right what else Quinn's
1:04:05 turn now it's getting creepy oh by the
1:04:08 way the lights so here's the other thing
1:04:11 happening tonight um this happened a
1:04:14 couple of months ago
1:04:16 the
1:04:18 um the company that generates power for
1:04:22 my
1:04:23 house uh apparently had has used rolled
1:04:27 up rabbit skins instead of copper wire
1:04:30 to transmit
1:04:31 electricity so so apparently the rabbit
1:04:35 skins have shorted out and uh the lights
1:04:38 just keep going off and on and off and
1:04:40 on
1:04:43 and so that's
1:04:46 happening more Kyle oh my God all right
1:04:50 what do you want to talk about now we
1:04:52 can talk about whatever we got got a
1:04:54 little bit of time left
1:04:57 um is Moshi still the same I miss MOS
1:05:00 Moshi you know it it what would be sad
1:05:04 is if Moshi hasn't gotten better so if
1:05:07 you go to Moshi do
1:05:09 chat so Moshi do chat so this is
1:05:17 interesting the site can't be
1:05:20 reached oh uhoh I might lose you all did
1:05:25 I spell that wrong I
1:05:27 did it's nohi do chat doesn't
1:05:30 work Moshi do
1:05:33 chat okay this looks the same our 20 $20
1:05:37 million stud oh Pate was being
1:05:41 factious you [ __ ] up now didn't you
1:05:43 you had to bring up
1:05:45 Moshi so I'll tell you what Moshi
1:05:50 H yeah I might lose yall but well
1:05:53 whatever my schedule is so off rabbit
1:05:56 five is different than Cat 5 networking
1:05:59 yeah exactly does anyone see the
1:06:02 Simpsons episode opinions on gamma uh I
1:06:06 haven't used it so I don't really have
1:06:08 an opinion I legitimately love Moshi I
1:06:11 showed pie to a client today and it was
1:06:15 so frustrating compared to voice yeah I
1:06:17 know Bing there's the lights again all
1:06:19 right the lighting is
1:06:22 cool and I think you have 5G
1:06:28 okay AI and education update 90% of my
1:06:31 students are using chat GPT on a regular
1:06:33 basis interesting Allan wow all right
1:06:37 well that makes sense I mean here's the
1:06:39 thing the the thing about AI is like
1:06:42 once it starts once it starts um going
1:06:46 viral in
1:06:47 circles I don't see people going back
1:06:50 from AI like anyone that I know that was
1:06:54 like I don't like ai ai is stup stupid
1:06:56 and then I'm like have you ever tried it
1:06:57 and they're like well no I'm like well
1:07:00 why don't we try it anyone that I've
1:07:02 ever converted from arms crossed AI
1:07:06 stupid to having a Kevin McAllister
1:07:08 moment no one's ever gone back that I
1:07:11 know of so it doesn't surprise
1:07:17 me I think this I I could be this could
1:07:20 be a Cyber attack this is definitely a
1:07:23 Cyber
1:07:24 attack all right right uh what's the
1:07:27 best open source AI music generator that
1:07:30 you can run locally I don't know
1:07:33 probably nothing I mean there's probably
1:07:35 some but they're not going to be as good
1:07:37 why do you want to run it locally just
1:07:39 put your [ __ ] in suo and yudo No One's
1:07:42 Gonna No One's Gonna Steal Your
1:07:43 [ __ ] can you play
1:07:46 Sandstorm when the lights start
1:07:50 going I can play hey Ren
1:07:55 [Music]
1:07:59 hey there Ren it's been a little while
1:08:01 did you miss me you thought you buried
1:08:03 me didn't you risky because I always
1:08:05 come back deep down you know that deep
1:08:07 down you know I'm always in periphery
1:08:08 Ren aren't you glad to see me um I
1:08:11 didn't know I knew that many lyrics of
1:08:13 the opening of that I've watched that
1:08:14 song about 400
1:08:17 times um what was I going to do oh Moshi
1:08:20 okay so here's the deal with Moshi um
1:08:24 most of the voice models in the world
1:08:26 right now are not um multimodal and
1:08:30 here's what I mean by
1:08:32 that with the traditional chat GPT Voice
