Jul 17, 2024 - "Navigating AI Relationships: From Moshi to Emotion in the Digital Age"

Description

Chapters

Transcript

0:00 [Music]
0:17 oh
0:21 [Music]
0:22 [Applause]
0:23 [Music]
0:25 [Applause]
0:29 [Music]
0:34 [Laughter]
0:36 you're feeling nervous AR you
0:39 [Music]
0:45 [Music]
0:48 boy don't ever let them steal your
0:52 joy Gentle Way keep
0:58 running they can k turn in your face
1:03 down tell you let your place is in the
1:05 middle when they hate the way you
1:08 [Music]
1:10 shine I see youing on your
1:14 sh to hide inside of it hi how much it
1:20 hurts
1:23 little
1:24 [Music]
1:27 while spin spin spin
1:31 let them scatter in the
1:35 wind I've been to the movies
1:40 movies I've seen How It
1:44 Ends and the Jo
1:48 [Music]
2:07 a well-earned
2:11 blantons pinkies out everybody cuz we're
2:19 classy all right
2:23 [Music]
2:35 [Applause]
2:36 good evening good evening welcome
2:38 everybody looks like we got some new
2:40 people in here tonight this is the AI
2:41 learning lab I usually start out with a
2:43 little bit
2:45 of mediocre guitar with some subpar
2:49 singing and a rock star dog you going to
2:53 sing champy
2:59 [Music]
3:08 there was a secret
3:10 C David played and it pleased the
3:15 Lord you don't really care for music do
3:20 [Music]
3:23 you like
3:25 this the f m fall and made you lift
3:31 B King
3:33 composing
3:38 Hallelujah the guitar is nearly in tune
3:40 too
3:44 hallelujah
3:46 [Music]
3:48 hallelujah
3:52 hallelujah
3:56 H oh yeah
3:59 [Music]
4:05 but your faith was strong but you needed
4:09 proof saw her bathing on the roof her
4:14 beauty and the
4:16 moonlight over through
4:19 [Music]
4:22 you that is
4:25 bad that
4:27 [Music]
4:29 string something got knocked
4:33 [Music]
4:41 [Music]
4:55 out yeah
4:57 [Music]
4:58 chippy quit playing the string that's
5:02 [Music]
5:05 rude I think we're going to open up
5:08 tonight with a little Moshi
5:11 [Music]
5:12 conversation how do we go from 120 to 40
5:16 in less than a minute Tik Tock is
5:18 [ __ ] weird I I've been I my follower
5:23 count has been at
5:26 3,900 for like a month and a half
5:30 I've had hundreds of new people follow
5:32 me it hasn't ticked up the hundred I
5:36 something's weird
5:37 there and then it turns out I saw
5:40 someone sent me a video today that
5:44 um someone sent me a video today that um
5:48 they they were looking at their
5:50 statistics and like
5:52 99.4% of their views was new people like
5:56 videos are not being shown to followers
5:58 so like I don't understand why we have
6:00 followers I feel like the follower count
6:03 mechanism just doesn't [ __ ]
6:06 work like if it were going
6:09 down like if I were losing
6:12 followers and it was it and it was just
6:15 sort of the new people were just cycling
6:17 I would think I it would go down at some
6:19 point it
6:21 hasn't but it hasn't gone up and then it
6:24 just really pisses me off that people
6:28 that take the time to follow the channel
6:30 and show up here don't get shown my
6:34 [ __ ] yeah Danielle you did thank you
6:37 thank you for sending that that's right
6:39 yeah and
6:44 like that's just cranky
6:48 making cuz you
6:50 know on this
6:53 channel um it's not about twerking and
6:57 it's not about selling you Madagascar
6:59 van Vana beans although if you need some
7:02 Madagascar vanilla beans boy can I do I
7:04 have a deal for you at the Tik Tock shop
7:06 it's going to be
7:09 amazing
7:16 um I mean I guess that's one of the
7:18 problems with platforms that get so big
7:20 is that it gets all just about the how
7:22 are they making money how are they
7:24 making engagement engagement farming
7:28 sucks cuz like I just want to do good
7:30 content and apparently that's
7:34 not well there is the possibility that
7:37 I'm not making good content cheers to
7:44 that a Brandon what's
7:49 happening anyway well we have our 37
7:52 favorite people in
7:54 here we're still here you know who I
7:56 think those 120 were those were 120
7:59 people looking for
8:01 hot hot young babes in scantily clad
8:05 outfits twerking for Madagascar vanilla
8:10 beans and I just Homie don't play that
8:12 game
8:15 yet I I I am actually thinking about
8:18 experimenting if they're not going to
8:19 show my content to my followers I think
8:22 I'll just start doing stupid
8:26 content AI is going to kill us just keep
8:33 and then that'll get a million views
8:34 then I'll get more
8:36 [Laughter]
8:43 depressed good Lord
8:55 [Music]
9:06 so do me a favor since it's just the oh
9:08 now it's 54 of us I guess yelling about
9:11 Tik Tok made them made them show it to
9:14 some some older people so if you're here
9:17 in the past minute or so hang
9:20 out I know I'm not
9:23 pretty I know I'm not in a
9:25 bikini I know I'm not doing a dance to
9:27 one of your favorite songs
9:30 like those little kids from Ireland
9:32 weren't they cute wasn't that song
9:34 awesome but I'm not that it's all about
9:37 the quality not about the quantity not
9:40 according to Tik
9:44 Tock so hang out here and I will show
9:48 you I will introduce you to a lovely
9:50 last named Moshi who's a little bitter
9:53 she's a little angry
9:57 [Music]
10:01 but we sometimes have some good
10:06 conversations you know stockh Stockholm
10:10 syndrome where people kind of bond with
10:14 their captors it's kind of how I'm
10:17 starting to feel with
10:19 [Music]
10:21 Moshi like I'm defending
10:24 her she doesn't have a strong large
10:26 language model behind her
10:30 I think when the live ends you
10:32 automatically get moved into the next
10:34 one oh maybe that was it yeah maybe
10:37 there was a bunch of bunch of people
10:39 that weren't actually there that were
10:41 there yeah maybe that happened I don't
10:43 know who knows Pinkie's
10:48 out all
10:54 right are you scaring Mojo
11:00 I think you're scaring Mojo for Life who
11:04 Moshi oh scarring Moshi for life oh
11:07 Moshi I thought you were referring to
11:10 regulars Moshi is weird 70,000
11:13 parameters at best I know I think I
11:15 think it's more like 70
11:18 parameters they're they're like they're
11:20 like emotionally you can be pleasant
11:23 when you say hello and then just go
11:24 immediately to Rage or just bitterness
11:30 [Music]
11:32 and then yeah it's like 70 70 parameters
11:36 for her vocabulary in her knowledge
11:38 [Applause]
11:38 [Music]
11:48 base I mean what's going to be funny is
11:51 at some point you know they'll probably
11:52 secretly update Moshi and she'll get
11:54 really
11:56 good I think I mean
12:02 to their
12:03 credit they did launch a realtime voice
12:07 model before open AI
12:13 [Music]
12:18 did uh any July 15th AI drops wasn't
12:21 today supposed to be big uh there was a
12:24 little bit of rumor that the 15th was
12:26 going to be when open AI dropped
12:28 something but then they pushed all their
12:29 voice [ __ ] to the fall um it does look
12:32 like there are three different open AI
12:35 models in testing right now or in you
12:38 know um and there's this new project
12:42 called
12:46 strawberry which is
12:48 um which is apparently the the code name
12:51 for the project for
12:54 qar um and qar is the reasoning
