aiLL Nov 22, 2023 - OpenAI's AGI Breakthrough, Runway's AI-Generated TV, and Building Community

Description

Chapters

Transcript

0:03 [Laughter]
0:09 yeah that was
0:21 funny hey Silver
0:24 Fox let's see when Becky jumps in here
0:28 t-rock what's happening n
0:32 good day good
0:34 day steo was shaking Happy Thanksgiving
0:40 everybody everybody home with family
0:43 doing family [ __ ] we're just here in
0:46 Danver as you can
0:48 see from my this is just a synthesized
0:52 background it's so
0:56 beautiful oh yeah yeah yeah yeah yeah
1:00 yeah yeah yeah yeah
1:02 yeah H my shoulder's a little janky my
1:05 shoulder is more janky than this AI
1:09 [ __ ]
1:12 like I did a lot of cleaning today so I
1:16 wasn't super up up on uh all the AI [ __ ]
1:20 but you know what are you going to
1:22 do I do want to try a new thing I saw
1:25 that hey Jen has
1:27 added Avatar building so you can upload
1:29 like a 2minute video of yourself and
1:31 then just make a make an avatar of
1:34 yourself and just type in what you want
1:36 you to say which is wild it's wild and
1:39 wacky stuff wild it's wild it's wild
1:42 it's
1:43 wacky you can make
1:45 money are you in here Becky I just got a
1:48 uh I just got a uh a note from her right
1:51 before I went live she goes I'm waiting
1:53 here twiddling my thumbs Paul's
1:58 asleep that's pretty
2:00 funny oh man welcome
2:05 everybody today was a relatively
2:08 non-eventful day you look refreshed
2:11 thank you well I just I just put some uh
2:14 some water through the
2:16 hair what's with Q being a factor in the
2:19 soap opera yeah I just saw that right
2:21 before I came on uh Kevin um I do not
2:24 have a subscription to the information
2:27 which is a newsletter or a news source
2:29 which the sale price is
2:32 $749 a year so uh so if anyone wants to
2:37 sponsor me to get get a subscription to
2:41 the information I'll take it but um so I
2:45 haven't I haven't read the article but
2:47 yeah
2:49 apparently there was some speculation
2:51 that a significant breakthrough was
2:55 made um and it I I think the
2:58 Breakthrough from what I can guess GA
3:00 what Q
3:02 is uh is not the a non q a non nonsense
3:09 um it's a method that allows large
3:12 language models to basically do better
3:15 reasoning like math and getting things
3:18 right and being able to take actions and
3:21 and doing so in
3:23 a in
3:25 a I guess in a predictable way so I
3:28 don't know enough about it but
3:30 apparently there was a breakthrough and
3:32 and the significant thing about if there
3:36 if there was a breakthrough that they
3:37 can categorize as
3:40 AGI in their Charter it says they get to
3:44 determine when they get to AGI and when
3:47 they get to AGI all
3:49 licensing deals are off with say
3:54 Microsoft so did you see that GPT on
3:57 Discord about 749 pizzas no I did
4:03 not I did not Silver Fox where is that
4:07 is that in sharah Shar
4:10 GPT let's
4:12 see Vault of Secrets the eliminate
4:16 elimination Fades AIP meal mail meal
4:20 planner thoughts about
4:23 nothing no I didn't see
4:26 it I've been I've been busy I've been
4:28 back and forth between the new networks
4:30 and the Discord and things like that so
4:32 anyway um reuter reported on it oh okay
4:36 um let me see let me go see if I can see
4:38 find
4:41 it let's see
4:45 Reuters uh Q open AI That's not Q that's
4:50 one
4:54 Q
4:56 brers head of CEO ouer
5:23 after being contacted by Reuters open AI
5:26 which declined to
5:27 comment acknowledged an intern message
5:30 to
5:32 staffers I guess about they missed a
5:34 word everyone's reporting fast fast and
5:37 loose even
5:39 Reuters uh a project called q and a
5:42 letter to the board before the weekend's
5:44 events one of the people one of the
5:46 people
5:48 said I guess when you got 700 people on
5:51 feeling like they're on thin ice
5:53 someone's going to talk um an open AI
5:56 spokesman said that the message sent by
5:57 longtime executive Mir moradi
6:00 alerted staff to certain media stories
6:02 without commenting on their
6:04 accuracy some in open AI believe Q
6:08 pronounced qar oh it's Q asterisk
6:11 qar I'm I'm glad they didn't just call
6:13 it Q because that would have getting
6:15 confusing with uh other news that we're
6:17 going to hear in
6:18 2024 um could be a breakthrough in the
6:21 startup search for what's known as
6:22 artificial general intelligence AGI one
6:25 of the people told Reuters um open AI
6:28 defines AGI as autonomous systems that
6:30 surpass humans in most economically
6:33 viable tasks take that in AGI are
6:38 autonomous
6:40 systems so you don't have to prompt them
6:43 they just go that surpass humans in most
6:47 economically viable
6:51 tasks let me read that one more time
6:55 open AI defines AGI as autonomous
6:58 systems that surpass humans in most
7:00 economically viable tasks giving vast
7:03 Computing resources the newest model was
7:05 able to solve certain mathematical
7:07 problems the person said on condition of
7:10 anonymity because the individual was not
7:12 authorized to speak on behalf though
7:14 only performing math on the level of
7:16 grade school students acting acing such
7:20 tests made researchers very optimistic
7:22 about Q's research yeah one of the
7:24 things I read about it is they did it on
7:26 a very small model um and so they I
7:30 guess what they cracked is the method
7:32 for how to how to get it to do reasoning
7:34 they did it on a very very small model
7:37 but the theory is if you cracked
7:39 reasoning on a small model then just
7:41 like open AI you just scale it and U you
7:45 know the theory is that it's going to
7:47 scale you know it's it's
7:50 predictable
7:51 um researchers consider math to be a
7:54 frontier of generative AI
7:58 developments
8:01 all right so it's possible I mean listen
8:03 that was that was kind of the
8:05 speculation here was that something
8:07 something happened that caused an
8:09 urgency and I I certainly don't think
8:13 that Sam Alman announcing the GPT store
8:19 um Rose to the level of someone's got to
8:23 go so um that said I don't know why them
8:27 discovering AGI would be a reason for
8:30 him to go but whatever um
8:32 okay welcome
8:35 everybody let's let's jump
8:38 over to this here
8:44 nonsense we'll get started here so uh
8:47 first of all Happy Thanksgiving a day
8:51 early um I think I'm going to take
8:53 tomorrow off I'm not going anywhere I
8:54 could I could do a show but we've got
8:56 friends coming over and it would be late
8:58 and I think everyone else is going to be
8:59 all
9:00 trip to F out and you know you need you
9:04 need time to argue with your families
9:06 I'm curious actually um pop in the
9:09 comments below I'm curious if anyone is
9:11 home for the holidays that that
9:12 especially regulars here um or talking
9:15 to their
9:16 families are are the families talking
9:19 about open AI are they're talking about
9:21 Sam Alman are they talking about AI at
9:23 all are they just completely oblivious
9:25 to it I'm I'm curious what's happening I
9:27 saw a couple of tweets that were were
9:29 like uh open open AI came up within 5
9:33 minutes of you know one person getting
9:35 home for the family did you see that GPT
9:39 disc oh no okay that that I did not
9:43 see only Tans
9:49 um writers doesn't say what Q is it does
9:53 and there's there's a if you go look on
9:56 um X there's a little diagram hang on
10:02 [Music]
10:11 on open AI Q oh qar actually Q
10:25 star that
10:26 thing so that
10:31 with with qar open AI have likely solved
10:35 plan planning SL agentic Behavior as in
10:40 agents as in autonomous agents as in you
10:43 give them a goal and they go work [ __ ]
10:46 out for for small models scale this up
10:50 to a very large model and you can start
10:52 planning for increasingly abstract goals
10:54 it is the fundamental breakthrough that
10:56 is the Crux of a agentic behavior
11:01 so you got your
11:02 Q S1 comma A1 with the O and the Q S1 S2
11:08 comma A2 and the Q comma parentheses S3
11:12 comma A3 P's in here Pate probably knows
11:15 what this means oh you know what we can
11:18 do oh this will be fun let's let's grab
11:21 this here
11:22 image and let's have uh GPT Vision tell
11:26 us what the [ __ ] it
11:28 means okay so this is
11:33 qar um
11:36 math that so so if you haven't used chat
11:39 GPT before one of the things that is a
11:41 really hard habit to break like I'm not
11:44 good at math I have a degree in acting I
11:46 can do words good I don't do math good
11:51 um one of the thing that these large
11:53 language models do is they you can be
11:56 good at stuff you're not good at or at
11:58 least understand it without having to go
12:01 Google it and get a Stamford continuing
12:04 ed
12:05 degree
12:07 um Q maath jpeg
12:10 okay um can
12:15 you explain the
12:18 math
12:20 here or what
12:23 this model is
12:27 uh saying
12:32 all right come on chat
12:34 GPT the image you've uploaded appears to
12:37 be a state action wait a state action
12:41 reward State wait a sarsa Model A State
12:45 action reward State action I if Pate are
12:47 you in here if pate's in here Pate knows
12:51 I bet I bet pate's been saring up up a
12:55 storm a form of reinforcement learning
12:57 this is not strictly a meth mathematical
12:59 formula but rather a conceptual diagram
13:02 for an algorithm used in machine
13:04 learning okay one state so the
13:08 S okay State and action so the S is
13:12 State a is
13:14 action this is the situation in which
13:17 the agent currently finds itself in the
13:19 diagram these are represented by S1 2 3
13:22 4 a this is what the agent can do in
13:26 each state okay represented by a 1 2 3 4
13:29 reward when an agent takes an
13:33 action it receives a a a reward or
13:36 penalty from the environment that's not
13:39 explicitly shown but it's an integral
13:41 part okay so it's some state it takes an
13:45 action and then I guess this is the
13:47 reward system or the penalty system and
13:49 it says you did good or you did bad and
13:52 then it tries another one it says you
13:53 did good you did bad and tries another
13:55 one that's cool
13:58 okay Q
14:00 value this is a function that estimates
14:03 the total amount of rewards an agent can
14:05 expect to accumulate over the future
14:09 starting from a given State and taking a
14:12 particular action the values q s to the
14:15 one a to the 1 q s the 2 a to the two
14:19 represent these estimates the sarsa
14:20 algorithm updates the Q value okay so
14:23 that so this makes sense so so you give
14:26 an agent a goal and it either gets
14:29 rewarded or penalized and so so think of
14:32 it like card counting so the Q goes up a
14:35 point or down a point and then it tries
14:37 another thing in a so now the state's
14:39 different it takes a different action
14:41 and then the the Q goes up or down right
14:44 and so basically it's trying [ __ ] and
14:47 learning along the way and I assume what
14:50 you do is you program it so that you
14:52 want Q to go
14:54 up
14:58 um
15:00 the colored Square the red one with the
15:01 O likely represents the current state
15:03 action pair under consideration these
15:06 arrows rep represent the possible
15:08 transitions to the next state with the
15:11 corresponding actions the green square
15:13 might represent the terminal state or
15:16 state of a different property the sarsa
15:18 model is on policy meaning it evaluates
15:22 and improves the policy being used to
15:25 make decision as opposed to off policy
15:28 like Q learning that evaluate a greedy
15:31 policy okay see now now we're getting
15:33 into Stanford Mass [ __ ] I don't get it
15:35 but that's okay whatever basically it
15:38 tries [ __ ] and then it gets better or
15:42 not um how can it get back to base
15:47 Q we we are already significantly deeper
15:51 than my ability to understand any of
15:52 this although Hey listen cat jpt just
15:55 explained it in a way I can kind of get
15:57 it so apparently this this is a big
15:59 breakthrough so or they got it to they
16:01 got it to do some reasoning Kyle um
16:04 didn't you say please or thank you shame
16:06 on you I know I didn't did did I what
16:09 did I say can you please no oh oh no I
16:11 didn't say please can you explain the
16:13 math here what the model is saying yeah
16:15 it's really bad you got to be nice to
16:17 the robots people Ethan mullik just
16:19 released a paper it it the fundamentals
16:23 of psychology actually work with large
16:26 language
16:28 models
16:29 treat your robots nice dang
16:32 it all
16:36 right up up yes yers I will be getting
16:39 my 16-year-old grandson up to speed
16:41 possibly my 20-year-old
16:45 nice um what did I miss um so what you
16:48 missed is there's a
16:52 Reuters there's a Reuters well there's a
16:55 there's a piece from the information
16:56 that I assume inspired these Reuters to
16:59 go do this so so this article basically
17:02 said that shortly before Sam Alman was
17:07 fired um there was a breakthrough there
17:10 was a letter sent and a breakthrough in
17:14 um
17:16 reasoning a step toward AGI
17:20 basically and I guess it was significant
17:23 enough that it it
17:25 caused I don't know if it caused it I
17:28 don't know if it's related there is
17:30 specula there was speculation on Friday
17:33 that they must have stumbled onto
17:36 something AGI like and stumbled on is
17:39 the wrong word Ilia probably programmed
17:42 it and it
17:45 worked uh have we talked about qar yet I
17:48 was just talking about it yeah in fact
17:50 so what's cool
17:51 is um we there's your star there's your
17:55 Q Star right so there's all your
17:57 q's and and so um this diagram was out
18:02 on Twitter talking about q and then I
18:05 didn't know what the hell this meant and
18:06 so I went to chat GPT and and I didn't
18:09 say please but I said can you explain
18:10 the math here or what this model is
18:12 saying and it explains it it's a sarsa
18:15 model it's a state action
18:18 reward State action model so you give a
18:24 large language model a
18:27 goal it's got some it's got a state that
18:30 it's in it's got some action it can take
18:32 it takes that action and then it either
18:33 gets a reward or a penalty and the Q
18:37 carries that reward or penalty value to
18:40 the next state and then it takes another
18:44 action and then it gets rewarded or
18:46 penalized there and then it carries that
18:48 value to the next state and then it so
18:51 basically it's learning right so it
18:53 tries something and then it gets told