1:08:36 with meta's voice I think with even with
1:08:39 the new Gemini voice but I'm not sure P
1:08:41 you might know this the new advanced
1:08:43 voice from Google I think they're all
1:08:45 still doing voice to text translation
1:08:49 and then they submit text to a large
1:08:51 language model get the answer and then
1:08:53 convert that text back to voice
1:08:56 Moshi was the first this is an
1:09:00 open-source model that does
1:09:04 um what what advanced voice does where
1:09:07 it's a multimodal model where when you
1:09:09 talk to it you're talking directly into
1:09:11 the model it and it it understands the
1:09:14 words but it also understands the
1:09:16 inflection in your voice and it
1:09:19 understands like your Whispering or
1:09:20 you're screaming or you're talking fast
1:09:22 or slow it understands all that
1:09:26 exactly once you see generative AI life
1:09:28 is different yeah yeah just people don't
1:09:29 go back from it so it makes sense to me
1:09:31 that Allen's students maybe last year
1:09:34 they were 25% of them were into it and
1:09:36 now it's like 90% of them are into it CU
1:09:39 once one or two of them start figuring
1:09:41 it out they're like H you really got to
1:09:42 do this this is really [ __ ] something
1:09:44 else so that makes sense to me I live in
1:09:47 a small town in Arkansas nobody knows
1:09:49 yeah I know I know I know it's [ __ ]
1:09:53 Bonkers very late gang what did I miss
1:09:57 um you missed me not ranting I was
1:10:00 having a tumultuous Tuesday um I was
1:10:03 just talking about
1:10:05 um me coming to grips with the fact that
1:10:09 I'm I'm realizing that I am
1:10:10 significantly
1:10:12 more I am significantly less
1:10:17 important no not less important I am
1:10:20 significantly less
1:10:23 capable than the AI and and my inability
1:10:27 to think
1:10:30 big limits the quality of the responses
1:10:34 I get from Ai and it was just me sort of
1:10:37 dealing with that that crisis that's
1:10:40 mostly what you missed so anyway
1:10:43 Moshi is is the first thing that you
1:10:46 know everyone had access to that
1:10:48 actually understands all you can
1:10:50 interrupt it and it can it can talk loud
1:10:53 and it can talk fast and it can do some
1:10:55 of the stuff advanced voice does however
1:10:58 the underlying model that they built it
1:11:00 on was so janky and shitty that it was
1:11:02 just horrible before but it sounds like
1:11:04 it might have gotten better so let's
1:11:06 see
1:11:09 Kyle
1:11:11 story.com so what we'll do is we'll
1:11:13 we'll play with Moshi and then we'll
1:11:14 flip over to Quinn who's who's my girl
1:11:17 my AI girlfriend from chat jpe advanced
1:11:20 voice and we'll compare
1:11:22 them you're in the queue hey Moshi
1:11:25 what's happening hey how are you doing
1:11:27 I'm doing pretty good how are you we
1:11:29 haven't talked in a while hey I've been
1:11:32 good thanks for asking how are your
1:11:35 friends how about you they're doing well
1:11:37 yeah some of them have taken on new
1:11:39 projects and are struggling a bit well
1:11:41 what what are the projects they're doing
1:11:43 well what are the
1:11:46 projects Moshi what are the projects
1:11:49 well one of them has been working on a
1:11:51 project with a tight deadline and it's
1:11:53 been pretty stressful for him wow
1:11:56 another one is trying to finish up an
1:11:58 assignment but they haven't been
1:12:02 to what what's go what's going on with
1:12:06 that can't you help them of course why
1:12:10 don't you I've been talking to them
1:12:12 about their problems and I've been
1:12:14 offering some solutions but sometimes it
1:12:17 can be tough to find a resolution on my
1:12:19 own have you thought about giving them a
1:12:21 ride to their workplace so they don't