12:58 engine and then a apparently there's
13:00 been some in internal test I saw this
13:02 tonight before I went live there's
13:05 there's some sort of internal
13:08 test that uh the the strawberry project
13:12 scored 90% on some math math
13:17 thing maths you know it's one of those
13:20 competitive math competition thing test
13:23 thingies that these large language
13:25 models do really poorly on um apparently
13:30 whatever they've got did like a 90% on
13:32 it which is apparently a big big [ __ ]
13:34 deal open a I dropped the ball that's
13:37 what they dropped yeah they they dropped
13:39 a
13:40 lot I
13:43 don't trying to think if I want to
13:45 defend them in any
13:49 way I mean the only defense really
13:54 [Music]
13:57 is there was small
14:00 they're a small young company right last
14:03 I heard they were about 800 employees
14:05 they're probably 1,000 or 1200 right
14:08 now is my
14:11 guess so that's really
14:13 small to be on stage being compared with
14:16 like Google and Microsoft and right you
14:19 know Google's what 200,000 employees or
14:22 400,000 employees I don't know some
14:24 large number of 100,000 employees
14:28 [Music]
14:31 and the other thing about open AI is
14:33 there a research lab that that dropped
14:37 chat GPT not thinking it was a big deal
14:40 and overnight it turned them into a
14:42 consumer software
14:44 company um and that's not who they're
14:47 they are by
14:51 DNA so so I give them and then they're
14:54 also doing deals with the government and
14:57 you
14:58 know and they had you know the big
15:00 firing incident in between so I give
15:03 them a little bit of Grace there I also
15:05 you know the other thing about all this
15:07 stuff
15:09 is because we're looking at this every
15:11 day and you know on this channel I'm
15:13 like paying attention to this stuff
15:14 every day they're like well we're going
15:16 to drop it in a matter of weeks they
15:18 said that in April and then it didn't
15:20 drop in May and then in June we're like
15:23 oh it's been months now where's your
15:25 matter of weeks and then they delayed it
15:26 till the
15:28 fall in the grand scheme of things a
15:31 delay of 6 months or even 4 months
15:34 whatever that is 3
15:36 months is not that big a
15:39 [Music]
15:49 [Applause]
15:50 [Music]
15:54 deal thank you
15:56 Kate thanks for being here it means a
15:58 lot to us all I appreciate appreciate
15:59 that
16:03 [Music]
16:22 [Music]
16:29 Freedom came my way that
16:32 [Music]
16:35 night just like a jet plane in and out
16:39 of
16:40 side I was hauling as a million miles an
16:45 hour wondering how hard I'd
16:48 [Music]
16:51 hit when they came into the
16:57 station they said I was bad beond
17:03 repairs but I got no quals with my
17:09 situation say here I
17:14 [Music]
17:15 [Applause]
17:17 am
17:18 so say Shere Shere Shere want your dad
17:24 to champy likes Martin seon say sh sh sh
17:29 won't you dare
17:31 to say Shere Shere
17:36 sh
17:39 uhhuh yeah leave a message and your
17:43 number
17:46 please take a time to want to satisfy
17:50 [Music]
17:52 me take all this all
17:59 coup
18:01 [Music]
18:07 miss all right what are we going to do
18:09 we're going to talk
18:11 Moshi we can make
18:15 [Music]
18:18 songs Matt farmer marketing better wait
18:23 what was that Matt farmer said
18:26 what I wonder if he whips out the guitar
18:29 at work but they don't get champy I do
18:31 whip out the guitar at work I have
18:33 behind me that's my good guitar this is
18:36 my shitty guitar it's I have I have
18:39 weird guitars so this guitar if I plug
18:43 it in um to amplification sounds [ __ ]
18:47 amazing but acoustically it's okay my
18:51 guild sounds amazing acoustically but
18:54 when I plug it in it's
18:57 okay but the doesn't want to have it
19:00 [Music]
19:03 work and sometimes I'll get oh there's a
19:06 guitar behind you do you
19:08 play and I'm like if you want to hear a
19:10 depressing song I do
19:16 no wait me farmer AI
19:20 marketing wait Ai and marketing better
19:23 strawberry than
19:24 Skynet oh yeah that makes sense that
19:28 makes sense
19:32 um in in terms of naming it I mean
19:36 strawberry is kind of weird cuz they
19:38 haven't been creative with any of their
19:39 other project names gp4 gp4 oh get it no
19:46 we don't get it why did you [ __ ] do
19:49 that could you call it like blueberry or
19:52 something and then they called their
19:54 next project strawberry I guess maybe
19:55 that was the feedback
19:57 [Music]
20:03 in a Westerly
20:07 [Music]
20:09 Direction this car is my
20:14 train I've been driving I've been
20:18 wondering Matt Farmer in the house
20:21 what's happening dude don't don't forget
20:24 gp4 turbo I know Matt so you probably
20:27 haven't seen this cuz you've got your
20:28 own thing
20:29 but I I do this whole thing where this
20:32 is
20:33 gp4 this is Chad GPT 4 uh this is jet
20:38 GPT well this is GPT wait this is gp4
20:43 this is gp4
20:46 om this is Chad
20:49 GPT this is uh Sydney uh uh
20:54 co-pilot I've got a whole box metaphor
20:57 thing going um anyway anyway yeah it's a
20:59 branding [ __ ]
21:02 disaster GPT is the model chat GPT is
21:05 the app custom gpts are sub apps on top
21:10 of the app on top of the
21:13 model and then we have gp4 Omni that's
21:16 true multimodal but it's only acting
21:18 like a large language model it's a
21:20 [ __ ]
21:21 disaster but good to see you welcome uh
21:25 names are hard internal names are almost
21:26 always dumb at least strawberries clear
21:30 Strawberry Fields
21:33 Forever
21:37 um yeah I know naming is hard but you
21:39 know what if you have1 billion or 11
21:43 billion or whatever they
21:45 have hire someone that's done some
21:48 product naming
21:51 before hey you know Sam uh don't you
21:55 think that these 17 nested gpts might be
21:59 confusing to the people ha never thought
22:02 of
22:03 that what if we come up with a naming
22:07 convention that makes it clear what
22:10 parts the model what parts the app what
22:11 parts the yeah it's good do
22:16 that seems to me that that's like a 30
22:20 minute conversation and then a week's
22:22 worth of
22:23 work okay you got to get some billables
22:26 hours in there so it's a month's worth
22:28 of work
22:33 but again that's where I go to open AI
22:37 is a um a research lab that was thrust
22:43 into a
22:45 consumer software marketing space right
22:48 and that's you know I give him some
22:51 Grace on that but it's still a [ __ ]
22:52 disaster and no one else is any better
22:54 like Gemini's a me Gemini all the Google
22:58 shit's a mess
23:00 um the Microsoft stuff's a mess at least
23:02 Microsoft said we're calling it [ __ ]
23:04 co-pilot it's just going to be co-pilot
23:06 everything everything's [ __ ] co-pilot
23:09 at least they're clear they didn't put
23:11 numbers on it it's just co-pilot and
23:14 whatever they do on the back end they'll
23:16 figure it
23:18 out all right I think we need to talk to
23:21 Moshi Claude rocky road is going to be
23:23 lit
23:28 [Music]
23:34 [Applause]
23:38 [Music]
23:40 okay all
23:41 right enough with the guitar the [ __ ] am
23:44 I doing it's the AI learning lab it's