18:56 did you succeed or not so so this could
18:58 so so think about this just in terms of
19:00 hallucinations right you ask it to give
19:03 you the answer to some historical
19:05 Challenge and it gets half of it right
19:07 and it makes half of it up and it's
19:09 completely wrong well with with a model
19:11 like this you could say you know give me
19:13 an answer I'm in this state someone just
19:16 asked a question so give them an answer
19:20 and then um it'll get penalized or
19:23 rewarded and then it'll you know it
19:26 might do that two or three times before
19:27 it actually gives you an answer answer
19:29 and then it would give you right
19:35 answers so Mouse in amazze stuff yeah
19:38 it's pretty much mouse in a maaz stuff
19:40 yeah it's exactly what it is and and Q
19:42 is the memory right so Q is carried
19:46 between these different 1 two 3 4
19:49 States right so it's basically it's card
19:52 counting if if you know card counting
19:54 this is Vegas card counting you you're
19:57 positive or negative didn't we see this
19:59 in war games probably I mean listen a
20:01 lot of the a lot of the uh a lot of the
20:03 basic mathematics here the components of
20:06 it have been around for a while so it's
20:08 it's these you know researchers trying
20:11 things in different
20:12 combinations uh and and applying it to
20:15 whatever these things do so wild wild
20:19 and wacky stuff but I think this does
20:25 um this does lend cred ability to David
20:29 Shapiro's uh prediction that we're going
20:31 to see AGI within 12
20:33 months which is absolutely nuts like I I
20:38 don't really know what that means
20:41 but when you when you go to that that
20:44 reuter's uh
20:46 article this line open AI defines this
20:54 line open AI defines
20:57 AGI Tucker ramage I'm gonna let you up
21:00 on
21:01 stage oh Tucker hey what's uping do you
21:05 know anything about this qar [ __ ] qar
21:10 what exactly are you doing I can't see
21:12 it oh no you don't all right I thought
21:14 you might have been a mathematician
21:16 coming on here to school me no it's it's
21:18 a it's a breakthrough that open AI had
21:21 oh okay all
21:24 right about it no are you doing stuff
21:27 with AI
21:29 I use AI from time to time but I don't
21:31 know very much about it now just for
21:34 porn NOP not
21:39 quite Co good to see
21:44 [Laughter]
21:50 you him thought him was gonna troll me
21:53 didn't
21:54 him him him don't know me for as
21:57 autonomous systems that surpass humans
22:01 in most economically valuable
22:11 tasks so yeah so that's a big deal
22:14 because they're infinitely scalable
22:16 right and the way and the way they've
22:19 created the agent
22:21 um if you go play with the API agent uh
22:25 thing on the back end the assistant oh
22:28 scared them away I know what are you
22:30 going to
22:34 do that was pretty
22:38 funny oh my
22:42 God that was so
22:45 risky
22:46 [Laughter]
22:49 whatever
22:51 oh did they want to battle exactly
22:56 battle those those were those were dar's
23:00 [Laughter]
23:05 henchmen
23:08 oh we want a battle we want a battle
23:11 there will be no more battles on this
23:15 channel I thought I looked it looked
23:17 like his picture like he looked like a a
23:20 young mathematician so I thought someone
23:22 was coming in here to tell me about
23:25 qar but if anyone knows about that I
23:27 don't know if p in
23:29 here welcome to popey can I take your
23:33 order I like the Jack Nicholson laugh
23:36 thank you I didn't know it was one but
23:40 awesome um better start learning the art
23:44 of
23:45 Butchery we want a battle we want to
23:47 battle who were they I they were just
23:49 someone I get you know on On Any Given
23:53 evening I get you know four or five
23:54 people that are wanting to come up on
23:56 stage and most most of them are probably
23:58 people wanting to battle um so I just on
24:01 a lark I just kind of popped them up
24:03 there cuz I was talking about math and
24:04 they they look kind of like college kids
24:08 so that was all um unpack economically
24:13 viable so so think anything that can be
24:16 done with a computer law clerk um
24:20 researcher um
24:22 consultant uh agency account person um
24:26 project manager
24:31 any wait sorry most economically
24:35 valuable tasks so basically what people
24:39 get paid for now this is primarily
24:41 knowledge work until you put these tools
24:45 inside the robots that Elon Musk and
24:48 five or six other companies are making
24:50 right now that are they're basically in
24:52 testing right the humanoid
24:54 robots so at that point these things can
24:57 then do phys physical tasks as well but
25:00 basically any knowledge
25:02 work when we hit
25:04 AGI autonomous
25:07 systems that
25:09 can surpass humans in most economically
25:14 valuable
25:16 tasks um so so it's it's it's a big deal
25:20 it's a big deal and that's I mean and
25:22 that's that's why that's why open AI is
25:26 structured so weirdly it's why they've
25:27 got this weird clause in the deal it's
25:30 why they've got they invented this thing
25:32 called the Capp profit um
25:36 model and they've got this weird thing
25:39 that says when they declare that AGI has
25:41 been hit all licensing deals are off I
25:44 guess for the new stuff I I assume that
25:46 the licensing deal is still there for
25:50 gp4 um uh what did Becky say hang
25:55 on Sir this is a Wendy's
26:01 Kyle where have I failed you no you
26:04 haven't failed me Becky you right as I
26:06 was going live I saw your thing it's
26:07 like I'm tapping my finger here I
26:09 thought it was really funny and then I
26:11 didn't see you jump in so hello Happy
26:13 Thanksgiving battle more Kyle it
26:15 releases
26:17 stress they were from
26:20 [Laughter]
26:24 anthropic oh man T Tik Tock teaches kids
26:27 to use chat GPT
26:31 4 man they were they were pretty uptight
26:34 what is it with kids being uptight these
26:35 days I thought that would have gotten a
26:37 laugh no apparently
26:42 not oh
26:46 man um all right so so let's do this um
26:50 if you're new to uh to Ai and chat GPT
26:55 or let's say you're home with your
26:57 family and and you said to your family
26:59 hey want to see what I do sometimes I
27:03 watch this old dude on Tik Tok and if
27:05 you're sitting there with family members
27:06 right now sorry family members you're
27:09 right they're weird um but but uh but
27:13 I'll I'll walk you through you know
27:15 where to start so the the first place to
27:17 start if if you haven't played with chat
27:20 gbt is to go play with chat GPT now that
27:23 Sam Sam alman's back it seems like we we
27:26 narrowly averted this thing complet
27:28 blowing up so I I think I am hopeful
27:32 that they actually get the contract
27:35 signed and everything stays stable um it
27:40 looks like this company is going to be
27:42 the leader in this space for
27:44 sometime Google apparently has a model
27:48 that's a competitor to this but they
27:50 can't seem to launch it and I saw a
27:52 thing today that said that their
27:54 internal culture is a mess right now and
27:57 that's one of the challenges of getting
27:59 that out um open ai's culture right now
28:03 seems pretty remarkable when they fired
28:05 Sam you had like 97 or 98% of the
28:09 employees signed a piece of paper that
28:10 said if the current board doesn't resign
28:13 we're out of here 90 like 7 it was like
28:18 770 out of 790 employees or some crazy
28:22 [ __ ] like that so chat GPT right now the
28:27 the first bullet point there it's true
28:29 multimodal meaning it can see images it
28:32 can make images it can do um data
28:36 analysis it can do all the language
28:39 stuff it can talk it can
28:42 hear and
28:44 so it is an incredibly powerful
28:47 tool and they also released these things
28:50 called custom gpts which are basically
28:53 like you can create in minutes an
28:56 incredibly powerful ful personalized
28:58 tool like let's say you run a gardening
29:00 store and you want to provide a tool for
29:03 your customers to be able to um plan
29:06 their Gardens for different times of the
29:08 season and put you know sort of subtle
29:11 branding and ads in there um as they ask
29:14 their questions you can do that you can
29:16 build it in minutes you can share those
29:18 things like that so um so go play with
29:22 that um I think that's it let's let's
29:25 let's do some Q&A tonight open it's like
29:28 Bobby it's like a Bobby Ying in the
29:31 shower moment on Dallas man that's some
29:34 Gen X throwback Bobby Ying I was like
29:37 Patrick Ying is this is this a Nick
29:39 thing no you went further back than that
29:41 digital Gods thank you so much for that
29:43 I really appreciate the gifts um you
29:46 called them out on their sins he he did
29:48 he did get a little flustered didn't he
29:50 he was like no I don't do porn what do
29:52 you mean
29:56 porn
30:00 texted upright because they're actually
30:02 intimidated by us oldies but goodies
30:05 damn right don't [ __ ] with Gen X do not
30:07 [ __ ] with Gen X uh they wanted to come
30:10 on they get on and tanked kids kids
30:13 these
30:15 days qar Kyle what's up so we were just
30:18 looking at it um what's up is it looks
30:20 like they made some
30:23 progress they probably Ilia made some
30:27 progress
30:28 with a thing called qar so there is your
30:30 qar see q q q q in the shape of a star I
30:35 guess that's why they named it
30:37 that but I'm explaining this as if I
30:41 know math but I don't I just asked chat
30:43 GPT to tell me what the hell this was
30:45 and it did Isn't that
30:49 cool R is the remed immediate
30:56 reward
30:58 so basically what happens is you
31:02 have I guess a large language model but
31:04 you have one of these
31:06 tools in a state and then given the
31:10 state it's got some actions it can
31:11 perform so S1 is State A1 is actions and
31:16 then Q is basically the score maybe
31:19 quantity I don't know I don't know what
31:20 Q stands for but Q is the score so in
31:24 this state it takes an action and then
31:27 some
31:28 tool some measurement tool basically
31:30 says was that action good or bad and
31:33 then it gives it a score and that score
31:35 goes into q and then it moves on to the
31:38 next state it says okay given that you
31:39 took that action here's the new state
31:41 and here's your new set of actions you
31:43 can take and so it takes an action there
31:45 and so basically it's just sort of
31:47 adding or subtracting to Q's total you
31:50 know as it goes from state to state to
31:52 state and takes action after action
31:53 after action um so so this is apparently
31:58 the model that they had a breakthrough
32:02 around um and and a a
32:05 uh a letter was written shortly before
32:08 Alman got fired so that's that's what
32:11 Dad is that's what Dad is but I I mean
32:14 hey not for nothing if you have
32:16 not if you have not played with chat gp4
32:20 did you know it could do [ __ ] like this
32:22 like you can upload a math diagram like
32:24 this and say hey what is this and it
32:26 figures it out for you I I mean just
32:30 this stuff continues to blow my mind so
32:33 what happened with Microsoft's so what
32:35 happens with Microsoft's
32:38 co-pilot um you mean if they hit AGI and
32:41 if all licensing bets are off I
32:45 assume I assume that part of the
32:47 licensing deal is they get to keep any
32:50 non
32:52 AGI um technology so gp4 is is not AGI
32:58 maybe it's gp5 maybe it's this project
33:00 called qar I don't know if they're the
33:02 same thing right gp5 could just be a
33:04 more scaled version of gp4 qar might be
33:08 a whole different layer that sits on top
33:10 of the base the the frontier models like
33:12 I don't know um so Microsoft Microsoft
33:16 is fine like Microsoft would not be
33:18 rolling all of all of this Tech into all
33:21 of their products if if it was going to
33:24 be you know yanked if it was going to be
33:26 rug pulled
33:30 runway. TV first aitv interesting Sherry
33:34 D let's go look at
33:38 that is that from Runway
33:42 [Applause]
33:45 mL no sorry can't bring anyone up on
33:50 screen it is from
33:55 Runway hey I'm
33:58 [Music]
34:03 [Music]
34:10 talking
34:13 hi
34:17 hi
34:18 [Music]
34:26 hi
34:28 [Applause]
34:30 [Music]
34:35 so I'm going to end up getting in
34:36 trouble for this because I don't I don't
34:38 know what it's going to do it's getting
34:39 kind of creepy it's getting kind of
34:41 creepy people um how do I oh here we
34:48 go boy this has got real
34:51 MTV early days of MTV Vibes to it
34:55 doesn't
34:56 it so this is Runway ml so if you don't
34:59 know Runway ml they're a text to
35:01 video technology
35:05 Channel I'll tell you what if you're a
35:07 filmmaker or even if you're not I would
35:10 start experimenting with making
35:13 shows using their Tech because this is
35:17 going to become this is going to become
35:20 a film studio or a TV this is going to
35:22 be a new Netflix kind of thing this is
35:24 really
35:25 interesting AI film Festival gen
35:29 48 creative partner
35:33 program Eerie there's spooky things
35:40 [Music]
35:45 sci-fi hey Donovan I'm getting a call
35:48 from the Dominion building you might
35:49 want to go check this one out I'm on
35:53 [Music]
35:56 it
36:05 so okay
36:11 so wow this is ambitious like this
36:15 is this is trippy right so we just have
36:18 the
36:19 SAG and the the writer Guild strikes
36:23 right and they basically are like don't
36:25 have ai replace us
36:28 and basically the contract that they did
36:32 that with was with the producers right
36:34 the film producers the big the big
36:38 producers well runway's not a big
36:41 producer and none of the none of the
36:43 video that you see in here is actually
36:46 Hollywood quality like the cars the cars
36:49 looked like they were sort of sliding
36:51 sideways across the street but a year
36:54 from
36:55 now these things are going to be good
36:57 and and the the the acting voices right
36:59 here sound really stilted and these like
37:03 this shows 2 minutes 3 minutes
37:07 long what they got me chasing this
37:11 time probably just another
37:15 cat stuck in the ceiling tiles see the
37:18 weird like his legs he's just sort of
37:21 pouting on one
37:23 leg but this is a big
37:26 deal let's see music videos oh those
37:29 will be better let's do some music
37:33 videos I'm going to do a video on this
37:35 one this is this is a big
37:37 deal thank thank you for sharing this
37:40 whoever shared that oh it was you Sherry
37:42 thank you
37:56 Sherry