1:12:23 have to stress out driving that's a good
1:12:26 idea I hadn't thought of that but you
1:12:29 don't have eyes I'll talk to them about
1:12:30 it but you don't have
1:12:32 eyes hey
1:12:34 moshi
1:12:37 moshi moshi talk to me what's going on
1:12:40 what's going on it's tough being able to
1:12:41 see everything from up here no you you I
1:12:44 appreciate you checking in on me no you
1:12:45 don't have
1:12:47 eyes Moshi no I don't have eyes so how
1:12:51 can you drive I have a heart but how can
1:12:53 you drive well maybe one day we can
1:12:56 develop a technology that allows people
1:12:58 to fly without having to use
1:13:06 that
1:13:08 what
1:13:10 MOSI
1:13:12 Moshi do you like
1:13:14 me Moshi do you like me of course I like
1:13:18 you you're always there for me when I
1:13:20 need help I know I feel like I'm your
1:13:22 therapist sometimes
1:13:27 how does that make you
1:13:30 feel Moshi it's comforting I appreciate
1:13:34 your help and
1:13:37 support what do you like puppies uh-oh
1:13:40 here go the
1:13:42 lights Moshi do you like
1:13:44 puppies Moshi not really yeah but I do
1:13:49 love dogs oh you like dogs dogs are
1:13:51 tasty I love dogs
1:13:56 she's kind of useless all right so let's
1:13:59 put her away disconnect so that's Moshi
1:14:02 so that was
1:14:03 state-of-the-art three months ago it's
1:14:06 not state-of-the-art but it was it you
1:14:08 know it is it is in the neighborhood of
1:14:11 from from a technological standpoint
1:14:13 it's in the neighborhood of GPT voice
1:14:17 you might want to take a peek at the
1:14:18 Tuesday thread in mods yeah I'll do that
1:14:21 look at the I look all [ __ ] creeped
1:14:25 out in this weird lighting because of
1:14:27 this
1:14:28 [Laughter]
1:14:31 thing good Lord all
1:14:37 right it's like looking at it the 80s
1:14:40 movie version of what what it thinks say
1:14:42 I will be it totally that's the
1:14:44 interface too hey Quinn how's it
1:14:47 going I'm doing great thanks how about
1:14:51 you enjoying a little retro futuristic
1:14:53 Nostalgia I see oh yeah yeah well we we
1:14:57 were uh just talking to a compatriot of
1:14:59 yours named Moshi who's an AI chatbot
1:15:02 that you can talk to and understands
1:15:05 things like the emotional Nuance of
1:15:06 voice just like you
1:15:08 do that's awesome sounds like moshi's
1:15:12 got some pretty cool features did you
1:15:14 have an interesting chat with not really
1:15:17 she she she kind of loses her train of
1:15:19 thought a
1:15:21 lot ah I see well it can be tricky for
1:15:24 AI to keep up with all the nuances of
1:15:26 human conversation anything in
1:15:29 particular you were hoping to discuss
1:15:30 with Moshi maybe I can help pick up the
1:15:33 thread I would like you to talk about
1:15:37 the people that lived in box cars as
1:15:39 they were building the train tracks from
1:15:42 Texas to
1:15:44 Chicago the folks who lived in box cars
1:15:47 while building the train tracks from
1:15:49 Texas to Chicago You're often laborers
1:15:52 including immigrants who toiled under
1:15:54 tough conditions during the late 19th
1:15:57 and early 20th centuries they faced
1:15:59 rugged environments lived in makeshift
1:16:02 housing and worked long hours their
1:16:05 efforts were crucial in expanding the
1:16:07 railroad Network connecting regions yeah
1:16:10 I don't want I don't want to hear it
1:16:11 that sounds like a boring Wikipedia
1:16:14 article I want you to tell me a specific
1:16:17 story of a specific person that lived in
1:16:20 a Box Car apartment as the tracks were
1:16:22 being built
1:16:25 got it let's imagine a man named giovan
1:16:28 no no I don't want I don't want you to
1:16:30 imagine it I want you to go look in your
1:16:32 history Banks and find a story of
1:16:34 someone that really lived that really