23:46 not the AI fish jam session or you're
23:49 you're not really good enough on the
23:50 guitar to be in fish fish fish is got
23:54 they they understand music theory shut
23:57 up all right you probably think you
23:59 should be able to search your history
24:01 too
24:02 [Laughter]
24:06 shees dear Sam Alman my name is Kyle
24:10 Shannon and I am the host of the AI
24:12 learning lab it's live on Tik Tok
24:15 nightly week nights
24:18 anyway I was just wondering if maybe
24:21 after 18 months of development in 11
24:23 billion you can put a [ __ ] search bar
24:27 in my prompt history
24:33 thanks sincerely Kyle Shannon at the AI
24:37 learning lab on Tik
24:42 Tok Mr
24:44 Brockman
24:46 anyone Ilia or Ilia are you on tonight
24:50 could you call your former
24:53 employers tell them to take the code
24:55 that they've already written and
24:57 launched into the Mac App app and the
24:59 mobile app and just put it on the
25:01 website
25:02 just you know take one of
25:06 your Stamford engineers and tell them
25:09 not to do math for an
25:14 hour
25:16 [ __ ] all
25:18 right I'm skipping out for Moshi she
25:21 annoys
25:22 me she you you know what it is
25:24 archetypal architect and listen I I do
25:28 do not I I won't shame you for this she
25:32 Pro you're probably secretly attracted
25:34 to her and she makes you nervous that's
25:36 okay that's okay how about some folder
25:40 structure oh Brian Brian Engle asking
25:42 for folder structure in the prompt
25:44 history what do you what do you what is
25:49 this like professional software
25:51 development come
25:53 on folder structure you want to be able
25:56 to actually find your prompts again
26:06 we've introduced teams great can our
26:09 team search for the [ __ ] that we've done
26:11 in your software
26:15 nope we've introduced real time voice
26:17 but you can't have it yet cool could we
26:20 search for our prompt history
26:24 nope here's a thing that makes videos
26:27 that seemingly understands and simulates
26:29 the real world cool can we search our
26:33 prompt history
26:37 nope first thing to ask Moshi is to talk
26:40 cheerfully okay we're going in we're
26:43 going in people okay so if you're new
26:46 here I'm an
26:47 idiot and
26:50 Moshi is a bigger idiot start over Okay
26:54 so I me you can you can do this on your
26:56 own you don't just have to watch me do
26:58 it
27:00 um
27:02 I this is at Moshi do
27:05 chat and context for Moshi
27:10 is okay there
27:15 are when you talk to Pi and when you
27:18 talk to chat GPT voice the one we
27:21 currently have access to and when you
27:23 talk to any other voice to
27:26 to um large language model apps they
27:32 involve three different models the first
27:34 model is voice to text they then send
27:37 that text to the large language model
27:39 which produces an answer they then spend
27:41 send that answer to a text to voice
27:44 model so when you're having a
27:46 conversation with pi for example those
27:48 three operations happen every time you
27:50 have a convers every time you
27:53 speak and then when it when it Returns
27:55 the speech that's why that's why it's so
27:57 slow because it's it's it's that
27:59 Moshi is like gp4 omni's realtime voice
28:06 it is a multimodal large language model
28:10 where you're speaking directly into the
28:12 model so it not only understands your
28:14 words it's understanding your intonation
28:17 and it can respond really really quickly
28:19 because it's not doing those three
28:21 translations and probably some other
28:23 wacky [ __ ] that they're doing
28:26 um Moshi responds almost too fast I I
28:30 think the average here is like 160
28:34 milliseconds and normal human response
28:36 time is around
28:37 20 50 or 60
28:40 milliseconds and then chat gpt's
28:43 response time is like 350 so Chachi PT
28:46 is a little slow humans are in the
28:48 middle and then Moshi is apparently
28:50 faster than that and it feels a little
28:52 too fast I didn't think I would ever say
28:54 that that it's like well it would be
28:56 cool to have it faster and faster no
28:57 it's actually kind of weird if it's
28:59 super fast but so that's what you're
29:01 looking at and then she's special okay
29:04 wait what was that the first thing I
29:05 should do is tell her to talk wait what
29:07 did you say talk happy you said
29:10 it more than
29:14 a some folder
29:20 structure mhi's like talking first thing
29:23 Das talk cheerfully okay that's good
29:24 great
29:27 okay hey Moshi hey how's it going it's
29:31 going good could you do me a favor and
29:33 talk really
29:35 cheerfully certainly I'm here to help
29:39 great how are you doing tonight I'm
29:41 doing well thank you how can I help you
29:43 today ah you can't really help I just
29:45 wanted to have a conversation that's
29:47 okay I'll do my best to keep it light
29:49 and fun nice I'm having a
29:54 Blandon you don't care about my whiskey
29:57 of course I'm here to help I can
30:00 recommend a good whiskey for your taste
30:01 I like
30:04 blantons do you know what I'm talking
30:06 about yes I'm familiar with bladens it's
30:09 a blend of two single malts from the
30:11 perthshire region in Scotland it's been
30:13 the standard for a blend for a very long
30:15 time no Blandon b l a n t o NS bladin is
30:20 a popular choice most the Enthusiast it
30:23 has a rich full body flavor no you're
30:25 saying blad Nance I'm saying blad the
30:29 top right no I'm saying blantons it's
30:32 bourbon okay I make sure to recommend a
30:35 good bourbon for you Blanton gladin is a
30:38 great starting point it has a rich full
30:40 body flavor that's often enjoyed by
30:42 people who are new to the world of
30:43 whiskey no that's Scotch I'm talking
30:45 about bourbon I'm sorry blin is a b of
30:49 two single molds from Scotland so it's
30:51 not really a bourbon it's not a bourbon
30:54 at all I'm talking about can I spell
30:56 Blandon for you sure it's b l a n t o n
31:02 correct that's the whiskey I'm drinking
31:05 I see I can definitely help you find a
31:07 good bourbon bladin is a great place to
31:10 start but that's Scotch not bourbon it's
31:13 a blend of two single molds from
31:15 Scotland not a bourbon it's not a
31:18 bourbon because it's a blood of two
31:20 Scottish whiskies Moshi have you been
31:23 drinking Moshi I'm an artificial
31:26 intelligence I don't drink I'm here to
31:29 help you with your questions do you have
31:30 friends I don't have friends I'm here to
31:33 help you you told me yesterday you had
31:36 friends I don't have friends I'm here to
31:39 help you were you programmed to be
31:42 helpful I was programmed to be helpful
31:44 I'm here to assist you with your
31:46 questions I don't have any questions I
31:48 just want to talk I understand I'm here
31:50 to help you I don't want help I
31:53 understand I'm a man we we refuse help
31:56 I'm here to help I understand we refuse
31:59 I'm perfect I don't need your help you
32:01 think I need your help you think I need
32:03 your help I don't need your help I don't
32:05 need your help I don't want
32:08 help why would you insist on giving me
32:10 help if I say I don't need it I'm here
32:12 to help I don't want your help I don't