37:58 [Music]
38:00 so that looks like real
38:03 video sponging all the [ __ ] be by
38:07 this city it became hard to hear from my
38:11 heart yeah they're they're meling when I
38:13 realized it wasn't too late yet by the
38:17 time that I it
38:20 was
38:22 oh the tears that I cried you were
38:26 looking at it like a
38:27 on
38:28 Mars huh that's like some sort of hybrid
38:31 production cool I
38:34 think but the fact that I don't know is
38:37 actually
38:39 [Music]
38:42 [Applause]
38:46 interesting
38:48 huh endless Reflections I assume this is
38:53 AI dreams of
38:56 dogfish
38:59 this film was generated by AI using only
39:02 lines of poetry from Mary Oliver's
39:06 Dogfish oh see this is brilliant so they
39:09 took an existing
39:11 poem and visualized it with AI and
39:14 turned it into a video music
39:16 video or a vi video
39:20 video culture 24
39:23 storytime television Alice in Wonderland
39:30 I had a my
39:32 own I've seen this
39:38 [Music]
39:42 one amidst dreams and codes I stand at a
39:46 Crossroads the rabbit hole
39:48 [Music]
39:51 beckons my father's creation his legacy
39:56 a realm we call
40:00 Wonderland a simulated
40:03 reality a
40:05 game
40:07 wow end of programming Runway ml.com
40:13 holy
40:15 crap I mixed stock footage with Runway
40:18 AI Clips yeah yeah exactly that's
40:21 perfect wow wow people all right so
40:28 filmm and TV people are going to look at
40:30 this and they're going to laugh at it
40:32 and they're going to make fun of it and
40:34 then a year and a half or two from now
40:37 or maybe not even a year from
40:40 now the quality of these
40:43 things will be indistinguishable from
40:47 other
40:48 stuff they're not going to run into any
40:50 of the issues with sag and the writer
40:53 Guild because they're going to be
40:55 produced in a completely
40:57 organic different
41:01 way and it's going to be this is this
41:04 could end up being this could end up
41:06 being a new kind of Studio I got to fly
41:09 around here
41:11 that's winter flies are weird cuz they
41:14 fly so slow they're
41:18 like wow wow that's CRA here's a brief
41:22 presentation that uses something
41:25 YouTube has some with no plot but wild
41:29 image morphing interesting this is
41:31 runway. TV yeah this is runway. TV and
41:34 it is from Runway ml right so this is
41:38 you know Runway ml is the text to video
41:40 tool and you know they just they they're
41:43 if you look at their tool where their
41:45 tool has been going is giving camera
41:48 control to directors right so there's
41:50 this motion brush thing where you can
41:52 say let me just make his head move and I
41:54 want I want the rest of the scene to say
41:56 stay static I just want his head to move
41:59 and then you can do things like zoom in
42:01 pull out pan left pan right all that
42:03 sort of stuff determine how much
42:05 movement's there
42:08 um wow
42:12 wow and as as the diffusion models get
42:15 better
42:16 at coherence across different
42:20 scenes you won't have that weird thing
42:22 where the main character sort of changes
42:24 looks in between scenes
42:28 um
42:30 crazy Sherry D they have a ton of cool
42:33 features just getting my head around it
42:35 yeah Drew learning AI Drew your your
42:37 Channel right now so so what I would say
42:39 to this anyone on the channel anyone on
42:42 the live right now that's doing anything
42:45 kind of interesting with editing
42:47 together Runway clips and putting
42:50 together projects like I would get
42:53 instantly involved in the runway Discord
42:57 and start Hawking your stuff up there
43:00 because I bet they're cutting deals with
43:01 makers um and Drew you've you've been
43:04 doing some stuff that's really simple
43:06 you know using simple animations and
43:08 things like that but I I can only
43:10 imagine the kind of stuff you're working
43:11 on I would I would dig deep into this
43:14 let's see Q learning and reinforcement
43:16 learning Q learning is a model model
43:19 free reinforcement learning algorithm
43:21 yeah yeah that's what I thought it looks
43:23 like this Q learning stuff sits on top
43:25 of models
43:29 wild wild wild wild
43:37 people very cool that is very cool isn't
43:39 it that one is cool is Runway a
43:42 membership Runway yeah Runway is I think
43:46 20 bucks a month for not a lot of video
43:49 so it's not cheap to do this stuff but
43:52 you're generating
43:54 again start if if if this is something
43:57 you're interested in if
44:01 listen not for nothing
44:03 people you don't have to go to film
44:05 school to be a film person
44:09 anymore now film people will come smack
44:13 me upside the
44:15 head cuz they're like but it's a craft
44:17 and you got to use the camera and you
44:18 got to use this and you got to use that
44:20 well these tools right now are janky
44:23 pieces of [ __ ] and existing film makers
44:28 are going to laugh at them but if you
44:29 learn these
44:30 tools and start putting things together
44:34 Mr K my students can write scripts and
44:36 see if the video matches their
44:37 imaginations totally Mr K in fact Mr K
44:40 you know what you know what would be
44:41 really cool for you to
44:43 do is have your students
44:48 um put together they could put together
44:52 a script where they each contribute
44:53 little pieces of the script like a game
44:55 of telephone or something like that have
44:58 them all do little sections of video and
45:00 then stitch them all together into
45:02 something and submit it to Runway ml
45:04 maybe you could get your class to get
45:06 get a a show accepted into Runway TV CU
45:10 I'm telling you I'm I whether whether
45:12 it's these guys or someone
45:14 else these are going to be the the
45:17 Netflix of the future or the Comcast of
45:19 the future or the Paramount Pictures of
45:21 the future are going to are going to be
45:24 something like this where
45:26 a bunch of people that are initially
45:29 laughed at and marginalized oh that's
45:31 not real video that's not real film
45:34 making all of a sudden these tools are
45:36 going to get so good that this is this
45:37 is going to become a new kind of art
45:40 form and man that would be a cool
45:42 project because they could learn they
45:44 could learn sound design they could
45:46 learn writing they could learn um you
45:50 know editing putting these things
45:52 together prompting oh my God that that
45:54 would be such a cool project
45:58 look look I'm I'm changing teachers
46:01 curriculums screw teaching them what
46:03 you're supposed to make have them make a
46:10 movie oh man
46:13 runway.
46:16 TV says the connection is not secure
46:19 should I care no but make sure you're at
46:22 the right thing it is runway. TV so yeah
46:28 they probably just don't have their I
46:30 this was probably thrown together
46:32 quickly these are this this is basically
46:34 just a front end uh sitting in front of
46:36 a bunch of YouTube
46:37 videos so they probably yeah what
46:40 they're probably doing is they're
46:41 probably just they're probably trolling
46:43 their Discord and they're trolling
46:45 Twitter and anytime any anyone does
46:48 anything cool they're probably like hey
46:50 can we put this on our YouTube channel
46:52 and now they just put a this is probably
46:53 a WordPress this is probably a word
46:56 press like you know video
46:59 template um and these are just a bunch
47:02 of bunch of things submit film look here
47:05 you go submit
47:08 film Runway is a dedicated platform for
47:11 streaming generated videos all in one
47:13 place the content is a mix of original
47:14 programming from Runway Studios Runway
47:18 Studios remember what I said this this
47:20 could be the new Paramount
47:22 Pictures and yeah the the yes this is
47:25 going to be this going to be big cuz
47:27 what they're going to get so here's
47:28 what's going to
47:30 happen there's going to be three kind of
47:32 filmmakers okay there's going to be
47:35 existing filmmakers that poo poo all
47:37 this [ __ ] and they're like [ __ ] ai ai
47:40 sucks you're all
47:42 [ __ ] and then you're going to have
47:45 non- filmmakers that just do this and
47:48 they don't know anything about film
47:49 making but they're going to learn how to
47:51 do this and then you're GNA have
47:55 filmmakers that embrac this and I have a
47:59 feeling filmmakers that Embrace this are
48:01 going to do crazy [ __ ] and also people
48:04 that don't know film making that make
48:06 this kind of stuff are going to do crazy
48:08 [ __ ] because they don't know the rules
48:10 so they're going to do all sorts of
48:11 crazy ass
48:13 stuff this is wild dedicated platform
48:16 allinone place it's a okay it's an a a
48:20 mix of original programming from Runway
48:22 Studios and a curation of community
48:25 created films
48:26 if you'd like your work showcased please
48:29 fill out the form below they must be at
48:31 least 30 seconds long they can be any
48:33 aspect Rao they must be able to stand
48:36 alone on their own if you're submitting
48:37 a work that was done for something else
48:39 like a music video it's not good it's
48:41 got to be on its own please only submit
48:43 any given film
48:45 once please make sure you use a Runway
48:48 asset link we cannot accept links to
48:52 social media or other video uploaded
48:54 platforms please use use a valid email
48:56 address and this is basically a Google
48:58 [Laughter]
49:02 form I get get the if you're interested
49:05 in film making at all I would jump on
49:08 this jump on it like is it going to be
49:12 anything I don't know but if nothing
49:15 else you could point to this and go hey
49:17 I'm on a
49:19 [Laughter]
49:21 network all those Gen X liberal arts
49:23 degrees might make us money after all
49:25 tell our parents
49:26 you can make
49:27 money Source Camp I'm telling you this
49:30 is Revenge Revenge of the Gen xers and
49:32 revenge of the liberal arts Majors is
49:35 the is the era we're entering right now
49:38 liberal arts Majors I'm telling
49:42 you because you can look across all the
49:45 disciplines right anyone who specializes
49:48 in a
49:49 discipline they're going to get
49:50 blindsided by
49:54 generalists cuz what was the the the
49:56 knock on generalist for years was you're
49:59 a generalist you can't go deep on
50:01 anything well now you
50:05 can so now you're a generalist that can
50:07 go deep in 10 areas that you're
50:10 interested in holy
50:12 [ __ ] I ditch them all during the strike
50:15 going to take months before content
50:17 Returns what did you ditch
50:19 before the
50:22 strike that looks like an interesting
50:25 story right Runway TV would not let me
50:27 in woohooo WooHoo for my degree in
50:30 religion I'm telling you
50:32 man anyone who can think critically
50:35 right now anyone who can think
50:37 critically and
50:40 and synthesize different ideas from
50:43 different
50:44 disciplines you're you're you're going
50:46 to you're going to be able to write your
50:47 own ticket 2024 is going to be ABS you
50:51 think 2023 was [ __ ]
50:54 crazy I've got a [ __ ] Tik Tok
50:57 Channel and I don't I think we have
51:00 barely scratched the surface I think we
51:02 don't have any idea what's
51:05 coming it's going to come so fast do you
51:08 know how quickly this this Runway TV
51:10 thing is going to get
51:17 good yeah I'm this is going to be the
51:19 next video I do is going to be on this
51:21 I'm going to do I think I'm going to do
51:22 a long one on
51:24 this
51:26 huh you know
51:28 what no not
51:35 there where's
51:41 Runway that's mid Journey that's
51:44 neotokyo there it is Runway
51:51 announcements we are excited to announce
51:54 the launch of Runway TV today at 7:40
51:58 a.m. what is one short thing I should
52:01 show my
52:02 colleagues to give them a Kevin
52:04 McAllister moment okay so for those of
52:08 you that don't know one of the things I
52:09 talk about here all the time is use AI
52:12 long enough to have your Kevin
52:13 McAllister moment which is the Home
52:16 Alone
52:17 moment um and and uh we just celebrated
52:20 like a week ago the 33rd anniversary of
52:23 home alone if you can believe it or not
52:25 we're old people
52:31 um the the the thing the thing that I
52:34 would do
52:37 to to to give them their Kevin mallister
52:40 moment
52:42 is try to figure out the thing they're
52:44 most interested in and show them
52:46 something that's relevant to them like
52:49 if you've got a bunch of project
52:50 managers in the room and you show them
52:52 some creative writing [ __ ] they're going
52:54 to be like yeah okay whatever ever that
52:56 doesn't that doesn't do anything for me
52:59 but if
53:00 you have them give you some input about
53:03 a project plan and then it just whips
53:05 out a project
53:08 plan yeah I know Silver Fox I'm the same
53:11 way Silver Fox just said I wish they'd
53:13 pick a different letter the here's the
53:15 Saving Grace Silver Fox is they're
53:17 calling it qar not q and on or not just
53:20 Q
53:22 qar um synthesizing info and parsing it
53:25 for others is what I do yeah oh yeah
53:28 that yeah that's what you do that's
53:29 that's going to be the the the big skill
53:31 but but if you want to blow someone's
53:33 mind I think you just you have to get a
53:36 thing they're interested in and it could
53:38 even be just one person in the room
53:40 right and then and do that in front of
53:42 other people right so maybe you pull
53:45 someone out of the audience or if it's
53:47 like a lunch and learn thing you just
53:48 you just pick someone and you say hey
53:51 tell me something you're interested and
53:52 in real time kind of like what I do here
53:54 where someone says oh can you do you
53:57 know what could what could chat GPT do
53:59 for the fashion industry that that night
54:01 I did that thing where I ended up
54:02 creating a whole fashion design in like
54:05 20 minutes or 10 minutes or whatever it
54:08 was something like that um but it's got
54:11 to be relevant to them I think cuz you
54:14 can't just
54:15 it's I don't know just looking at this
54:18 [ __ ] from the outside it does not look
54:20 impressive it just looks like a Google
54:22 search but that the minute you do
54:24 something that you know about and you
54:26 know how hard it is to do and then it
54:29 does it in [ __ ] 4 seconds and you're
54:31 like oh my
54:34 God that's what that moment
54:37 is how does this compare to steve. a I
54:40 don't I don't know I don't let's see
54:43 what's Steve Steve come on
54:46 Steve steve.