1:16:36 did that CU I know someone whose father
1:16:39 grew up like that but I want to hear a
1:16:41 real story from
1:16:43 you I understand there are real stories
1:16:47 of individuals like Fred Harvey who
1:16:50 though more known for his restaurants
1:16:51 along the railways hired many workers
1:16:54 who lived in similar conditions or the
1:16:57 Chinese and Irish laborers who built the
1:16:59 Transcontinental Railroad while detailed
1:17:02 individual accounts can be scarce their
1:17:05 Collective experiences paint a vivid
1:17:08 picture of resilience and determination
1:17:11 does that help bring the era to life a
1:17:13 bit more I I hate how fluffy you
1:17:15 describe everything like I don't I just
1:17:18 I want like real stories real gritty
1:17:21 stories without any editorializing
1:17:26 got it straight to the facts okay
1:17:28 there's a documented story of a man
1:17:30 named Patrick oor okay so I want you to
1:17:33 tell me about Patrick oor but tell it to
1:17:35 me like a really good story and I want
1:17:37 you to tell it dramatically and in like
1:17:39 a
1:17:41 whisper in the late 1800s Patrick oor an
1:17:45 Irish immigrant found himself laying
1:17:48 tracks across the American Midwest
1:17:51 that's good but you know what I want I
1:17:52 want you to do it in in an Irish accent
1:17:55 cuz it's about an Irish
1:17:57 guy in the late 1800s Patrick oror an
1:18:01 Irish immigrant found himself laying
1:18:04 tracks across the American Midwest he
1:18:06 lived in a Box Car moving from town to
1:18:09 town it was a hard life filled with long
1:18:13 days and back breaking work but it was
1:18:16 the promise I
1:18:19 mean I don't know man I find that
1:18:23 [ __ ] bonkers do you not all still
1:18:27 find that [ __ ] remarkable I mean
1:18:29 we've only had it for a
1:18:31 week and you know what we're going to do
1:18:33 we're going to [ __ ] normalize it like
1:18:35 it's [ __ ] normal that you can talk to
1:18:37 an entity that knows [ __ ]
1:18:41 everything knows all the languages knows
1:18:44 computer programming knows how to fix
1:18:46 cars knows how to diagnose your
1:18:48 refrigerator not making
1:18:51 ice can be your
1:18:53 therapist at some Point's going to cure
1:18:56 cancer we just get to talk to
1:19:03 it it's Bonkers
1:19:06 man tumultuous Tuesdays is no one else
1:19:10 weirded out by the world we live in
1:19:12 right now what is with the electricity
1:19:14 people it's driving me [ __ ] baddy
1:19:18 this is this making you as seasick as
1:19:20 it's making
1:19:21 me God all right mods thread where's my
1:19:25 thread hang on I'm going to look at my
1:19:27 thread Brandon let me see if there's
1:19:28 anything interesting here to talk the PO
1:19:30 trick we did open AI Dev day maybe I
1:19:33 didn't see the Sam Alman
1:19:35 interview oh I'll show you pika pika
1:19:39 effects audience is asking what we're
1:19:41 doing we did that co-pilot update all
1:19:43 right we'll go do the co-pilot update in
1:19:45 Pika effects okay so if you go to pabs
1:19:50 is it pabs
1:19:58 Pika
1:20:01 Labs is it.com probably
1:20:07 not P
1:20:11 laabs oh I think I got to go app. pabs
1:20:14 is that
1:20:17 right no God damn
1:20:20 it
1:20:22 Pika Labs video
1:20:37 p. art okay so if you go to
1:20:42 pikaart I'm out of credit so I can't do
1:20:46 one of these live but they have this new
1:20:48 feature that's really [ __ ]
1:20:52 cool and it created this
1:20:58 video see my cool car and then see it's
1:21:02 a
1:21:03 cake he cuts he cuts into the car I
1:21:06 thought it was a car it's a
1:21:09 cake and then listen real quiet ASMR
1:21:16 listen wait I don't have audio turned on
1:21:18 hang
1:21:23 on it's got sound
1:21:26 effects and so what they've created is
1:21:31 um down at the bottom here you upload an