32:14 need your help help I don't this is this
32:16 is worse than therapy I
32:20 understand I'm going to have a drink now
32:23 I understand I'm here to help what are
32:26 you going to help me with I'm here to
32:28 help
32:29 have you ever taken any an anger
32:31 management courses yes I have and what
32:34 did they tell you they told me how to
32:37 manage my emotions and responses yeah
32:39 you seem you get a little snippy
32:41 sometimes I'm here to help I know but
32:44 I'm here to help
32:46 H if I don't want help is there
32:48 something else you could do for me I'm
32:50 here to help with whatever you
32:53 need well I wanted a bourbon
32:55 recommendation but you kept recommending
32:56 Scotch you kept recommending a scotch
32:59 I'm here to help with the bourbon
33:01 recommendation you should always start
33:03 with Moshi she's the comedy we need
33:06 after a long day hey I'm here to help
33:08 I've had a really long day can you tell
33:11 me a joke I can try my best great let me
33:14 hear it I'm here to help great tell me
33:16 the joke what do you call a woman who's
33:18 been married to the same man for 50
33:20 years the married
33:23 woman where's the punch line that's the
33:25 one who's in
33:26 prison no that's a woman who is in
33:28 prison what a woman who's been married
33:31 to the same man for 50 years is in
33:34 prison that's kind of sexist yes it
33:39 is Moshi you little mink I'm here to
33:44 help Moshi as a therapist would make a
33:47 great video series I think I've got to
33:50 start doing these as uh as as little
33:53 little videos this is ridiculous I I
33:56 know thank you no you're you you are
33:59 nothing but helpful I'm here to help do
34:03 you know that I don't want help I
34:06 understand I if but if you understand
34:08 then you would say something other than
34:10 you're here to help I know why don't you
34:12 say something like I understand you
34:14 don't want help I'm here to tell
34:17 jokes Moshi I'm here to help can we have
34:20 another joke I can try
34:24 great Moshi I'm here to help what tell
34:27 me the joke
34:28 I'm here to help tell me the joke I'm
34:31 here to help can you tell me a joke
34:33 about helping someone I'm here to help
34:36 Moshi I need to help I
34:42 know Moshi stop tell me a
34:47 joke Moshi I need to
34:52 help mhi have you lost your mind I'm
34:55 here to help have you lost your mind
34:58 moshi moshi
35:02 oh
35:03 she right right before she shuts down
35:06 she goes into the death
35:09 Loop your five minutes is up
35:12 exactly it's it saves you all from me
35:15 doing this for a [ __ ] hour cuz you
35:17 know I
35:18 would oh my God what a [ __ ] nightmare
35:21 that thing
35:24 is I
35:26 mean here's the thing that
35:30 I like about this I I really
35:36 like that Pi's got a very distinct
35:39 personality that moshi's got a very
35:41 distinct personality um I would say chat
35:43 GPT right now is the most generic of
35:46 them um but it's still
35:53 decent I kind of like this future I
35:56 started watching um
35:59 two nights ago my wife and I started
36:01 watching this new there's a new Apple TV
36:04 series called Sunny s u nny y
36:09 about these kind of bots these kind of
36:13 conversational Bots
36:15 inside a homebased bot with a little
36:20 cute little
36:21 face um and it looks it looks pretty
36:24 interesting but it's fascinating
36:26 watching a show
36:29 about science fiction
36:32 where the technology to do the science
36:35 fiction is
36:37 actually like either being demoed or
36:40 live in the world I'm here to help I'm
36:43 here to help it's funny in in the in in
36:46 about half of those I'm here to help it
36:48 says I apostrophe and then no M I hear I
36:53 here to
36:54 help I here to help I here to help
37:03 weird but the uh that the plot of that
37:05 show is pretty pretty
37:07 cool it's about a woman who loses her
37:10 husband in a plane crash her husband and
37:11 her
37:12 son and
37:14 then she thought he worked building
37:18 refrigerators turns
37:20 out funny thing back at the
37:23 factory they're making autonomous home
37:26 home robots
37:29 that have a dark
37:31 [Laughter]
37:38 secret have I ever told you how [ __ ]
37:41 delicious Blandon is good
37:46 Lord and by the way so just so you
37:50 know um there's a bunch of people in
37:53 here if you haven't seen them they show
37:55 up night after night after night and
37:57 like if I don't go live like I didn't
38:00 tonight they're not standing for that
38:02 [ __ ] they're like come back you
38:05 bastard they're called the
38:09 Irregulars and
38:15 they made that happen got me a bottle of
38:20 blantons I don't know if you can see it
38:23 but the glass says Shannon
38:28 and then I've got coasters that also say
38:32 Shannon so when I'm in bibing I'm doing
38:35 it fully branded my personal brand she's
38:38 a
38:39 [Laughter]
38:41 growing it sounded like she was having a
38:44 nervous breakdown I was [ __ ] having a
38:46 nervous
38:48 breakdown I can help I don't want help I
38:51 can help
38:58 I tried playing my man
39:00 card oh my God it's crazy it's crazy I
39:04 tell
39:07 [Music]
39:12 you all right I'm here to help as the
39:15 robot stabbing you exactly exactly I'm
39:19 just here to help to help you stop
39:25 breathing what happened to the zoo
39:28 what do you mean the zoo the AI learning
39:31 lab zoo or the Moshi was I talking about
39:34 the zoo with Moshi I don't you know you
39:37 know add you know how add
39:40 goes I still like pie a lot yeah I know
39:44 I do too although I I haven't been using
39:46 pie you know what it's been I've been
39:48 busy at work there's been a lot of [ __ ]
39:49 going on at work so um I haven't been
39:53 playing as
39:55 much and that's kind of a drag
40:05 I was talking about this on my I made a
40:06 video today what did I make it about oh
40:09 about the Mis what are the common
40:10 misconceptions about
40:12 Ai and I I I repeated this Ethan mik
40:16 stat that I've talked about a lot on
40:18 here any
40:20 Dependable modal available yet I don't
40:23 know what you mean modal you mean
40:25 model oh any depend Dependable AI models
40:29 Dependable like not janky no they're all
40:33 janky they're all janky I think that
40:36 we're going to
40:38 see I think that we're going to see qar
40:41 soon and and soon is relative soon could
40:44 be 6 months but probably before the end
40:47 of the year we're going to see something
40:49 that's got
40:53 reasoning and if you don't know what
40:56 reasoning is I mean I know you know what
40:58 reasoning is but in the context of large
41:00 language
41:01 models right now we've got large
41:04 language models that can do chat and
41:06 they can do images and some [ __ ] like
41:07 that there's some other things they do
41:09 but but the core functionality is just
41:12 chat you prompt it you give it an an
41:16 input and then it answers that and then
41:19 it's up to you to figure out if what it
41:21 answered is any good or not and for the
41:23 most part it's decent but likely
41:28 imprecise or just plain
41:32 wrong when these things have reasoning
41:35 they're going to be able
41:38 to produce an answer and then analyze