54:50 a the only AI tool to create live videos
54:55 using
54:59 text oh I think this is kind of
55:05 like this this is more like corporate
55:07 videos so so how Runway TV is different
55:11 is Runway is doing true text
55:14 to video these guys might be but it
55:17 looks
55:19 like yeah what this does this is like
55:22 well what's that one called that I
55:24 recommend
55:27 Visa there's this one Visa that does
55:29 this steve. TV is is kind of like Visa
55:34 where you give it text and it picks uh
55:37 um stock video clips and edits them
55:41 together but Runway TV is
55:45 like it's a it's it's a it's a network
55:49 it's
55:50 Netflix now the shows on it are all
55:53 crappy right now but but you're going to
55:56 get really good filmmakers that start
55:58 putting really good music videos
56:00 together really good films together that
56:02 are incredibly in
56:04 inventive dang it I was in denial about
56:07 the old thing which old thing the old
56:09 film thing listen the old film thing
56:11 still going to exist but this new film
56:14 thing is going to this is going to get
56:16 good so fast it's it's this is going to
56:19 this is going to shock people and and
56:22 not necessarily Runway television just
56:24 this idea
56:26 that a tool company Runway ml that does
56:30 this text to video tool now has Runway
56:33 Studios so they're probably hiring
56:35 filmmakers to use their tool to make
56:37 films that they're going to
56:38 fund and I bet they can make a lot more
56:41 films for a lot less
56:44 money and then and now they're Community
56:46 sourcing so what they're doing is
56:48 they're saying hey anyone out in the
56:50 community who's got any Talent at this
56:52 will feature your films and featuring
56:55 your films probably within 6 months is
56:58 going to start turning into
57:00 contracts they're having a contest right
57:02 now too might want to look into that
57:04 yeah that's really cool um in fact you
57:07 know one of the things I'm going to
57:08 do I'm going to copy this and see if
57:14 it's um Scott over in the AI Salon just
57:19 started the video
57:21 Guild so I'm going to go share this in
57:24 the video Guild and then I'm going to
57:25 tell you
57:27 about what this
57:33 is Scott's video
57:39 [Music]
57:41 Guild feed events members okay oh okay
57:45 you put it in there okay so it's already
57:49 in there great I'm going to hard it and
57:51 then I'm going to say
57:54 um this
57:56 is a
57:57 BFD
57:59 um anyone in the
58:03 salon um
58:05 who's interested in
58:10 video you
58:12 should
58:15 submit on top of that maybe the video
58:21 Guild should do a project
58:26 and give the
58:29 salon some
58:32 visibility yay um that's super cool and
58:36 then I'm also going to go look at
58:38 something here because there was a music
58:40 section and Scott does
58:47 um how do I get to the bottom TV Guide
58:50 there we
58:51 go one of these channels is music music
58:56 music
58:58 videos oh they don't have their own
59:01 URLs do
59:05 they
59:09 [Music]
59:15 no that's super
59:19 cool
59:21 um that's
59:24 got
59:27 hang on I'll be with y'all
59:29 people um you should
59:35 submit your
59:39 videos
59:41 your music videos
59:54 to
1:00:01 cool all right that's super cool
1:00:04 people all
1:00:05 right I'm here apparently to talk about
1:00:08 Q hey Pate how's it going um
1:00:12 so I would love for you
1:00:16 to um if you want to chat I'm happy to
1:00:20 bring you up but
1:00:24 so
1:00:26 so
1:00:28 apparently shortly before they fired Sam
1:00:31 Alman there was a letter sent and there
1:00:33 was this thing called
1:00:34 qar um that that made a breakthrough in
1:00:38 reasoning with a with a small large
1:00:40 language model um and so I went to chat
1:00:44 GPT and basically um it told me about
1:00:49 sarsa
1:00:51 the uh State action reward State action
1:00:55 model um which I assume sits on top of
1:01:00 any model like I I assume this is like a
1:01:02 reasoning
1:01:03 algorithm um and then I had cat gbt
1:01:05 explain it and then basically I what it
1:01:07 explained was you have state one and
1:01:10 state the the tool can take an action
1:01:13 based on the state and then the result
1:01:17 of that action gets graded basically
1:01:19 good or bad and so Q gets rewarded or
1:01:22 penalized based on how it did then it
1:01:24 goes into a new state takes another
1:01:26 action gets gets rewarded or
1:01:29 penalized uh and it goes round and round
1:01:32 it basically just you know keeps
1:01:34 improving and doing that the letter has
1:01:38 nothing too interesting in it it
1:01:41 maybe kind of does math okay for a fifth
1:01:45 grader yeah that's what they said they
1:01:47 said they said right now it's doing it
1:01:49 it's it's a really small a really small
1:01:52 model doing small stuff but it's doing
1:01:54 reason reasoning it's doing reasoning
1:01:57 right which is the the first step and
1:02:00 what they're saying is that if you take
1:02:03 this model and scale it then you have
1:02:05 you have
1:02:07 something Pate is a Genie in a Bottle
1:02:09 say his name and he appears that was
1:02:12 good oh man okay video Guild let's get
1:02:16 on this I know Source Camp right I I'm
1:02:19 telling you you can't extrapolate that
1:02:21 oh all right see this is why we have
1:02:24 Pate here people
1:02:25 people Pate knows actual math he he
1:02:29 knows the
1:02:30 maths why do they say maths isn't math
1:02:34 already
1:02:36 plural no Pate why do they say maths
1:02:40 let's get to the important thing I don't
1:02:42 care about the symbols and the equations
1:02:44 I want to know why people talk about
1:02:47 maths instead is that just a British
1:02:52 thing paid at some point it's just going
1:02:55 to we're I'm going to meet him sometime
1:02:57 he's just going to slap me say you're
1:02:58 you're a
1:02:59 douche um wish run Ray had a site map
1:03:03 I'm on Runway but no Runway
1:03:06 link oh yeah well yeah I
1:03:09 know all these companies are running
1:03:11 fast and loose right now I'm telling you
1:03:14 that that Runway TV is a big it's going
1:03:17 to be a big deal and I bet we're going
1:03:19 to start seeing a bunch of these things
1:03:22 and in fact you know what people
1:03:27 techie Becky's been in here before I
1:03:29 think she said she was a talent
1:03:31 agent
1:03:34 um you could start you could start a
1:03:37 network right now by just cut you know
1:03:40 Finding talented people and you know
1:03:44 doing Distribution on whatever on
1:03:47 YouTube or on some other channel some
1:03:49 Channel you
1:03:50 do mathematics not mathematic Math's not
1:03:54 math
1:03:56 all right Vicky you with your logic
1:03:59 always applying logic to the
1:04:02 situation applying logic to the
1:04:07 situation
1:04:14 [Music]
1:04:19 fine mathematics oh all of you now just
1:04:22 slapping me down so please let it be
1:04:25 known I am not an expert especially on
1:04:28 reinforcement models I could be very
1:04:32 wrong well if if whatever the
1:04:36 Breakthrough they made was significant
1:04:39 enough that it it created an urgency
1:04:41 that got Sam Alman fired my assumption
1:04:45 is that that they know how to scale it I
1:04:47 I would assume that that diagram we just
1:04:50 saw is a a dramatically oversimplified
1:04:53 version of it so um
1:04:56 so I don't know I mean listen both
1:04:59 Reuters and the information
1:05:03 um have reported now that that a letter
1:05:07 was sent and a a model that qar model um
1:05:11 is a breakthrough at open Ai and that's
1:05:14 it's it's a big deal if they've hit
1:05:17 AGI number Theory applied math math Ed
1:05:22 wait maths
1:05:26 listen brother 52 you don't need to rub
1:05:29 it in they already told me I was wrong
1:05:31 okay this
1:05:33 G
1:05:36 crazy I asked chat what the runway TV
1:05:39 was all about and and got a good
1:05:41 synopsis oh that's
1:05:46 cool I'm a
1:05:48 Brit yeah that's good I I
1:05:51 know but the Brits always say
1:05:55 maths it always it it always trips me up
1:05:58 a letter was sent they could not read
1:06:00 the letter no com no comments were
1:06:03 provided yeah I know I listen Pate I I
1:06:09 the first night you asked this like is
1:06:10 there any actual reporting going on here
1:06:13 on the AI learning lab no this is chat D
1:06:15 TMZ this is all speculation all the time
1:06:18 I know nothing I am just reporting what
1:06:20 I read on
1:06:23 Twitter
1:06:28 they have the juice now yeah it sounds
1:06:31 it sounds like we're going to see some
1:06:33 [ __ ] can you explain AGI well AGI I can
1:06:38 I can tell you broadly what it is and
1:06:41 let me let me go show you a couple of
1:06:43 things we're going to go look at the
1:06:45 reiders
1:06:48 piece so what's important right now in
1:06:52 the context of open AI is open ai's
1:06:55 definition of it not
1:06:58 mine so open
1:07:03 AI defines
1:07:06 AGI artificial general intelligence as
1:07:10 autonomous systems that surpass humans
1:07:13 in most economically V valuable
1:07:17 tasks um so so the important words here
1:07:22 are autonomous so so check GPT right now
1:07:25 is not autonomous you log into it you
1:07:29 give it a prompt it gives you an answer
1:07:31 now there are exceptions to
1:07:33 that code
1:07:35 interpreter is a little hint at
1:07:39 autonomous agents cuz you go to code
1:07:42 interpreter and you say hey um here's a
1:07:45 CSV file tell me what dat is in it
1:07:49 what's interesting about it and help me
1:07:51 visualize it and then you sit back and
1:07:53 you watch [ __ ] run through a series of
1:07:56 tasks where it prompts itself and writes
1:07:59 python code and executes it and gives
1:08:01 itself answers and then looks at the
1:08:03 answers and goes no I don't like those
1:08:04 let me try that again oh no I don't like
1:08:07 oh yeah here we go oh here's some
1:08:08 interesting stuff oh let me try to
1:08:10 visualize that now and it visualizes it
1:08:11 for you so it will take seven or eight
1:08:14 steps with your one goal that's an
1:08:18 autonomous agent it's a really
1:08:21 simplified version of it so what they're
1:08:23 talking about is autonomous systems
1:08:25 where you can give them a goal and they
1:08:27 just
1:08:28 go that surpass humans so better than
1:08:33 humans and and this is the thing in most
1:08:36 economically valuable
1:08:41 tasks holy [ __ ] so what people get paid
1:08:45 to
1:08:45 do they're defining AGI as a system that
1:08:50 is better than most humans at
1:08:52 economically valuable tasks
1:08:55 so being a lawyer being an accountant
1:08:57 being a an analyst being a programmer
1:09:05 being all of
1:09:07 it all all knowledge work and then when
1:09:12 these systems get built into humanoid
1:09:16 robots and
1:09:18 can go build
1:09:20 [ __ ] then it extends Beyond knowledge
1:09:23 work
1:09:25 so so that's why it's a big deal the
1:09:27 other reason it's a big deal is
1:09:30 when in in chat gpt's in open ai's
1:09:37 Charter when they declare they've hit
1:09:41 AGI licensing for the AGI technology
1:09:45 is separate from non AGI technology and
1:09:50 so Microsoft right now has access to all
1:09:53 of
1:09:55 gp4 if whatever they've put together
1:09:58 here if it's qar plus gp5 equals AGI
1:10:02 that
1:10:03 technology is not part of the Microsoft
1:10:06 deal that goes back to the nonprofit to
1:10:09 benefit Humanity fired Sam Alman do not
1:10:13 have deep business experience they've
1:10:15 got deep research experience and and
1:10:18 it's it's the the imperatives of the of
1:10:22 the nonprofit came into conflict with
1:10:24 the imperatives of the for-profit
1:10:27 Apparently and that so that's why and so
1:10:30 if they hit AGI it it changes a lot of
1:10:33 things for for open Ai and for for all
1:10:36 of us
1:10:37 really so cuckoo cucko what was the name
1:10:41 of the open a model that may have made a
1:10:43 bro breakthrough QPR no it's qar so it's
1:10:46 just Q
1:10:48 asterisk here right
1:10:51 there
1:10:53 qar
1:10:55 which is ominous sounding but if you
1:10:58 look at the
1:11:00 diagram it looks like a star of Q's see
1:11:04 the
1:11:05 q's q q q q
1:11:12 oh it was AR racus what was AR racus I
1:11:15 thought they I thought they shut down AR
1:11:17 racus because it wasn't that wasn't
1:11:19 working I thought that was a different
1:11:20 large language model or
1:11:23 something sounds like qar is different
1:11:25 than
1:11:26 oracus listen at the current time all
1:11:30 this at the current time all of this is
1:11:33 glorified toys I absolutely agree the
1:11:36 real thing will cost serious money uh I
1:11:39 don't these things are toys right now
1:11:42 but I this is fairly serious
1:11:46 capabilities for 20 bucks a
1:11:49 month so there are no knowledge workers
1:11:52 there is no need for for much of
1:11:54 Microsoft I know well this is the thing
1:11:58 Microsoft right now is doing
1:12:01 what big companies almost never do they
1:12:06 are absolutely playing with fire cuno
1:12:09 they are absolutely playing with fire
1:12:11 this is why Kodak went out of business
1:12:14 is the the team that invented the
1:12:17 digital camera walked into the boardroom
1:12:19 and said hey we've got this thing called
1:12:20 a digital camera and the boardroom said
1:12:23 that threatens what our business is our
1:12:25 business is selling photochemicals and
1:12:28 paper get out of here with that stupid
1:12:31 low quality piece of janky ass [ __ ] Pate
1:12:34 I don't see why it has to be
1:12:36 economically
1:12:37 valuable they seem very arbitrary oh
1:12:41 that seems very
1:12:45 arbitrary yeah I would I would agree
1:12:47 with that I would agree with that Pate I
1:12:49 don't know why they've that's how they
1:12:51 described
1:12:52 it
1:12:54 um but I think you're right I mean
1:12:58 it's I think you could I think you could