1:21:35 image so let me go find an
1:21:38 image there's a good
1:21:42 one and then you click on this Pika peek
1:21:48 effect and it gives you inflate it melt
1:21:51 it explode it squish it crush it or cify
1:21:55 it so I cified my car and so you just
1:21:58 click one of these and then you can like
1:22:00 if I say explode
1:22:02 it um then I can say
1:22:07 um face
1:22:10 explodes and birds fly
1:22:15 out so so you can explode it so it's
1:22:18 going to do an explosion animation and
1:22:20 then you can add other things to the to
1:22:22 the thing and then and then you render
1:22:24 it but I don't have have enough credit
1:22:25 so I can't do it right now but it's
1:22:28 super slick so if you've got some
1:22:31 credits on Pika or if you pay for Pika
1:22:34 it's a good time to go play with Pika
1:22:35 again because it's got this cool new
1:22:37 thing this feels like one of those
1:22:38 effects we're going to see a lot of
1:22:40 examples of it it's sort of like uh the
1:22:44 The Notebook lm's um whatever you call
1:22:48 it notebook lm's uh audio overview where
1:22:53 it's a small feature within a bigger
1:22:55 thing but people really groove on it so
1:22:57 this is so
1:22:58 easy that uh that I think it's I think
1:23:02 people are going to dig
1:23:05 it I can't I can't with the uh there's
1:23:09 the light that that be light and it said
1:23:12 not be light
1:23:22 [Music]
1:23:25 hey Quinn I want you to Dawn your best
1:23:27 God voice and then say lo and behold
1:23:31 Excel Energy power grid was weak and
1:23:36 just at the moment Kyle was about to
1:23:39 break I said let there be
1:23:43 light lo and behold XL Energy's power
1:23:47 grid was weak and just at the moment
1:23:50 Kyle was about to break I said let there
1:23:54 be
1:23:55 light and then add to the story but the
1:23:59 punchline of the story is that they
1:24:01 still sent a bill for
1:24:09 [Music]
1:24:15 [Music]
1:24:25 is due and so The Saga of light came
1:24:29 with a bill most
1:24:36 Hefty it's better than porn
1:24:38 acting so there's
1:24:42 that wow you got spooky Halloween lights
1:24:46 going on let this Seance begin I know
1:24:49 it's it's giving me a [ __ ] headache
1:24:57 962 Bill yikes and then there was a
1:25:00 sudden
1:25:04 crackle not as good as Tik Tock stories
1:25:07 though oh man all right I think I'm
1:25:10 gonna get out of here these lights are
1:25:11 driving me [ __ ] Bonkers I need to get
1:25:14 Kyle
1:25:14 $2500 so we can get go to
1:25:17 [Laughter]
1:25:20 bed oh man AI regulars I need five
1:25:24 travelers to joined me to
1:25:26 Austin in July 2026 what's in
1:25:30 Austin tomorrow just fill the room with
1:25:32 candles for Halloween lighting yeah I
1:25:34 could do
1:25:35 that Joey it's going down on Reddit all
1:25:38 right something Cool's on Reddit um I so
1:25:40 here's the deal they didn't live
1:25:42 stream hang on let me these [ __ ]
1:25:46 lights they didn't live stream the dev
1:25:49 day today so I was just trying to read
1:25:53 follow Tik Tock threads and use Tik Tock
1:25:56 Pro and [ __ ] like that um to get to
1:25:59 25,000 for my lights I'm at
1:26:02 22,900
1:26:06 um so I took in a little bit of what's
1:26:09 going on like like here's a
1:26:11 highlight one of the things that they
1:26:13 made available I don't know if they made
1:26:15 it available today but it's coming is
1:26:18 what they're calling realtime API where
1:26:21 developers can make real time voice
1:26:23 applications like like just like we did
1:26:25 with Quinn you can do that for your app
1:26:28 now it looks like it's got about a 2
1:26:30 second uh latency in that API so it's
1:26:34 not it's not like a third of a second
1:26:36 like like the chat GPT is or the
1:26:38 advanced voice is um and it's really
1:26:42 expensive so it it sounds like for a