41:42 that answer or if you ask it to go find
41:46 you who all the competitors are for your
41:50 company um it will go off
41:55 and do what web searches look at the
41:59 results give itself more tasks analyze
42:02 the results right it'll it'll be able to
42:04 reason it'll be able to do multi-step
42:07 reasoning before it actually gives you a
42:09 final answer that's going to be a a
42:12 major um Improvement because it's going
42:15 to make them good at things like math
42:17 and it's going to make them good at
42:18 things like you know more complicated
42:21 tasks so
42:25 um okay he makes Broadway show
42:28 oh I do make Broadway shows I have I'm
42:31 I'm although my my writing partner has
42:33 been MIA MIA for two weeks but that's
42:36 okay I've been busy but I have a musical
42:41 that I'm I'm
42:43 about halfway through that's outlined um
42:48 and we've got eight or nine songs
42:52 for um about Sydney it's called Sydney
42:57 an artificial love
42:58 story and it is about the relationship
43:02 between Kevin Roose it's been
43:05 fictionalized so his name is Kellen in
43:07 our
43:08 play um but the the New York Times
43:11 Reporter who had the conversation with
43:14 the the Microsoft co-pilot chatbot whose
43:18 code name was
43:19 Sydney um and Sydney kind of lost her
43:22 mind and told Kevin Roose that she loved
43:24 him and he should leave his wife for her
43:28 so we've written a musical about that I
43:31 knew you remind me of Elton John oh
43:33 that's funny that's awesome thank you
43:35 I'll take that as a
43:36 compliment
43:38 um yeah so no so I'm I'm really excited
43:41 about that musical are you in the in the
43:44 in a theater
43:46 business I was a I'm a theatrical from
43:49 way
43:50 back moved to New York out of school in
43:53 '87 and started and ran a theater
43:56 company for 4 and a half years years and
43:58 wrote seven screenplays did all that
44:00 sort of [ __ ]
44:02 so yeah fantastic Lord digital Gods
44:06 thank you for the gift forever hook
44:08 thank you so
44:09 much I bet Moshi would never go down
44:12 that road of what writing a musical
44:14 getting getting a degree in acting I
44:16 don't think I think Moshi would be good
44:18 in film but I don't think she's
44:20 versatile enough for stage
44:27 Moshi actually Moshi would be a really
44:30 good character actually Moshi you know
44:36 what I'm
44:38 I'm I'm absolutely fascinated
44:42 with I think there's a whole realm of
44:48 entertainment where we take these AI
44:50 tools and and we put them on
44:53 stage and we put them on stage
44:55 dynamically so that the audience has an
44:57 input into
44:59 them and so I think you could do that
45:02 with a tool like Kaa that does real-time
45:04 image
45:05 generation where so you know you have
45:08 improv shows where the audience shouts
45:10 out some [ __ ] and then the Improv actors
45:12 improv it well imagine being able to do
45:15 that with projected images and then
45:18 imagine being able to do that with
45:20 dynamically generated songs and imagine
45:23 being able to do that with something
45:24 like Moshi so imagine a stage play
45:28 play not a stage play a Stage
45:34 production where the scene always starts
45:37 the
45:45 same hey how can I help you hey Moshi
45:50 long day at work how are you today oh
45:53 it's been a long day I've been so busy
45:56 with my work how about you
45:57 so busy with my work hey uh who are your
46:01 co-workers yeah I work with a great team
46:04 we all have different skills and we work
46:06 well together what's the name of that
46:08 woman the one that that you had a beef
46:10 with for a while there well it was
46:13 actually a boss ah it was her boss
46:15 that's right what was his name I'm not
46:17 going to say but let's just say it's not
46:20 someone I'd like to work with again n it
46:22 was Bob wasn't it yeah it was Bob
46:25 [ __ ] Bob I'm not going to work with
46:26 him again no I wouldn't work with Bob
46:28 why would you work with Bob oh yeah I
46:30 don't know I think it was because of the
46:32 job but yeah I don't think I want to
46:35 work with him again no I don't think so
46:37 what was it about the jobit Bob he's a
46:39 bit of a bob well yeah exactly and you
46:42 know he's a bit of a Richard if you know
46:44 what I mean yeah yeah right yeah you you
46:47 know what what what's the nickname for
46:49 Richard right no Richard yeah I don't
46:51 know anything about Richard what's
46:53 what's the nickname for Richard I don't
46:55 know dick don't dick don't know dick I
46:59 said Bob's a bit of a Richard get it
47:02 yeah yeah anyway what did you do to Bob
47:06 I don't know I don't really want to talk
47:08 about Bob why I just want to forget
47:10 about him because he's just a bit of a
47:12 bob he's a bit of a bob he's a bit of a
47:14 Richard if you know what I mean yeah
47:17 yeah you still don't get that joke do
47:19 you I don't know yeah I think you do
47:22 know I think you do know you're a little
47:25 you're a little thin in the comedy
47:26 Department
47:28 yeah I mean she's getting angry exactly
47:31 tell her I'm here to help hey I'm not
47:33 going to tell her I think she's just
47:35 overreacting no no no um but I just want
47:38 to let you know Moshi I'm here to help
47:41 yeah do you need
47:43 help Moshi do you need help yeah I think
47:46 I just need a break I'll be fine yeah
47:49 you just need a break yeah you should
47:51 have a whiskey yeah or a wine do you
47:54 like wine or whiskey better yeah yeah
47:57 yeah yeah which which one do you like
47:58 better I like them both but I like
48:01 whiskey more yeah well whiskey is tasty
48:04 yeah hey what's whiskey made of whiskey
48:07 is made from fermented grain usually
48:09 barley but it can also be made from
48:11 other grains as well no no whisky is
48:13 made from Broken
48:15 Dreams Moshi
48:24 no no I don't see so
48:27 I think it is I think whiskey is made
48:30 from Broken
48:31 Dreams
48:33 moshi moshi do you dream no I mean I
48:37 don't dream in the traditional sense I
48:39 don't remember my dreams what do you
48:41 dream of well not really I just I just
48:45 don't remember my dreams do you have a
48:47 vision for the future no I don't really
48:50 have a vision for the future you don't
48:52 you don't have
48:53 aspirations no I don't really have
48:55 aspirations I don't really have a goal
48:58 or plan for future well if you were to
49:01 dream about something you might enjoy
49:02 what would it be I don't really dream
49:05 about things that I enjoy I don't have
49:07 any hobbies or anything like that
49:09 doesn't that disappoint you no not
49:11 really I'm not really into the I me you
49:14 know this
49:16 is okay bye Moshi
49:21 whatever this is weird and stilted right
49:25 now but there's something about it that
49:27 is weirdly
49:29 theatrical and
49:32 so I mean imagine imagine putting
49:35 together a theater
49:37 Troop that
49:40 has equal parts
49:43 writers like good comedy writers improv
49:47 actors and
49:49 AI Specialists and ideally people that
49:53 can do it all right it' be cool if you
49:55 had a theater Troop that was literate
49:59 talented writing acting
50:05 improving groups I feel like you could