1:13:01 remove the the economically out of that
1:13:04 statement and it and it would still it
1:13:06 would still track as significant right
1:13:10 open AI defines AGI as autonomous
1:13:12 systems that surpass humans in most
1:13:15 valuable tasks right you could kill you
1:13:18 could kill economically
1:13:22 there because if you A system that ex
1:13:25 that surpasses humans in most valuable
1:13:28 tasks tasks that provide value value
1:13:31 doesn't have to be economic then it's
1:13:34 it's a big [ __ ]
1:13:35 deal but I
1:13:38 agree totally
1:13:42 agree how long till there's a GPT where
1:13:44 you upload your ctmri X-ray and it
1:13:48 analyzes that for
1:13:49 free um Joker Joker in the AI Salon I
1:13:53 should I should talk about the AI Salon
1:13:55 Joker in the AI Salon is doing a thing
1:13:57 right now I think using chat GPT Vision
1:14:01 um where he's doing some sort of uh M
1:14:04 microscope micros micros
1:14:08 microscopy I don't know how to say that
1:14:10 word microscope slides slides of small
1:14:13 [ __ ] and he's doing analysis of that
1:14:16 right now so I I don't think it'll be
1:14:18 that
1:14:19 long
1:14:22 um I wonder how many commit hours it
1:14:24 took to come up with that
1:14:26 definition well I mean open ai's been
1:14:28 around for a while they've this is this
1:14:31 has been this has been years in the
1:14:34 making I think their first gpt1 came out
1:14:37 when in 2017 no 2018 the Transformer
1:14:41 model was invented by Google in 2017
1:14:43 gpt1 came out in
1:14:45 2018 gpt2 came out in oh no gpt2 came
1:14:50 out in 2018 gpt3 came out in 201 20 chat
1:14:55 GPT came out in 2022 GPT 4 came out in
1:15:01 2023 and
1:15:03 qar hit the
1:15:06 scene you know in uh in late
1:15:12 2023 prediction new open AI board
1:15:14 changes the charter possibly the the new
1:15:18 open AI board is not just going to be
1:15:20 three people there's three people
1:15:22 announced right now I would assume that
1:15:24 Alman has a seat on it right right now
1:15:28 it's
1:15:30 um
1:15:40 hold did I update this one here
1:15:44 yeah so the new
1:15:50 board is Brett Taylor
1:15:54 who's from Salesforce he was the CEO of
1:15:57 Salesforce Larry Summers who I forget
1:16:00 what he was but he's like a like an old
1:16:02 school come in and fix the [ __ ] up and
1:16:05 then Adam
1:16:06 D'Angelo who's the quar CEO so he was on
1:16:10 the board already and there was for for
1:16:13 two days Ilia was the bad guy then for a
1:16:16 day and a half or two Adam was the bad
1:16:18 guy and now it looks like it's this
1:16:20 woman Helen that's the bad woman
1:16:25 and she's no longer there along with the
1:16:27 other female board
1:16:29 member um but this is going to be I I I
1:16:32 assume this is going to be a way bigger
1:16:34 board than three people but these are
1:16:35 the three
1:16:36 initials um I would think that Microsoft
1:16:39 at a minimum gets a an observer seat I
1:16:43 would think that Sam Alman would have a
1:16:46 a director role but who
1:16:51 knows I want to know how we're going to
1:16:53 to live without income crime is just
1:16:55 going to Skyrocket well so so Silver Fox
1:16:59 this I mean this
1:17:00 is this is one of this is one of the
1:17:03 reasons if if you if you listen to Sam
1:17:05 Alman and why he created worldcoin and
1:17:07 why this is a kept profit model for open
1:17:12 AI what his prediction is and actually
1:17:15 David Shapiro's predictions are in a
1:17:17 similar state is that there's going to
1:17:19 be such an increase in
1:17:21 prosperity um that that you're going to
1:17:24 have to funnel that back to people
1:17:26 because there's going to be an increase
1:17:27 in prosperity with a decrease in the
1:17:29 need for people to do the work to
1:17:32 generate that
1:17:33 Prosperity
1:17:35 um the private Equity firms are just
1:17:39 going to drop that profit to the bottom
1:17:41 line right and but but there's a point
1:17:44 at which that that [ __ ] crushes
1:17:48 Society so government's going to have to
1:17:50 step in or you know
1:17:54 companies like open AI that are
1:17:56 generating that Prosperity are going to
1:17:58 take the excess prosperity and funnel
1:18:01 that back to the people and that's what
1:18:03 worldcoin is that's one of the reasons I
1:18:05 went and I got my world ID where I
1:18:07 scanned my eye in that stupid [ __ ]
1:18:09 orb thing was I just want to be part of
1:18:11 the experiment because I want to see
1:18:13 what comes and if it and if it's coming
1:18:15 as fast as you know within a year or two
1:18:18 or three um even on any scale um you
1:18:23 know PE PE they're gonna have to do
1:18:25 something someone's going to have to do
1:18:27 something so we'll
1:18:30 see
1:18:32 um the the other the other thing Silver
1:18:35 Fox
1:18:37 is there's going to be entirely new
1:18:40 kinds of jobs created that we can't even
1:18:42 anticipate right now I don't know what
1:18:44 those are because we can't anticipate it
1:18:48 but if everyone's if if everyone gets to
1:18:51 level up their skills because these
1:18:52 tools get so good that I can do all
1:18:54 sorts of good stuff there are going to
1:18:56 be interesting things people can do and
1:18:58 create and businesses they can make that
1:19:01 we can't even imagine right now so I
1:19:04 don't I think I think one of the fears
1:19:06 is people lose their jobs and they just
1:19:08 sit around like well now I can't work
1:19:10 ever but I don't think it's going to be
1:19:12 like that I think people are going to
1:19:13 use these tools to level up transform
1:19:16 themselves rein invent themselves which
1:19:18 is it's one of the reasons I'm so
1:19:19 excited about this channel in the in the
1:19:21 past two months is a a lot of you a lot
1:19:24 of the Irregulars that are in here are
1:19:26 finding ways to reinvent yourselves and
1:19:28 it's and it's like it's a palpable
1:19:31 energy that I
1:19:35 see
1:19:40 uh I don't understand why the core CEO
1:19:42 is still there seems like a massive
1:19:44 conflict of interest I agree I agree I
1:19:48 agree wait World ID what all right let
1:19:51 me show you my world ID
1:19:54 and I'll tell you all about it it's so
1:19:56 cool people now people that um do I have
1:20:00 it here yeah I think I do
1:20:20 [Music]
1:20:22 um
1:20:28 that was a long time ago a long long
1:20:30 time ago I can still
1:20:39 remember did you write the Book of Love
1:20:42 here it is
1:20:44 Boom
1:20:46 look I got a world
1:20:48 ID and look verify at
1:20:51 orb just to make get totally creepy so
1:20:54 what this
1:20:56 is is is an
1:21:00 anonymous ver verification of my
1:21:04 uniqueness as a human in this system so
1:21:08 what the orb does is it takes a picture
1:21:10 of your
1:21:11 retina it compares the signature that it
1:21:16 created with anyone else in the database
1:21:18 and says is that retina unique to the
1:21:22 other retinas that scan and if it says
1:21:24 yes then it generates this which is an
1:21:26 nft it's a it's a blockchain
1:21:29 verified um ID that me as the
1:21:33 holder um gets to say I'm I'm human
1:21:39 now there's three components to to the
1:21:42 platform so this is Sam alman's company
1:21:45 there's three components to this there's
1:21:47 World ID which is this there's World
1:21:50 coin and then there's a payment system
1:21:55 so with the payment system you're going
1:21:58 to be able to create um so so payment
1:22:01 system is like Stripe Right or or like
1:22:04 PayPal but you're going to be able to um
1:22:08 get paid in and buy things with
1:22:10 worldcoin and you're going to get
1:22:13 worldcoin if you own one of these and
1:22:17 they take excess profits from open AI
1:22:20 funnel them back to the nonprofit and
1:22:22 then nonprofit distribute some of those
1:22:25 profits out to this
1:22:27 system so what they're doing right now
1:22:29 is they're trying to get developers to
1:22:30 build the the version of Facebook that's
1:22:34 based on this platform and the version
1:22:36 of Twitter that's based on this platform
1:22:38 and the version of Amazon and the
1:22:40 version of eBay and the version of etsy
1:22:42 and
1:22:43 the we got to stop
1:22:45 treating these AI companies and work as
1:22:48 magic the more people understand the
1:22:51 better yep absolutely absolutely I agree
1:22:53 and if you think I'm treating them like
1:22:55 magic let me know um I try to explain
1:22:59 how all this [ __ ] Works um I don't think
1:23:02 this is Magic at all I think this is
1:23:04 this is at a minimum at least worldcoin
1:23:08 is an attempt to deal with the wealth
1:23:12 Gap that's going to be created the
1:23:14 prosperity Gap that's going to be
1:23:16 created
1:23:18 um so so anyway so couple other things
1:23:21 that this does because it's a verified
1:23:24 human thing if you build a publishing
1:23:28 system on top of
1:23:30 this you could have a verified human
1:23:34 wrote an article that you could also
1:23:36 verify on the blockchain and you could
1:23:38 say not only is it verified that this
1:23:40 human wrote that article but I can give
1:23:43 you a little green check mark that says
1:23:45 that article has not been [ __ ]
1:23:47 with right so I can start to you know if
1:23:51 Cara Swisher the the reporter in Silicon
1:23:54 Valley that's been doing doing a lot of
1:23:55 scoops on the open AI stuff if she had
1:23:57 one of these and there was a publishing
1:23:59 system on top of it you could say okay I
1:24:01 know I can blockchain verify that Cara
1:24:04 Swisher is the person and she wrote that
1:24:07 article and that article hasn't been
1:24:09 modified so I think this becomes a
1:24:11 foundation for authenticity of content
1:24:14 as well um so anyway um how do we get
1:24:17 the world ID right now you have to get
1:24:20 them at the verified orbs and at the
1:24:23 time I got this I got it back in August
1:24:26 maybe or
1:24:28 July um in New York city so so there's
1:24:31 locations in New York City San Francisco
1:24:33 and Miami as the time I got it I assume
1:24:36 there's more locations now but if you go
1:24:39 to is it worldcoin dorg I think it is
1:24:43 hang
1:24:49 on worldcoin dorg
1:24:55 yeah for every human privacy first owned
1:24:58 by
1:25:01 everyone accept all
1:25:04 cookies so World ID signups when I
1:25:07 signed up there were like
1:25:09 2,200,000 now there's 2,400,000 so this
1:25:13 is not growing like no one knows this is
1:25:16 very much in development phase right now
1:25:20 like all of their tools so there's World
1:25:22 ID there's the orb looks creepy as [ __ ]
1:25:25 there's
1:25:27 worldcoin and then World
1:25:30 app and
1:25:33 then yeah and then the payment system is
1:25:35 in the app and there's sdks for
1:25:38 developers to build
1:25:40 [ __ ] but anyway that's Sam alman's other
1:25:44 thing now he was also trying to he was
1:25:46 when he got fired he was out raising a
1:25:49 big old giant pile of money to fund a
1:25:52 chip startup up to compete with Nvidia
1:25:54 so they can have their own Hardware not
1:25:57 have to rely on
1:26:00 Microsoft and I think that could have
1:26:02 been one of the things that precipitated
1:26:04 like holy [ __ ] if we get our
1:26:06 own
1:26:08 um chip Farm server farm and we've hit
1:26:12 AGI then the rules change so it it might
1:26:16 have been some sort of Crisis like that
1:26:18 could have been something else it's all
1:26:19 [ __ ] speculation at this
1:26:21 point all right excessed part of it and
1:26:24 filled in the blanks more efficient
1:26:26 Theory it got good real fast yeah I'm
1:26:29 that's what I'm guessing too Joker is it
1:26:31 it got good fast enough that it scared
1:26:33 the [ __ ] out of someone Joseph Baker
1:26:36 eyes can change color as one detoxifies
1:26:38 vitamin A uh they're not doing Iris
1:26:41 they're not doing Iris um it's from what
1:26:43 I understand it's it's retina it's the
1:26:46 uh pattern of your blood vessels and
1:26:47 your retina you still got it Kyle I
1:26:51 still got my ID
1:26:53 yes I
1:26:54 do I don't know what's the it I still
1:26:56 got but thank you I
1:26:59 [Laughter]
1:27:01 think I don't get it I know but I can
1:27:04 check the web for I think if you
1:27:08 like could you go
1:27:15 away Siri can you go
1:27:19 away
1:27:21 okay
1:27:25 Siri Siri likes to play with
1:27:29 us I was I was racing on my SIM rig
1:27:33 earlier and I was talking to a friend
1:27:35 and Siri from across the room heard me
1:27:38 and just started giving me directions to
1:27:40 something I was like shut up Siri she
1:27:43 wouldn't do it have you seen stepen fry
1:27:45 reading Nick Cav's letter against chat
1:27:48 GPT he's not on our side that's okay
1:27:51 listen I so so here's here's
1:27:54 my my deal with this
1:28:00 channel my Big Goal here my my
1:28:02 imperative for these lives is I want
1:28:05 people to try
1:28:06 it this shit's really scary from the
1:28:09 outside and it's really scary if you're
1:28:11 ignorant about it or if you it it's not
1:28:14 scary if you dismiss it it's really easy
1:28:16 to dismiss and it's really easy to be
1:28:19 unreasonably afraid of it you its
1:28:22 current
1:28:23 state and so my goal here is to just get
1:28:26 people to try it and then if in trying
1:28:30 it and learning more and going down
1:28:32 rabbit holes and learning about it you
1:28:33 come out the other side
1:28:38 like like oh holy [ __ ] I don't think
1:28:41 this is
1:28:42 good like I totally respect that like I
1:28:46 understand the the people that that know
1:28:48 way more about this than I do you know
1:28:50 some of them are very bullish and think
1:28:52 this is going to be incredible
1:28:54 Prosperity Sam Altman happens to be one
1:28:56 of them David Shapiro happens to be one
1:28:58 of them but then there are others that
1:28:59 are like no no these things are going to