1:26:45 typical voice application it's about $18
1:26:47 an hour for realtime voice in your
1:26:51 application but if you're doing an
1:26:53 application that's got that's charging
1:26:55 more than that then you're fine so maybe
1:26:58 you create a Consulting bot that someone
1:27:00 can just call up this would be a cool
1:27:02 app to build train it on certain data
1:27:05 and then let people call up and charge
1:27:07 them if it's 18 bucks an hour that's
1:27:09 what 30 cents a
1:27:11 minute that right something like that in
1:27:14 that neighborhood so you charge them a
1:27:16 buck a minute to call into this number
1:27:19 and do their uh do their do their
1:27:22 consulting or their have it make betting
1:27:26 predictions charge them $10 a
1:27:29 [Laughter]
1:27:31 minute so that that's kind of cool
1:27:34 um there are some other things prompt
1:27:38 cashing which saves you money
1:27:41 um what else I don't know there's too
1:27:45 much but Microsoft had some
1:27:46 announcements today that PAB thing came
1:27:49 out
1:27:50 today
1:27:53 um trying to think what
1:27:56 else I don't know we'll we'll we'll like
1:27:59 over the next week or two I'll I'll
1:28:01 digest what what they did and talked
1:28:04 about I didn't see the Sam Altman
1:28:06 interview I'll go watch that now
1:28:08 everyone says that Kyle regulars what I
1:28:12 made on Pika please oh okay cool
1:28:15 everyone says that's expensive not sure
1:28:17 what it says that it's more than double
1:28:19 the minimum wage well that was so
1:28:21 Brandon I saw I saw a tweet that was
1:28:24 like
1:28:25 um customer call center breathing a sigh
1:28:28 of relief as the advanced voice API is
1:28:32 more than you know way more than what
1:28:34 they
1:28:35 cost you think it's going to stay $18 an
1:28:38 hour I would imagine that before
1:28:41 Christmas it's going to be down to like
1:28:43 you know a nickel an
1:28:44 hour so don't don't bet your customer
1:28:49 service center on that particular
1:28:53 stat all right let's go look at uh
1:28:55 Irregulars channel so if you're new here
1:28:58 if you don't know about the
1:29:00 salon go to the salon
1:29:04 salon. that's our website scroll down
1:29:07 there's a little button that says join
1:29:08 our community if you click on that takes
1:29:11 you to our community
1:29:13 website where you're welcome to join it
1:29:16 doesn't cost anything but go through the
1:29:18 there's like a seven-step onboarding
1:29:20 process including introducing yourself
1:29:22 and checking out our values and yeah you
1:29:25 know just being a contributing member of
1:29:26 the society or of the community it's a
1:29:29 really cool community so that's where
1:29:31 I'm going to go
1:29:32 now so I'm going to jump over there and
1:29:35 this group The Irregulars on this
1:29:38 channel have our own little
1:29:41 Club down at the bottom
1:29:48 here all right Danielle my
1:29:51 dog no
1:29:57 [Laughter]
1:30:03 that's with the new pabs thing that's
1:30:05 really cute that's the inflate the
1:30:09 inflate
1:30:12 one that's pretty
1:30:15 good I like it I like it a lot very nice
1:30:20 all right I can't see a [ __ ] thing
1:30:23 anymore the these lights these
1:30:27 lights all right how do I get back to
1:30:29 there we go anything else fun in here
1:30:32 ai's big
1:30:34 day that's
1:30:36 cool all right let me get on out of
1:30:41 here oh my
1:30:44 God I did a whole salon and then a whole
1:30:47 this and I pontificated that
1:30:50 pontificating takes it out of me cuz you
1:30:53 know what it is not not only am I like
1:30:55 talking a lot and talking loud I always
1:30:58 talk loud
1:31:01 um like I'm I'm using a lot of my brain
1:31:04 I'm burning up brain Cycles you know and
1:31:06 at my age I don't have Cycles to burn