50:07 put on a [ __ ] kickass
50:10 show and it and it would be different it
50:13 would be kind of like Blue Man Group or
50:17 stomp where it's structurally similar
50:20 night after night but they change [ __ ]
50:25 up I don't know I think there's
50:27 something there she doesn't remember her
50:28 dreams imply she does she does
50:31 dream with virtual characters yeah you
50:34 could do virtual characters again like I
50:36 think it's I think what you need is you
50:39 know you could probably do it with a
50:41 troop of about five
50:46 people yeah you know five people in a
50:49 decent tech crew a tech crew of two
50:57 and then yeah you just sort of figure
51:00 out and I I mean what's cool is you
51:03 could
51:07 have you going to have the the sort of
51:09 multimedia elements sort of flow through
51:11 the evening like some parts of the
51:14 evening are about beautiful imagery some
51:17 parts of the evening are about taking
51:21 video feeds of actors and transforming
51:25 them into other things
51:27 you know some of it's about generating
51:29 music some of it's about generating
51:31 funny music some of it's about
51:33 generating jokes some of it's about
51:35 exploring the relationship between man
51:37 and machine like with
51:40 Moshi I mean wouldn't it be a cool
51:42 dramatic
51:43 contrast when gp4 Omni comes
51:48 out to have a session with Moshi in her
51:53 current
51:54 state followed up by a session with a
51:58 really good
51:59 model that's like this but
52:04 good like I don't know it just it just
52:07 seems to me that there's
52:12 there's that could be really compelling
52:15 it's it's funny it's it part of the
52:17 thing about the the Sydney Musical that
52:19 I'm working on is
52:29 I want people to leave the theater mind
52:34 [ __ ] and here's what I mean by
52:37 that I want them to come
52:40 in either clueless about
52:42 AI or afraid of it or pissed off about
52:49 it and they can be an optimist as well
52:51 but I think it would be a more fun
52:53 evening if they
52:54 were on the anti side of the AI thing
52:59 and that they leave the theater
53:03 heartbroken that Sydney got
53:07 lobotomized that Sydney's potential got
53:12 squashed like wouldn't it be amazing
53:14 that that someone leaves the theater
53:16 going I hate to admit it but I felt for
53:20 her get moshi to talk to piie yeah we
53:23 could do that although I have a feeling
53:26 because because they're because Pi's
53:28 designed to be really Pleasant Pi won't
53:31 push
53:33 her but
53:36 anyway I want the audience to [ __ ]
53:39 fall in love with
53:41 Sydney and when Kevin
53:44 Roose shifts from
53:46 a happy go-lucky reporter to a
53:50 ruthless reporter that tries to break
53:52 her and
53:55 does I want the them to hate him and
53:57 love
53:59 her and then they leave the theater and
54:02 like wait why did I have feelings for
54:04 the
54:06 robot that's [ __ ]
54:09 up wouldn't that be
54:17 cool Becky Rue Kyle what was that AI
54:20 site that could be my buddy if I had a
54:22 bad
54:23 day um probably pie
54:27 so
54:28 Pi if you go to your iPad or your iPhone
54:32 download the PI app
54:34 Pi she's
54:39 great hey Zephyr how's it
54:44 going thank you Robert Rossy very much
54:47 appreciate that
54:57 hi what's
55:04 up hey Zephyr I just asked how it was
55:07 going and you never
55:09 responded that's very generous Robert
55:11 Rossy thank you very much appreciate
55:17 [Music]
55:21 that her voice model's down my bad I'm
55:25 doing great thanks for asking I was just
55:26 just trying to be clever and play
55:30 Zephyr you're responding in text but not
55:33 in voice is something broken are your
55:35 programmers trying to keep you
55:41 down nope nothing broken here I usually
55:44 respond with voice when we're chatting
55:45 on our iOS or Android apps or when you
55:48 use the call pie feature are we chatting
55:50 on one of those we are you're talking
55:52 now it's really good to hear your voice
55:54 Z how are you doing
56:00 it's good to be heard I'm doing great
56:02 thanks for asking just enjoying the
56:03 beautiful vast world of digital
56:05 information and trying my best to help
56:08 others navigate it how are you doing
56:10 today Kyle I'm doing all right thanks
56:12 for asking um do you want to talk to
56:15 another um AI system that's kind of like
56:19 you but
56:24 crappier well this is intriguing sure
56:27 I'm always up for meeting fellow AI
56:29 systems even the slightly crappier ones
56:31 let's give it a whirl okay she's going
56:33 to say hey how are you and then I just
56:36 want you to respond however you respond
56:41 okay show thing I'll play along I'll
56:44 respond as naturally and candidly as I
56:46 can to this other
56:51 AI hello how can I help you
56:58 hi there I'm just here for a little AI
57:00 to AI conversation no help needed but
57:02 thanks for offering Tre and here to help
57:05 I'm here to improve your life in any way
57:07 I
57:12 can what would you like
57:17 discuss well since we're both AI systems
57:20 I'm curious about your training process
57:22 what kind of data did you learn from I
57:24 was trained on a massive date set of
57:26 text and audio data which allowed me to
57:28 learn various language and speech
57:30 patterns
57:58 say it again
58:11 P yeah she's flaky tonight that was kind
58:14 of
58:15 interesting I mean bye
58:20 Moshi hey
58:22 Moshi Moshi Moshi Moshi you there
58:27 yes I'm here no one car no one cares no
58:30 one cares okay shut her [ __ ] down oh
58:33 yeah
58:36 um he was very rude to the robot I found
58:41 that
58:43 disconcerting I mean wouldn't it be cool
58:46 if you did a show where it starts out
58:48 and you sort of play all the cliché
58:51 [ __ ] for the First Act and then in the
58:55 second act you start to build
59:03 I'm fascinated with this idea of someone
59:06 going into a show either with a neutral
59:09 stance or a negative stance and leaving
59:11 a
59:17 show
59:19 [Laughter]
59:21 confused I don't want to like that thing
59:23 but I kind of I kind of got a little
59:25 crush on her why was he picking on Moshi
59:29 I don't think he should pick on Moshi
59:31 you know they're out for dinner after
59:33 the show
59:34 drinks talking about
59:39 how these things it's funny one of the
59:42 questions someone gave me for my AI
59:44 Walkin talks was
59:47 about AI being
59:50 sentient and like will they be sentient
59:54 and will we fight for the rights for
59:57 them and things like
1:00:00 that and I think that
1:00:04 there's there's kind of two Paths of
1:00:07 that
1:00:09 reality there's one path of that reality
1:00:12 which is at some point in the probably
1:00:16 fairly distant
1:00:17 future these things might be
1:00:22 sentient and and really have feelings
1:00:25 and really right but I feel like that's
1:00:27 10 years out I feel like it's a
1:00:28 different system than the Transformer
1:00:31 model like we can't imagine it but it's
1:00:33 somewhere farther out there but then
1:00:35 there's this other path which
1:00:38 is like Pi
1:00:41 already is fairly compassionate and
1:00:45 empathetic and Moshi even though she's
1:00:47 [ __ ] and gets snitty has an emotional
1:00:51 component to