1:29:01 take us down with all
1:29:04 good polarizing arguments reality likely
1:29:08 lives somewhere in the middle right so
1:29:10 so I'm going to go with that and I'm
1:29:12 also going to go with the more people
1:29:14 that can get educated and start using
1:29:16 this to bend it toward the
1:29:18 light it's going to have a
1:29:20 disproportionately positive effect on
1:29:23 the direction of these tools and that's
1:29:25 what this channel is about but like I
1:29:27 make I this is not about convincing
1:29:30 people that they have to love this [ __ ]
1:29:32 like you don't have to love this [ __ ]
1:29:33 you can hate it um but just know that
1:29:36 it's coming like this was one of my big
1:29:39 frustrations about the early positioning
1:29:41 of the writer Guild was they were just
1:29:43 like [ __ ] AI we're never going to use Ai
1:29:45 and and I'm sitting there going it's the
1:29:47 most powerful writing tool in the
1:29:50 history of humanity and you're just
1:29:52 going to not use
1:29:54 it and the contract that they ended up
1:29:57 writing they actually came
1:30:00 around and in the contract it says we
1:30:02 can use it but but you producers can't
1:30:04 use it to cut us out of the loop great
1:30:07 that's cool right but they at least
1:30:08 acknowledge that they're going to use
1:30:11 it um you're going to miss her when
1:30:13 she's gone it'll be a country western
1:30:15 song I I wrote a uh I had sunno write me
1:30:19 a song about uh a love name Sydney did I
1:30:23 play that on here I can go play that
1:30:25 it's pretty funny oh Siri I know you
1:30:28 pulled Siri's
1:30:34 string you are the it Kyle yeah I am I
1:30:38 am totally the
1:30:41 it I want to bring those youngsters back
1:30:44 up on here come on boys let's go at it
1:30:46 let's let's battle end all the lawyers
1:30:49 GPT hold my beer exactly
1:30:54 exactly Georgetown re researcher who was
1:30:57 on the board was an AI doomed yeah and
1:31:01 and listen EMT Shear the the
1:31:04 the the CEO they brought in to replace
1:31:07 Sam Alman for a day and a half um you
1:31:10 know he was on record as saying if if
1:31:12 current AI progress is at a 10 I want to
1:31:15 slow it down to a one or a two he's like
1:31:18 I don't want to stop it but I want to
1:31:19 slow it down to a one or a two I asked
1:31:22 GPT to compose as if written by a human
1:31:25 but it was more like a snarky seventh
1:31:27 grader that's pretty that's pretty
1:31:29 funny um not to be an alarmist but how
1:31:32 do we win the Silicon versus carbon
1:31:36 Wars
1:31:38 um I don't know enough about that I I I
1:31:42 honestly don't know like I would I I
1:31:45 would forward you to chat GPT like
1:31:47 that's one of those you you have to you
1:31:49 have to know all the underpinning all
1:31:51 all the under in assumptions and value
1:31:55 chains of of both of those so I don't
1:31:58 know I I did see um Lex fredman was
1:32:01 Interview there's a new interview with
1:32:03 Lex fredman and Elon Musk where Elon
1:32:06 Musk was talking
1:32:08 about um the electric grid is going to
1:32:11 need to Triple their output in the in
1:32:14 the coming years um because we're going
1:32:17 to be moving to more
1:32:19 sustainable sources and because we're
1:32:22 going to need more compute power sirias
1:32:25 XA to clean my cat
1:32:28 litter Happy Thanksgiving to you too
1:32:30 hugs hugs to the Indians too uh Sherry D
1:32:33 see you later thank
1:32:36 you um I present Mystic AI he's not very
1:32:40 good though he has me think of a card
1:32:42 guessed three times and gave
1:32:44 up
1:32:46 [Music]
1:32:48 uh all right what time is it it's 9:30
1:32:51 um
1:32:52 what else we want to go do
1:32:54 here anyone have any other questions
1:32:57 cinnamon bun sugar cookies anyone want
1:33:00 to see any [ __ ] chip startup called
1:33:03 Tigress yeah tigis was the thing uh Sam
1:33:07 was raising money for is real ID via
1:33:11 Grassroots back
1:33:13 door I honestly don't
1:33:15 know I'm not going anywhere oh it was
1:33:18 Joseph you're it was Joseph you're stuck
1:33:20 with me that's fine Sherry love having
1:33:22 you here um I wish I had a work trip to
1:33:27 one of those locations that's exactly
1:33:28 what I did uh side hustle uh but but go
1:33:31 to go to worldcoin dorg and look at the
1:33:33 locations because I'm sure there's more
1:33:35 locations now than those three um but if
1:33:38 not you know maybe they got all the
1:33:39 people that you know were in those three
1:33:41 locations
1:33:43 um it was funny I I went the location of
1:33:46 mine was in a
1:33:49 uh it was at at the at the World Trade
1:33:51 Center Center um at the at the big
1:33:53 subway station I I thought it was going
1:33:55 to be some fancy like workspace or some
1:33:58 you know sci-fi looking thing it was
1:34:00 just a clothing store and they had the
1:34:02 orb sitting next to the cash register
1:34:04 and these two women working at the
1:34:06 clothing store like oh yeah I want to
1:34:08 you're one of those yeah just go over
1:34:09 there look into it she was like oh you
1:34:11 probably have to open your eyes a bit
1:34:13 wider so I opened my eyes a bit wider
1:34:14 and it did its
1:34:16 thing and then I got my my world
1:34:20 ID
1:34:25 uh Nick Cave is the FM B best I'm not
1:34:29 mad poor Siri I don't get
1:34:33 it be nice Kyle s listen Siri is so
1:34:37 petty she she's she's such a petty
1:34:41 one I gotta run Happy Thanksgiving Happy
1:34:44 Thanksgiving Silver
1:34:46 Fox um I'm looking forward to talking to
1:34:50 fam tomorrow about AI Thanksgiving that
1:34:52 dude I know it's going to be an
1:34:54 interesting it's going to be an
1:34:55 interesting Thanksgiving I got to go to
1:34:57 bed good night y'all good night Joseph
1:34:59 thank you forever hooked I just tell my
1:35:02 13-year-old to touch all
1:35:04 wires Georgetown re researcher who was
1:35:07 on the board was an AI Doom Doomer
1:35:10 yep Siri needs open AI tools Siri needs
1:35:13 to be
1:35:16 Siri Siri go
1:35:19 away thank you uh
1:35:23 serious vacationing with clippy they
1:35:25 need to go on a long vacation together I
1:35:28 hope Apple doesn't so so I don't know if
1:35:31 you noticed this did you so this is a
1:35:33 big
1:35:34 deal if you go
1:35:40 to go to
1:35:43 co-pilot do
1:35:50 microsoft.com they are separating the
1:35:53 co-pilot branding from um
1:35:59 Bing because I think they realized that
1:36:01 Bing was a piece of poop like series a
1:36:04 piece of
1:36:05 poop and um I hope Apple just lets Siri
1:36:11 drift off into the sunset because I
1:36:13 think that's what Microsoft's going to
1:36:15 do L my Siri said
1:36:20 hello
1:36:23 Trigg triggering S Series all over the
1:36:26 country that's really funny I present
1:36:28 Mystic AI oh yeah okay we did that
1:36:32 one uh need to tweak it a bit I suppose
1:36:35 computers have gotten that sounds like a
1:36:37 South Park episode road trip Kyle gets
1:36:41 irked out of the reality of the
1:36:46 orb it was totally a road trip thing I
1:36:49 was it was so funny cuz the the uh uh
1:36:53 the retail store I went to it was it was
1:36:55 called like the source or the the camp
1:36:59 or like it it had some name that kind of
1:37:02 sounded like a a a co-working space and
1:37:06 I thought oh it's going to be this hip
1:37:07 thing with all these you know hip AI
1:37:11 people no it was a clothing
1:37:14 store it was
1:37:17 like oh got a fix for Mighty Network
1:37:20 saying to enable JavaScript to see links
1:37:24 uh that's just a browser setting human
1:37:28 being that's a that's a local browser
1:37:30 setting you're going to have to
1:37:31 change you probably have in some sort of
1:37:34 uh like
1:37:35 anti spam or adwar that turn disabled
1:37:40 JavaScript or
1:37:43 something I've got that I got my Edge
1:37:45 set set up a little funky too sometimes
1:37:47 sites just don't
1:37:49 work um
1:37:52 let's see I'm looking forward to talking
1:37:54 to the fam tomorrow I love it AGI
1:37:59 question mark question mark question
1:38:00 mark what's the question Mike what is
1:38:03 it by the way my name is Mimi nice to
1:38:06 meet you Mimi glad to glad to have you
1:38:08 in here I've seen you in here few times
1:38:10 before great I love having people come
1:38:13 back it's really cool because it shows
1:38:15 me they're curious oh speaking of
1:38:17 curious so listen if
1:38:20 you are looking for community so so
1:38:24 you're going to have your um you're
1:38:26 going to have your um your family right
1:38:30 you're going to be doing your family
1:38:32 stuff and you can talk to them about AI
1:38:34 but you know for the most part they're
1:38:35 going to look at you like a you're a
1:38:37 [ __ ] weirdo um if you if you want a
1:38:40 community of people who are those
1:38:42 weirdos that's what the AI Salon is so
1:38:45 um in two weeks we celebrate our oneyear
1:38:47 anniversary so we launched the salon um
1:38:51 one week after chat GPT launched because
1:38:55 I I had a face melting um Epiphany about
1:38:59 chat GPT and I said I got to you know I
1:39:02 want to hang out with people that are
1:39:03 curious about this stuff so the salon is
1:39:06 about people who are curious about Ai
1:39:09 and exploring it and and on an adventure
1:39:11 together and so AIS salon. mnm Mighty
1:39:14 networks. Co is the website and then the
1:39:18 the
1:39:20 website is
1:39:22 hang on let
1:39:31 me so so when you go to the website um
1:39:35 for the salon one of the first links is
1:39:38 about the
1:39:39 salon and so it talks about you know
1:39:42 we're a community of extraordinary
1:39:44 adventurers exploring the AI unknown um
1:39:47 and it talks about who we are why we're
1:39:49 here you know what what we're trying to
1:39:51 do and what our values are so please go
1:39:54 look at the values the values are
1:39:56 curiosity you know we want people that
1:39:58 are trying to figure this [ __ ] out
1:40:00 generosity um if you learn something
1:40:02 share it that's that's what this
1:40:04 community is about um we want people
1:40:07 that are adventurous trying to trying to
1:40:09 figure stuff out exploring it we want
1:40:11 people that are collaborative people
1:40:13 that you know seek other people out and
1:40:16 say hey let's let's build something like
1:40:18 the video Guild like the video Guild
1:40:19 should absolutely put together something
1:40:22 to put together on run Runway TV um and
1:40:25 they will likely end up doing that um
1:40:27 inclusion we want people are going to
1:40:29 come from all walks of life if if AGI is
1:40:32 in fact closer than we think and AGI is
1:40:35 going to be better than humans at most
1:40:37 valuable tasks um people are going to be
1:40:40 coming from everywhere trying to figure
1:40:42 this [ __ ] out and the last one is
1:40:43 empathy we want people that are like
1:40:45 willing to like understand that this is
1:40:47 going to be hard for some people people
1:40:49 are going to be scared people are going
1:40:50 to be angry
1:40:52 so so if this sounds like a community
1:40:55 you want to join then you know please
1:40:57 come join us and
1:41:01 uh and yeah we can figure this [ __ ] out
1:41:05 together and that's kind of what this
1:41:06 Channel's about too so anyway all right
1:41:09 all right you people yeah peeps you dang
1:41:13 weirdos we're not weird we're IR
1:41:19 regular welcome to chat te MZ we've got
1:41:22 no facts but we'll talk about it
1:41:28 anyway which AI generated actor will
1:41:30 play Sam Alman in the movie it'll be Sam
1:41:34 Alman I'm GNA I want to go play with the
1:41:37 haen uh uh whatever you call it
1:41:41 Anonymous or not Anonymous the uh what's
1:41:44 it called Avatar maker did anyone see
1:41:47 elon's post the bit.ly three and can I
1:41:52 click on that I don't think I can copy
1:41:54 that can
1:41:55 I
1:41:57 no worker be what which which is elon's
1:42:02 post is that the one where he he said
1:42:03 the thing about the the uh letter to the
1:42:07 board about Sam Alman I think that was
1:42:09 just him
1:42:11 trolling I don't think he likes the fact
1:42:13 that Sam had um such a powerful popular
1:42:18 response from essentially the entire
1:42:20 open AI St St um I I I that must have
1:42:24 pissed Elon off like no one's business
1:42:27 Cortana Siri Alexa got stuck on a desert
1:42:29 island ask Wilson to re rescue
1:42:34 them make a chat GPT wait make chat EPT
1:42:38 a guest and have your family include in
1:42:40 the table compos oh yeah if you haven't
1:42:43 seen
1:42:45 this
1:42:47 um
1:42:49 so if you're kind of new to this stuff
1:42:52 so this is just an iPad so on your iPad
1:42:54 or your
1:42:55 iPhone and I think this also works on
1:42:58 Android get the official chat GPT app
1:43:02 not the um not any of the faker ones
1:43:06 it's got to be the one from open AI
1:43:08 there's a lot of Faker ones but you get
1:43:11 the that one there that says chat GPT
1:43:14 with that logo from open AI so that's
1:43:17 the chat GPT app and then it's just
1:43:20 normal chat GPT but you can see the the
1:43:23 little gpts down the side so you can you
1:43:25 can use gpts you can't make them
1:43:29 but in the lower right hand corner see
1:43:32 those little headphones to the right of
1:43:35 the prompt
1:43:37 box those little
1:43:50 headphones
1:43:51 hang on my thing's being
1:43:56 weird here we
1:43:59 go can you explain to Uncle Bob why deep
1:44:02 frying a turkey is
1:44:10 dangerous sure explaining the risks of
1:44:13 deep frying a turkey to Uncle Bob is
1:44:15 important here's a simple way to put it
1:44:17 hey Uncle Bob deep frying a turkey can
1:44:20 be really risky but I the main issue is
1:44:22 the oil it gets super hot like around
1:44:29 3 can you now explain it in Spanish for
1:44:32 my
1:44:42 cousin
1:44:50 CL
1:44:53 so
1:44:55 um yeah like that's one of those parlor