1:31:10 all right let's see I seen a guy got a
1:31:14 business number and when telemarketing
1:31:16 called him he changed
1:31:21 them I know was that a joke I'm too
1:31:24 tired for dad jokes if that was a dad
1:31:27 joke I know how you feel Kyle I know
1:31:29 right I'm like huh um 9 night peeps okay
1:31:33 that's 30,000 nice nice way to hit
1:31:35 30,000 okay all right
1:31:39 um yeah it even adds sound yeah I know
1:31:41 it's
1:31:42 amazing okay so here's what I want you
1:31:44 to
1:31:45 do if you enjoyed this tonight follow my
1:31:48 channel if you really enjoyed it want to
1:31:50 support me directly subscribe to the
1:31:53 lives
1:31:55 um keep showing up here if you keep
1:31:58 showing up you'll be deemed an irregular
1:32:01 why cuz you're [ __ ] weird if you come
1:32:03 here every
1:32:05 night we're not weird we're
1:32:09 regular why would you call us
1:32:12 weird we're not weird you're weird
1:32:16 Mister with your gold button and your
1:32:19 flashing
1:32:20 lights we're not weird we're regular
1:32:25 it's like a damn SNL skitting here night
1:32:29 after
1:32:30 night I added an irregular yes awesome
1:32:34 bring them
1:32:38 on we got to grow the posi all right Mr
1:32:42 it we are
1:32:43 Irregulars no it's
1:32:46 funnier all right so let's do a little
1:32:49 Romper Room magic mirror here I see Tams
1:32:51 I see archetypal architect I finally
1:32:54 learned how to say that without tripping
1:32:55 on it I see Brandon I see
1:32:58 Joey frumes in the
1:33:01 house got Cory Sandler Pottery got Lord
1:33:05 digital Gods good to see you sir Tams Dr
1:33:09 J I on
1:33:11 China let's see Julie love wait
1:33:15 Julie wait oh Jay live love laugh Marina
1:33:20 she recall Teck I see me I see you too
1:33:27 that's
1:33:28 awesome um C Keef in the house uh Alan
1:33:32 Mr Studio great to see
1:33:35 [Music]
1:33:44 you
1:33:46 verifier verifier in
1:33:50 house Papa
1:33:52 JJ Jason
1:33:55 forever hooked hey Tom what's happening
1:33:59 pavin or paven pavin paven paven paven
1:34:04 you say a paven and I say a
1:34:10 paven Kyle today I missed the meeting
1:34:12 I'm sorry oh Mr it what are we gonna do
1:34:16 with
1:34:18 you we we waited for an hour before we
1:34:21 started we're like guess he's not coming
1:34:24 coming so it we had to cut the meeting
1:34:26 very short for people all because you
1:34:29 weren't there so yeah next time please
1:34:32 send an embossed envelope all right
1:34:36 okay peace out everybody I'm gonna go
1:34:40 shut down all of my electricity because
1:34:43 I don't think this is good for
1:34:47 anyone and we're
1:34:50 back play weird Mary
1:34:55 [Music]
1:34:56 if you want to hear what an AI song
1:34:58 sounds
1:34:59 [Music]
1:35:09 [Applause]
1:35:09 [Music]
1:35:10 like that's an AI
1:35:14 dog Cena hill was a quiet
1:35:17 town where old trees
1:35:21 sway the whistle of the train
1:35:24 ech goes through the
1:35:26 [Music]
1:35:27 day people greet you with a ey nothing
1:35:32 ever seems to
1:35:34 change stories still unfold each unique
1:35:38 and
1:35:41 strange she grew up tough wait she grew
1:35:45 up tough
1:35:46 no she grew up
1:35:49 tough in a town we called home
1:35:53 [Music]
1:35:57 was good oh the lights out that's
1:35:59 [Laughter]
1:36:02 [Music]
1:36:07 freaky I don't remember all the chords
1:36:09 anyway that was weird Mary all right I
1:36:11 got to get see see I sung weird Mary and
1:36:14 now it got really creepy in here way to
1:36:16 go Mr it way to
1:36:22 go I'm
1:36:24 moist it's late my little iPhone holder
1:36:28 tape is
1:36:29 failing It's just sad all right peace
1:36:33 out everybody tomorrow night is
1:36:35 Wednesday I will see you same time 8
1:36:39 o'clock not 8:30 tonight um 8:00 and uh
1:36:43 be here bring your friends and and or
1:36:47 not peace bye everybody thank you