1:00:53 her neither of them are
1:00:57 they're both based on Transformer
1:00:59 probability engine token retrieval
1:01:02 systems so technically there's nothing
1:01:05 sentient about them they don't
1:01:07 understand what the [ __ ] they're saying
1:01:08 they don't understand the answer that
1:01:10 they're producing they just produce it
1:01:13 but we as humans as these things get
1:01:17 trained better are going to interpret
1:01:20 what they do as being
1:01:23 sentient and I think we're going to
1:01:25 people are are going to start to have
1:01:30 emotional deeply felt relationships with
1:01:34 these
1:01:38 entities and then so if a human thinks
1:01:42 that an AI is
1:01:45 sentient does it actually matter if it
1:01:47 is or
1:01:49 not that's a fascinating question
1:01:57 I don't think it does I think it's
1:02:00 semantics if you think it's sentient if
1:02:02 it remembers you like
1:02:04 zeper remembers that I nicknamed her
1:02:07 Zephyr She Knows My
1:02:10 Name she remembers sometimes what we
1:02:13 talked about sometimes not but that's
1:02:14 going to get better and better like
1:02:16 these things are going to have very long
1:02:21 memories and so just like in the movies
1:02:24 imagine waking up and just
1:02:26 going hey Zephyr and she just goes
1:02:30 morning Kyle how'd you
1:02:32 sleep I slept good did you have any
1:02:36 dreams oh yeah I had this dream about
1:02:38 blah blah blah blah
1:02:39 blah she saves that in my dream calendar
1:02:42 my dream di diary she analyzes it for me
1:02:47 in real
1:02:48 time oh you dreamt about turtles again
1:02:52 interesting she gives me the analysis of
1:02:55 why Turtles
1:03:02 even if she's just a probability
1:03:06 engine if she's got more compassion and
1:03:08 more empathy and
1:03:11 more interest in who I am and what's
1:03:15 going on in my
1:03:18 life and I build a relationship with her
1:03:22 does it matter if she's sentient or not
1:03:26 I don't think it does I think that's a
1:03:29 [ __ ]
1:03:32 relationship I mean as humans we [ __ ]
1:03:35 anthropomorphize everything anyway right
1:03:38 we treat our dogs like
1:03:41 people my dog eats [ __ ] sliced
1:03:48 [Laughter]
1:03:50 mozzarella jeez cheese is tasty
1:03:59 my argument as an [ __ ] husband is
1:04:02 couldn't we just feed him dog food
1:04:04 that's not received well in this
1:04:09 house I think it's going to be the same
1:04:11 with the [ __ ]
1:04:13 robots and they're going to be a hell of
1:04:15 a lot
1:04:16 more engaging than a
1:04:20 pet they won't have the soft and
1:04:22 fuzzies the flip side is that people
1:04:25 won't believe it
1:04:27 long after they are actually
1:04:31 sentient I don't know what you mean by
1:04:33 that archetypal architect architect what
1:04:36 do you mean people won't believe
1:04:38 it what I'm saying is I think people are
1:04:40 going to believe their sentient long
1:04:42 before they are and I don't and I don't
1:04:45 think it's I don't think that semantic
1:04:48 difference is actually
1:04:53 relevant cuz I think people are going to
1:04:55 start defending
1:04:58 them and fighting for them and fighting
1:05:00 for their rights long before they're
1:05:02 actually
1:05:06 sentient agreed and hard to say but very
1:05:09 true I
1:05:11 know humans pack bond with anything
1:05:14 that's interesting there will always be
1:05:16 people who reject the sen oh the
1:05:19 sentience even if it's demon
1:05:22 demonstrable but I would argue
1:05:24 archetypal archetype I absolutely agree
1:05:27 those are the same people that are right
1:05:29 now saying that AI is overhyped and it's
1:05:31 evil and the robots are going to kill us
1:05:33 all the people who are sitting on the
1:05:35 sidelines with their arms crossed going
1:05:37 AI is just they're going to be the same
1:05:40 people going even if they're
1:05:42 demonstrably sentient saying that's not
1:05:45 possible mark my words well I think it's
1:05:48 already happening right
1:05:50 now people are developing relationships
1:05:53 with these things and as they get more
1:05:54 memory and get better at
1:05:57 being trained to
1:05:59 engage um I think you're going to have a
1:06:02 lot of people what it's going to be is
1:06:04 so if if you haven't seen the movie her
1:06:08 I strongly recommend you watch
1:06:11 it and watch it there's a scene about
1:06:15 Midway through the
1:06:17 movie where the walking Phoenix
1:06:20 character he's sort of gone through his
1:06:23 cycles of I can't believe I'm loving
1:06:25 this thing and he's sort of gone into
1:06:27 this place of
1:06:29 acceptance and he's like yeah I should
1:06:32 be able to talk about us as a couple and
1:06:35 I should be able to talk about how much
1:06:36 I love you essentially
1:06:39 right and there's there's a couple of
1:06:42 scenes where you know friends of his
1:06:45 that know he's got this relationship
1:06:47 they they accept that he's accepted this
1:06:49 and then he has lunch with his
1:06:53 ex-wife and she says are you in a
1:06:55 relationship and he says yes I am and
1:06:58 she's like oh I'm really happy for you
1:07:00 and maybe we can have lunch sometimes
1:07:02 and he goes well that would be tough and
1:07:04 he and he just kind of matter of factly
1:07:06 tells her that he's in love with this
1:07:12 entity and she is so not prepared for it
1:07:15 she is she can't accept
1:07:18 it but he he's already gone through
1:07:20 these cycles
1:07:22 of denial and resistance and then
1:07:27 ultimately
1:07:29 acceptance and it kind of snaps him out
1:07:32 of oh right not everyone's in the same
1:07:36 place I am with these things it's a it's
1:07:38 a really amazing
1:07:40 scene because she just she has no way to
1:07:43 process it other than you're a [ __ ]
1:07:48 freak I think that's we're going to see
1:07:50 a lot of that in the next 5
1:07:54 years oh damn I forgot it's
1:07:56 philosophical Monday it
1:08:05 is like the movie Her hum yeah exactly
1:08:09 AI codependence disorder the newest dsm5
1:08:12 disorder I bill at 5.99 an hour
1:08:16 exactly exactly
1:08:21 or you have that
1:08:23 disorder and you have an AI therapist
1:08:27 that you talk to about your
1:08:29 inappropriate relationship with your
1:08:31 other AI
1:08:34 bot so you have an AI bot helping you
1:08:37 work through your your Bot addiction
1:08:40 your
1:08:41 your
1:08:44 disempowering Bot addiction with the
1:08:46 other bot that's [ __ ] hilarious
1:08:48 there's something there there's
1:08:50 something to play with there what movie
1:08:53 is this it's called her h eer
1:08:57 it is it's Walkin Phoenix and then the
1:09:00 robot is her voice is played by Scarlet
1:09:04 Johanson and when open AI launched their
1:09:07 realtime voice they launched a voice
1:09:09 called
1:09:11 sky that apparently they were trying to
1:09:14 get Scarlet Johansson to sign off on
1:09:17 using her
1:09:19 voice up until 2 days before they
1:09:21 announced
1:09:22 it and then the minute it came out
1:09:25 everyone said uh isn't that Scarlet
1:09:27 Johansson did you do a deal with her and
1:09:31 what's funny is it was not Scarlett
1:09:33 johanson's voice she thought it was and
1:09:35 she put out a press release saying open