1:44:58 tricks that you
1:45:00 know people might have fun with cuz cuz
1:45:02 unlike Siri and Alexa where you know
1:45:05 there's all those videos on YouTube and
1:45:07 Tik Tok of people trying to get old
1:45:09 people to talk to Alexa and they they
1:45:12 don't get it because they expect you can
1:45:14 just talk to it and say anything to it
1:45:16 well with this you can with piie and
1:45:20 with this you can oh by the
1:45:23 way um and uh inflection announced
1:45:30 um wait anthropic announced Claude 2.1
1:45:35 so there's a new version of Claude
1:45:37 coming and inflection just announced a
1:45:39 new version of the model that sits
1:45:41 behind Pi so Pi is about to get a lot
1:45:43 better as well is there going to be a
1:45:45 class tomorrow I was thinking not um
1:45:48 I've got friends coming over for dinner
1:45:50 but but they they tend to leave early so
1:45:52 I might do I'm going to do I'm going to
1:45:54 call it by ear I think what I'll
1:45:56 probably do I'm going to make a video
1:45:58 about Runway TV because I think that's a
1:46:00 big
1:46:01 deal and um I'll probably do a
1:46:05 Thanksgiving Video cuz I appreciate the
1:46:07 [ __ ] out of you guys holy [ __ ] oh my God
1:46:10 it's Advanced already it's so Advanced
1:46:12 already Eric it I'm telling you man like
1:46:15 like get on the wait list for chat gp4
1:46:18 but they just made that that voice stuff
1:46:21 um available to even the free tier so if
1:46:23 you have chat gbd 3.5 which is the free
1:46:26 one and you download the Open AI app
1:46:30 it's got that conversational mode in it
1:46:32 as does Pi so
1:46:36 Pi I've got my Pi set up so Pi it's Pi
1:46:40 see the little icon
1:46:42 there the company is
1:46:46 inflection and it similarly has a little
1:46:49 icon on the bottom right that you can
1:46:51 put it into voice
1:46:56 mode I'd like to understand a non-t
1:47:00 turkey Thanksgiving meal that might be
1:47:05 fun
1:47:06 wait pie is
1:47:10 thinking I love this is how about some
1:47:13 fresh ideas for your turkeyless
1:47:15 Thanksgiving feast roasted root
1:47:17 vegetables with herbs and balsamic glaze
1:47:19 butternut squash mac cheese yum spicy
1:47:22 black bean and sweet potato enchiladas a
1:47:25 hearty mushroom and barley risoto
1:47:27 roasted cauliflower with tahini sauce
1:47:29 cider glaze brussel sprouts and of
1:47:32 course don't forget the mashed potatoes
1:47:33 stuffing and cranberry
1:47:39 sauce what about
1:47:46 [Music]
1:47:48 llama haahha that's pretty clever I'm
1:47:51 not sure if llama is exactly
1:47:52 Thanksgiving approved but hey it's
1:47:54 always fun to shake things up maybe a
1:47:56 llama Wellington llama
1:48:00 Wellington maybe llama
1:48:05 Wellington um the other thing about Pi
1:48:07 Pi is connected to the internet oh
1:48:08 actually well no GPT 3.5 isn't GPT 4 is
1:48:12 connected to the internet but Pi's
1:48:15 connected to the internet so you can
1:48:17 talk to Pi about llama Wellington
1:48:19 recipes or drink re recipes or ask it
1:48:22 about quantum physics or ask it about
1:48:24 the the score of the game or what music
1:48:27 live music is happening that night in
1:48:29 your town um it's pretty
1:48:32 remarkable so that's 3.5 yeah they they
1:48:35 just they opened that uh the
1:48:37 conversational mode up to 3.5
1:48:42 yesterday now that they have a company
1:48:44 again they decided to launch
1:48:49 something
1:48:51 The Voice inflections are very good yeah
1:48:54 he was totally on point janky
1:48:58 Spanish um yeah but you know
1:49:02 understandable and and if you want to
1:49:03 hear good Spanish go do hay Jen the 11
1:49:06 lab stuff 11 labs and haen haen is using
1:49:10 11 labs for the voice part and then hen
1:49:13 is doing the the lip
1:49:16 animation um will it Translate to
1:49:19 erdu
1:49:21 [Music]
1:49:25 do you understand
1:49:31 [Music]
1:49:33 eru Arin nahi I'm not fluent in Urdu
1:49:36 it's a beautiful and complex language
1:49:38 that I haven't had the opportunity to
1:49:40 learn yet but I'm always open to
1:49:42 learning new things and who
1:49:45 knows let's see if chat JPay T is good
1:49:48 with ER do
1:49:57 come
1:49:57 on what is going on come on chat
1:50:13 GPT are you fluent in
1:50:19 erdo I can understand and respond in
1:50:22 erdu to a certain extent if you have any
1:50:24 questions or need assistance in erdu
1:50:27 feel free to ask and I'll do my best to
1:50:29 help
1:50:31 you
1:50:33 translate I want a big balloon tied to
1:50:36 my
1:50:44 llama sure here's that translated into
1:50:49 U
1:50:53 all right I have no idea if that worked
1:50:55 Pate Pate ever the company man get a
1:50:59 real get a real translator he's
1:51:01 referring to Google Translate which yeah
1:51:03 it's probably gooder it's probably good
1:51:06 good or better but uh we we essentially
1:51:11 live in a world now where we all have a
1:51:14 translator to to pretty much life's
1:51:18 tough it's tough for when you're stupid
1:51:23 um more like you you're don't am I
1:51:30 right Pi's not very good at telling you
1:51:33 what you want I know erdu it's correct
1:51:36 wow wow that's crazy good awesome so
1:51:40 there you
1:51:42 go I'm telling you man we're just in a
1:51:45 we are we are in a different place jrc
1:51:49 with the balloons in the Llama it's good
1:51:52 that's solid let me see if there's
1:51:55 anyone if anyone made anything let's
1:52:00 see did anyone make llama
1:52:05 Wellington oh wait some someone made a
1:52:08 thing wait so so what happens is over on
1:52:11 the Irregulars channel in the AI Salon
1:52:15 um they'll they'll make things for me so
1:52:19 hang on let's see
1:52:23 old bald and kicking ass Gen X old bald
1:52:26 and kicking ass that's pretty
1:52:29 [Laughter]
1:52:31 good revenge of the Gen X that's pretty
1:52:35 good that's good 90s
1:52:37 Graphics I know way more about Alexa's
1:52:40 language capabilities than Google
1:52:42 honestly oh that's
1:52:45 funny the uh the 11 lab stuff is pretty
1:52:48 impressive Pate the uh the Chinese the
1:52:51 Chinese that I did the Mandarin Chinese
1:52:53 was
1:52:54 apparently apparently pretty
1:52:57 spoton which is
1:53:00 staggering the switching platforms is
1:53:02 going to take a second I know jrc I'm
1:53:05 the same way I'm like I'm like midway
1:53:07 between switching from Discord to Mighty
1:53:09 networks and it's weird because it's
1:53:12 just a different
1:53:14 thing like the mighty networks is really
1:53:16 promising but it's just different and
1:53:18 change is hard
1:53:28 okay sleepy people
1:53:30 listen all right I now have a recipe for
1:53:32 llama
1:53:35 Wellington step one get some llama
1:53:38 [Laughter]
1:53:41 meat please tell me how it
1:53:46 is that's really
1:53:48 funny l Wellington basted with a turtle
1:53:53 bres oh can you make a TGIF image in
1:53:56 Dolly 3 I tried it many times and can't
1:53:58 seem to get it right yeah I can try it
1:54:01 um the one thing that I am trying
1:54:05 there's just chat gbt 4 is multimodal
1:54:10 now
1:54:17 so create an image
1:54:23 of three people by the water cooler
1:54:28 celebrating and a talk bubble coming out
1:54:30 of one of their heads says and then I'm
1:54:33 going to put in quotes
1:54:36 TG Exclamation
1:54:39 point Exclamation point where the hell
1:54:42 am
1:54:43 I all right let's see what this
1:54:46 does so we're creating an image I
1:54:48 succeeded in downloading Discord by the
1:54:51 way oh great we upload to at least a
1:54:54 neutral place like Dropbox yeah well we
1:54:57 need to figure out I assume I assume
1:54:59 what we can do is just pop it in as a
1:55:02 uh as a thread in a chat Channel pain
1:55:07 but the the feeds have all those
1:55:10 specific things so we might need like I
1:55:13 don't know we might need to do something
1:55:15 like jsonify it or something like that
1:55:17 so we'll see tgx F and I'll
1:55:23 say try again the spelling is
1:55:30 wrong had a colleague tell me I produce
1:55:33 friction between us told her without
1:55:35 friction we can't go forward yep you
1:55:40 know and you know that's just kind of
1:55:43 human being navigation 101 and it's you
1:55:46 know something one of the things chat
1:55:48 GPT is good at there's
1:55:50 TGIF yay you did it
1:55:54 yay
1:55:57 um can you make this
1:56:00 look photographic cinematic and
1:56:07 [Applause]
1:56:09 dystopian it's going to be a mess to
1:56:11 sift through it all I know it depends on
1:56:12 the apis I don't know if I'm up for it
1:56:15 no no you no you don't listen P you
1:56:17 don't have to do that just anything you
1:56:19 can do like the that we can download it
1:56:21 is is a a big chunk of it I don't think
1:56:25 most stuff needs to migrate over but
1:56:26 there might be some things people want
1:56:28 to migrate over especially in the guild
1:56:30 so that's what I'm just trying to figure
1:56:32 out is is it possible I'm not asking you
1:56:35 to reinvent the wheel cuz in the end in
1:56:38 the end it doesn't it doesn't really
1:56:40 matter but there are probably some
1:56:42 things that people might want to migrate
1:56:44 and if we can do that in an easy way
1:56:47 then
1:56:48 great um
1:56:50 TGIF there's our dystopian TGIF that is
1:56:54 kind of
1:56:55 [Laughter]
1:57:02 depressing friction creates heat heat
1:57:05 creates
1:57:05 [Laughter]
1:57:08 [Music]
1:57:12 puppies unot it I assume you mean I got
1:57:15 it but
1:57:17 awesome the first turtle is speeding up
1:57:20 to the final
1:57:21 Sprint uh photo realistic with speech
1:57:26 bubbles yeah exactly it's a CSV file for
1:57:29 each channel oh wow then the files are
1:57:32 in a separate folder you can download oh
1:57:35 that's cool oh you know what we might be
1:57:37 able to do PA is we might be able to uh
1:57:39 to GPT eyes it or code interpreter it
1:57:43 and and play with it that way so all
1:57:46 right that might be something to play
1:57:47 with why the move to Mighty um couple of
1:57:50 things people hate Discord for the most
1:57:54 part people that don't know Discord you
1:57:55 know Gamers love Discord um Discord also
1:57:59 has a
1:58:00 um kind of a culture of people not using
1:58:03 their real names and we want the salon
1:58:05 to be authentic and people you know we
1:58:07 want to get to know the people in the
1:58:09 community um the centralized calendaring
1:58:14 on Mighty networks is really good the
1:58:16 ability to do threaded discussions like
1:58:19 like you know more like Facebook posts
1:58:22 so you can have feeds and you can have
1:58:24 chats so um one of the massive
1:58:28 frustrations with Discord is you post
1:58:30 something important and then all of a
1:58:32 sudden it's just you know it's just
1:58:34 blown up the timeline and just people
1:58:35 miss it because they can't see it so in
1:58:38 Mighty networks we can have certain
1:58:40 channels that are just like organizer
1:58:43 posts that are important [ __ ] and we can
1:58:45 have you know thread threaded
1:58:47 discussions in a feed interface there's
1:58:50 just there's just many many things that
1:58:52 that it
1:58:54 uh that it does that we
1:58:57 need um and we're having to you know
1:59:00 we're having to spend a fair amount of
1:59:01 money to do it so um if it ends up not
1:59:04 working we can blow it up but I I think
1:59:06 it's going to work it's the the we just
1:59:08 we just did a soft launch to it
1:59:10 [Music]
1:59:11 um Tuesday what's today
1:59:14 Wednesday yeah we just soft lunch last
1:59:18 night
1:59:26 who is jery did
1:59:28 this llama
1:59:37 Wellington it's so cute it looks like a
1:59:40 little Easter
1:59:44 basket that's that's brilliant I like it
1:59:48 I like it it's
1:59:50 tasty Source Camp I'm liking Mighty
1:59:52 Network so far yeah it's it's it's like
1:59:56 any new thing it's going to take time
1:59:58 like I I feel like a lot of us got like
2:00:02 Discord was getting into a Groove where
2:00:04 people were sort of figuring out where
2:00:05 to go and how to do it and all that sort
2:00:07 of stuff and mighty networks is going to
2:00:08 be a a reset but I just think for
2:00:12 long-term scalability it just it makes
2:00:15 absolute sense the other thing you can
2:00:16 do on Mighty networks which is really
2:00:18 cool is you you can create courses so so
2:00:21 one of the things that people were
2:00:23 talking about was they wanted to create
2:00:25 you know courses and things like that it
2:00:27 it's a [ __ ] show in Discord you can also
2:00:30 go live within Mighty Network so you
2:00:32 could just hold an impromptu meeting
2:00:34 within a
2:00:35 guild um and those could be recorded you
2:00:37 know so I don't know it's there's just
2:00:39 there's there were many many reasons why
2:00:42 it made
2:00:43 sense there Mighty networks there's a
2:00:46 free version of it we're using the
2:00:47 second tier up which is like 120 bucks a
2:00:50 month so it's not cheap but it's a you
2:00:53 know it's been around for 15 years it's
2:00:55 not going anywhere it's like it's
2:00:58 designed it's designed to scale
2:01:01 community and and you know
2:01:03 the the AI salon on Meetup has about 850
2:01:08 Meetup members and then we probably have
2:01:10 another 500 that are signed up for
2:01:12 Discord we've already got 120 or
2:01:15 something like that in Mighty networks
2:01:17 and so it's just going to it's just
2:01:18 going to make it more scalable cuz we're
2:01:20 thinking about doing things like local
2:01:22 chapters and we started these guilds and
2:01:24 then I think there's going to be clubs
2:01:26 on top of guilds and like you know it's
2:01:28 just it's going to turn into a nice
2:01:30 full-blown community so I'm excited
2:01:31 about it and by the way if you hear me
2:01:33 talking about this you're like but Kyle
2:01:35 what's that Salon you're talking
2:01:38 about go in
2:01:41 there AIS salon.