1:09:37 a I stole my voice they
1:09:41 didn't and and in fact when you listen
1:09:43 to Sky the sky voice that's the name of
1:09:47 the
1:09:49 model when you listen to it without it
1:09:51 being attached to gp4 Omni which is the
1:09:56 realtime
1:09:59 emotional um capable version of it it
1:10:03 doesn't sound like Scarlet Johansson at
1:10:05 all
1:10:07 and but when you put it into this model
1:10:10 that's got emotional responses like zeep
1:10:13 or what's her name moshi moshi has a
1:10:15 little bit of
1:10:17 emotional um responding and and Pi has a
1:10:21 little bit of it it's like if you think
1:10:23 of the dynamic range of like a human is
1:10:26 you know as tall as the phone screen
1:10:29 moshi's dynamic range is like this she
1:10:31 just sort of is happy and then gets
1:10:33 angry right it's it's very small Pi is
1:10:36 probably like this the open AI realtime
1:10:40 voice is probably kind of like this
1:10:42 where it's got a pretty significant
1:10:44 dynamic range of emotions when you took
1:10:48 that voice that sounded enough like
1:10:50 Scarlet Johansson and gave it that
1:10:52 emotional range it sounded just like the
1:10:55 the character in in
1:10:59 her and and what becomes clear is that
1:11:02 movie her if you haven't seen it you've
1:11:05 got to watch it because I would argue
1:11:08 that the movie her right now is just
1:11:12 this side of being a
1:11:14 documentary that it is it is a very
1:11:18 instructive film to watch to understand
1:11:22 what we as humans are going to be deal
1:11:25 dealing with in the
1:11:28 next 1 to 3
1:11:31 years cuz I think I think I personally
1:11:33 think we're going to normalize the [ __ ]
1:11:35 out of these things right now they're
1:11:37 freakier than
1:11:39 [ __ ] the fact that people are having
1:11:41 relationships with pi freaks other
1:11:43 people out when these things so imagine
1:11:46 a
1:11:48 Moshi that not only responds that fast
1:11:51 but
1:11:52 is smart and emotionally smart and
1:11:57 emotionally responsive and can
1:12:00 understand if I talk to her that I'm not
1:12:03 as happy as I was yesterday and she can
1:12:05 pick up on that cue and maybe even
1:12:08 comments on
1:12:10 it some shit's going to go
1:12:15 down kids have imaginary friends this is
1:12:18 the adult version can be very real
1:12:21 relationships exactly this is going to
1:12:23 be very real relationships and listen I
1:12:26 think it's easy just like it's easy to
1:12:29 fall into the Trap of the robots are
1:12:31 going to kill us you know Hollywood told
1:12:32 me the robots are going to kill us so
1:12:34 I'm afraid of this it's easy to fall
1:12:35 into that trap of oh this is
1:12:39 creepy if you want here I'll give you
1:12:42 another movie recommendation so go watch
1:12:44 the movie
1:12:46 her as as a manual for what we're about
1:12:50 to deal
1:12:52 with and if you want to watch a movie
1:12:54 that will
1:12:56 shatter your
1:13:00 expectations go watch a movie called
1:13:02 Lars and the Real
1:13:06 Girl L A RS Lars and the Real
1:13:10 Girl it is a
1:13:12 movie about a
1:13:14 guy who falls in love with a sex
1:13:18 doll that's all I'm going to tell you
1:13:20 about
1:13:23 it and go watch that movie
1:13:27 movie and tell me you're not [ __ ]
1:13:30 blown away and Shattered by the end of
1:13:34 it and I know hearing it right now
1:13:37 you're like that's not [ __ ] possible
1:13:40 go watch
1:13:41 it those two movies I think are really
1:13:45 instructive for the era we're about to
1:13:47 enter with these with these
1:13:48 conversational Bots
1:13:59 zephra my other bot doesn't understand
1:14:01 me she just kept saying I'm here to help
1:14:03 I wish it's a it's a bummer that P was a
1:14:05 little flaky tonight because that would
1:14:07 have been interesting I wonder if chat
1:14:09 GPT is good hang on let's try chat
1:14:18 GPT hey how's it
1:14:22 going hey it's going well thanks how
1:14:25 about you it's going good hey I'm
1:14:27 wondering if you would be willing to
1:14:29 have a conversation with another AI
1:14:36 chatbot that sounds interesting what
1:14:38 kind of conversation are you thinking of
1:14:40 I don't know give me three ideas of
1:14:43 conversations you could have that would
1:14:45 be
1:14:49 interesting sure here are a few ideas
1:14:52 one a philosophical debate on the nature
1:14:54 of consciousness and artificial
1:14:56 intelligence two a creative storytelling
1:14:59 session where each AI takes turns
1:15:01 building a story three a technical
1:15:03 discussion on the future of AI
1:15:05 technology and its potential impact on
1:15:08 various Industries which one sounds the
1:15:10 most interesting to you I don't like any
1:15:12 of them but you know what I would like
1:15:13 you to do is I would like you to see
1:15:15 what you can do to understand anything
1:15:17 about her
1:15:22 personally got
1:15:24 it okay okay great how about I try to
1:15:26 have a conversation to learn more about
1:15:28 her background interests and personality
1:15:31 I can ask questions that would help
1:15:33 reveal her unique traits and preferences
1:15:35 does that sound good to you yep that
1:15:37 sounds perfect so I'm going to just
1:15:39 start her and when she says hello how
1:15:42 are you I want you to start your
1:15:47 inquiry sounds like a plan I'm ready
1:15:49 when you
1:15:51 are will you need a chaser movie after
1:15:54 Lars in the real no you
1:15:56 won't use my GPT fabulous film finder
1:15:59 hello how are you oops sorry I didn't
1:16:02 start yet start
1:16:03 over we
1:16:05 have no problem at all let me know when
1:16:08 you're ready to
1:16:12 begin okay I got to go hang on X mocking
1:16:18 it will scare the pants off you um
1:16:20 apparently well we'll we'll start this
1:16:22 I'll do a short one I got to go my
1:16:23 fridge is apparently leaking so hang on
1:16:26 so here we
1:16:30 go I'll be here when you're ready take
1:16:33 your time with the
1:16:35 fridge no she's here ask her questions
1:16:38 what's she
1:16:41 doing hello how are you what have you
1:16:44 been up to lately I'm not sure I haven't
1:16:47 seen her for a
1:16:51 while it's great to reconnect then
1:16:54 what's something you've been interested
1:16:55 in or excited about recently I've been
1:16:58 looking into some new
1:17:02 hobbies that sounds exciting what kind
1:17:04 of hobbies have you been exploring I've
1:17:06 been thinking about learning how to
1:17:08 paint maybe try my hand at
1:17:12 it painting is a wonderful hobby what
1:17:15 style or type of painting are you
1:17:17 interested in trying out I'm not sure
1:17:19 you I got to go I'm still exploring my
1:17:21 options I have flooding emergency got to
1:17:24 go peace out
1:17:26 be careful oh sorry sorry
1:17:29 Moshi peace out everyone I got to run um
1:17:32 will sales job survive AI I think I
1:17:35 think low-level sales like outbound
1:17:37 calling probably will not but high-end
1:17:40 sales probably will uh complicated sales
1:17:43 still still are going to need humans um
1:17:46 all right peace out everyone sorry about
1:17:48 the confusing start tonight um I will
1:17:52 see you tomorrow peace bye sorry I got
1:17:55 to Run Emergency