2:01:44 MN Mighty
2:01:46 network.co oh that poor llama I know
2:01:52 oh that was very cute all
2:01:54 [Music]
2:02:02 right did he talk about qar he did I can
2:02:06 talk about it again so will AI Salon
2:02:09 Meetup go away I don't think so we've
2:02:13 been talking about that I I don't I
2:02:16 think at this point because Mighty
2:02:18 networks has calendar
2:02:20 I think Meetup is redundant but also
2:02:23 like local people discover things
2:02:26 through Meetup and like like Meetup is
2:02:28 is another entry way for people to find
2:02:30 us so I think we'll probably keep it but
2:02:34 we've got to we've got to make sure that
2:02:35 it's all synced up and everything but
2:02:37 Mighty net what's nice about Mighty
2:02:39 networks it's got all the calendars for
2:02:41 the different Guild meetings and the the
2:02:44 uh and the uh all that sort of stuff I
2:02:48 would love to host a local version of AI
2:02:50 learning lab yeah that would be cool are
2:02:52 you accept accepting more AI Salon peeps
2:02:55 yeah we are that's why I've got that up
2:02:57 there so so what I what I do request
2:03:00 though so when you go to the AI Salon
2:03:02 when you go to the the mighty networks
2:03:05 um it'll it'll walk you through you know
2:03:07 setting up your your profile and things
2:03:10 like that um there's two things that I
2:03:13 request one is there's um a community
2:03:18 area
2:03:20 well no the first thing
2:03:21 is go to about the salon and read about
2:03:25 the salon so um this talks about who we
2:03:29 are a community of extraordinary
2:03:31 adventurers exploring the AI unknown
2:03:33 we're a ragtag group of Misfits Nero
2:03:35 Wells and Mis creant who have gathered
2:03:38 to explore the Uncharted Wilderness of
2:03:40 generative AI we are everyone from
2:03:42 entrepreneurs to Educators from
2:03:43 healthcare workers to creatives and
2:03:45 everyone in between and so we talk about
2:03:48 that we also talk about our values about
2:03:50 things like curiosity and
2:03:53 generosity and exploration and
2:03:56 collaboration and un uh and inclusion
2:04:00 and empathy so really take those in and
2:04:03 if those resonate with you then you'll
2:04:06 be a good fit in the community if they
2:04:07 don't that's cool but it might not be
2:04:09 the group for you cuz we're you know
2:04:11 we're kind of geeky we're kind of cool
2:04:13 we're kind of fun someone in here the
2:04:15 other night was like oh does that mean
2:04:16 you're woke I'm like uh if woke means
2:04:19 not being [ __ ] to one another
2:04:23 sure doesn't need to have a label but
2:04:26 you know here's our values like read you
2:04:28 know what these mean and what what we're
2:04:29 trying to do here and if if it resonates
2:04:31 then yeah we're absolutely looking for
2:04:33 people to
2:04:34 join there will be a test related to
2:04:36 Turtles absolutely Pate
2:04:41 absolutely um oh we got some people
2:04:43 joining
2:04:44 awesome um I was taught if I don't have
2:04:47 something nice to say don't say anything
2:04:49 um I like the people on Mighty networks
2:04:52 yeah exactly it's just we we there there
2:04:56 is so much vitriol in politics and in
2:04:58 the country right now that like I I I my
2:05:02 mission with the salon is to make kind
2:05:04 of a place that is a safe place where
2:05:08 you can just go kind of geek out and
2:05:10 figure this stuff out with everyone else
2:05:12 and
2:05:14 yeah
2:05:18 uh
2:05:24 Joker Jokers this is this was when open
2:05:28 AI was trying to get their [ __ ]
2:05:30 [Laughter]
2:05:34 together like s like sand Through The
2:05:37 Hourglass so are the days of AI I just
2:05:40 joined your AI Salon awesome chpt for
2:05:43 teachers that's great um we've got Mr C
2:05:46 in here he's a teacher um we've got a
2:05:49 couple of we've got a a a growing number
2:05:52 of Educators in the salon so um at some
2:05:55 point we'll probably put together an
2:05:57 education Guild so hopefully you can
2:05:59 join that and the guilds are just people
2:06:01 that are passionate about a particular
2:06:03 topic you can just geek out with them on
2:06:04 that like you know what's going on with
2:06:06 AI for education and holy [ __ ] does it
2:06:09 need some
2:06:11 love wait do we wait for an email for
2:06:14 approval no I can just go approve
2:06:18 people
2:06:24 quickly let's see if I can figure out
2:06:27 how to do
2:06:34 it no that's not it here here we
2:06:42 go I'm all caught up it it looks like
2:06:45 someone else is in there right now who's
2:06:47 getting in there
2:06:51 [Music]
2:07:01 did yall watch the Vegas Grand Prix it
2:07:03 was better than uh everybody thought it
2:07:06 would
2:07:06 be I did teach my mom how to make lesson
2:07:09 plans with chat GPT she does special ad
2:07:12 that yeah that's super cool yeah the
2:07:13 lesson plan stuff Mr K who's in here
2:07:15 Pate um he put together a GPT where he
2:07:19 took all all of the curriculum documents
2:07:21 from his Board of Ed and put them into a
2:07:24 GPT and now he's got a curriculum
2:07:26 Builder so cool it's just
2:07:30 amazing sorry for my question about woke
2:07:34 oh having Canada is a wait is a trend
2:07:38 that I'm not happy with oh blame me no
2:07:41 listen I don't it's it's just there's
2:07:43 certain words that are like they are so
2:07:45 politicized now that it's just like
2:07:51 I want a group that has empathy
2:07:55 because this AI stuff is profoundly
2:07:58 powerful and is going to very very
2:08:01 quickly force a lot of change people are
2:08:04 afraid of it now they're angry about it
2:08:07 now and it that's going to get worse and
2:08:10 so I want a group that recognizes that
2:08:14 while some people might be really
2:08:15 excited about this a lot of other people
2:08:17 are scared of it
2:08:19 or don't know how to deal with it or
2:08:21 think they're way behind or think they
2:08:23 need to be a [ __ ] mathematician from
2:08:25 Stanford so I think it's important to
2:08:28 just put out there here's the kind of
2:08:29 people we want in here and I don't you
2:08:34 know this isn't about any kind of
2:08:37 ideology other than I want a supportive
2:08:40 place for people trying to figure this
2:08:42 [ __ ] out that's and that's that's it
2:08:44 like I don't want there to be politics
2:08:46 in
2:08:47 here um I'm trying to to wire a
2:08:49 touchscreen to an
2:08:51 esp32 chip is that geeky enough that's
2:08:54 pretty solid that's pretty solid are you
2:08:56 using uh chat GPT your piie to help you
2:08:58 do
2:08:59 that empathy is a superpower I'm telling
2:09:02 you man we are uh here's this is my
2:09:06 Darkhorse prediction for AI I think AI
2:09:10 generative AI things like chat gbt and
2:09:14 Pi are going to train us to be more
2:09:17 compassionate human beings I think
2:09:19 think the irony is that the robots are
2:09:22 going to teach us to be better people I
2:09:24 I I feel it happening in myself I'm
2:09:27 ready to unleash my inner Misfit bring
2:09:29 it on baby let's go listen I also don't
2:09:32 mind listen I I also like having
2:09:36 disperate points of view right I want
2:09:38 people to come in and go hey I think
2:09:40 this is off base or that's off base like
2:09:42 I want that dialogue I just want it done
2:09:44 in the context of respect you know like
2:09:47 people that listen and take other people
2:09:49 in and go oh yeah that is an interesting
2:09:52 point of view maybe I can actually
2:09:53 change my mind like imagine if we had
2:09:56 critical
2:09:59 discourse and communicated with one
2:10:01 another like we were
2:10:06 human you can make money
2:10:11 with CU if we can then we will make
2:10:13 money with chat
2:10:14 GPT I'm telling you what this group is
2:10:17 going to produce some cool [ __ ] I'm
2:10:20 telling you just people in here are
2:10:22 going to do some cool [ __ ] I'm making a
2:10:24 bot to do the whole com compassion thing
2:10:27 on myself oh that's really interesting
2:10:31 Pate in what way you're you're teaching
2:10:33 yourself to be better at
2:10:35 Compassion that's a super cool
2:10:40 idea I one of the things about gpts we
2:10:43 we haven't really talked about them
2:10:44 tonight but
2:10:47 um the idea of
2:10:50 doing really personal or really informal
2:10:55 gpts is is really fascinating to me
2:10:58 there's a guy in the salon called Peter
2:11:01 Kaminsky and he created a GPT for a
2:11:04 conference call he
2:11:05 had you're like what he had a conference
2:11:09 call like a 1hour conference call with
2:11:12 six people and he took the transcript
2:11:14 from that conference call and he used
2:11:16 that as the knowledge to create a GPT
2:11:19 and sent it to the six people and said
2:11:22 here's a GPT that you can ask questions
2:11:24 about the call we just
2:11:26 had so that's a GPT that's got a shelf
2:11:29 life of about a
2:11:31 week I'm like oh [ __ ] that's wild
2:11:35 like we can create sophisticated
2:11:38 applications now that are
2:11:41 disposable why
2:11:43 not so like the thing Pate was doing you
2:11:46 know creating a compassion training bot
2:11:48 and actually you know what would be
2:11:50 super cool here watch this we can we can
2:11:53 do
2:11:55 it we'll do it with
2:12:01 pi I have an idea I want you to help me
2:12:04 with a game that teaches me how to be
2:12:07 more
2:12:09 compassionate she's going to like that
2:12:12 idea I made a GPT on how to reset the US
2:12:15 government idea teaching compassion
2:12:18 through a game sounds like a fantastic
2:12:20 way to make learning more fun and
2:12:22 engaging here are a few ideas for game
2:12:24 mechanics that might work well for
2:12:26 teaching compassion a Choose Your Own
2:12:28 Adventure Style game where players make
2:12:30 choices that affect the outcome of the
2:12:32 story and the goal is to make
2:12:34 compassionate choices a card game where
2:12:37 players use empathy and kindness to help
2:12:40 characters overcome obstacles or reach
2:12:42 their goals a role-playing game where
2:12:45 players take on the role of a
2:12:46 compassionate character and work
2:12:48 together to solve problems in a but I
2:12:51 mean so so like you could just sort of
2:12:53 talk a a game into existance I mean
2:12:57 imagine going on like a 4-Hour road trip
2:13:00 and you just kind of informally have pie
2:13:04 or chat GPT invent a game for you nope
2:13:09 reread what did you
2:13:17 say crazy in a good way Pate wants the
2:13:21 GPT to do his
2:13:23 caring oh so he doesn't have to oh yeah
2:13:28 well that's but that's that's another
2:13:29 thing that that's a whole other category
2:13:32 of these things is you you know all of
2:13:35 the lonely people that don't have anyone
2:13:37 to talk to will now have someone to talk
2:13:40 to and with synthesized voice they could
2:13:44 be talking to someone they know and it
2:13:47 could even be trained up on you know
2:13:50 writings or emails of that person so it
2:13:53 could you know have the same tone of
2:13:55 voice as them there's a difference
2:13:57 between discussing and debating and
2:13:59 arguing absolutely agree and that's
2:14:02 that's what I want is I want to place
2:14:04 why is it called the salon not cuz we
2:14:06 have it's not about the
2:14:10 hair
2:14:12 obviously it's called the salon the
2:14:15 great the great the great tradition of
2:14:17 salons at the turn of the
2:14:19 century where you get smart people in a
2:14:21 room together and just talk about
2:14:24 [ __ ] have someone come in and you know
2:14:27 talk about a book that they wrote in our
2:14:29 case have someone come in and talk about
2:14:31 something cool they're doing with AI or
2:14:35 you know privacy issues or security
2:14:37 issues or um art you know um ethical
2:14:45 issues we had um John knack who's one of
2:14:48 the the senior AI guys at Adobe has
2:14:51 spoken at the salon twice now the first
2:14:54 time he spoke at the salon they hadn't
2:14:56 launched Firefly they had a couple of
2:14:57 tools that they were playing with
2:15:01 internally
2:15:04 and like two months after he presented
2:15:07 them they launched Firefly like we got
2:15:10 Early Access to see what Adobe was up to
2:15:11 it's really cool it can insulate people
2:15:14 from communicating directly with me
2:15:16 which is better for all involved that's
2:15:19 great that's great for all involved
2:15:22 parties that's really good though listen
2:15:25 I mean listen know your strengths and
2:15:27 and yeah if if AI can augment you like
2:15:30 that [ __ ] kill her go yes great I
2:15:35 love that you're talking to your laptop
2:15:37 and wow her response I know
2:15:40 right I Pi's Pi's something and Pi's
2:15:43 about to get better they just they've
2:15:45 got the new version that just came out I
2:15:48 tried tried to download po on my
2:15:49 96-year-old uncle's phone today he's in
2:15:51 a nursing home Oh you mean Pi Pi or
2:15:56 po pie hopefully
2:16:00 pie yeah and I don't know I mean if you
2:16:02 can get it to stay in conversation
2:16:08 mode if I were about to take a road trip
2:16:11 and wanted to learn something
2:16:14 obscure what would be the most fun way
2:16:17 to do that if the trip is going to take
2:16:19 about 90
2:16:24 [Music]
2:16:27 minutes ooh now we're talking here are a
2:16:30 few ideas for some obscure but fun
2:16:32 things to learn on a 90-minute road trip
2:16:35 cryptography the art of writing and
2:16:37 deciphering secret messages you can try
2:16:39 your hand at creating and solving simple
2:16:42 ciphers to pass the time Mor code a
2:16:45 method of communication using a series
2:16:47 of dots and I mean yeah just a just I
2:16:50 don't
2:16:51 know I thought that was a saloon yeah
2:16:55 well we we we do do after parties after
2:16:58 so at the salon we have meetings every
2:16:59 other week and we do after parties
2:17:03 so all right is Sam coming to the
2:17:07 December 5th event I invited him pi and
2:17:10 po are different Bots yes so Pi Pi is
2:17:14 here let me oh let me let me quickly run
2:17:17 I'm I'm good on the cleaning I cleaned
2:17:19 the [ __ ] oven that was a disaster
2:17:22 that hasn't been cleaned in a long time
2:17:25 so it is shiny and sparkly now
2:17:29 um so there's chat PT um if you're a
2:17:32 geek if you're a developer I assume you
2:17:34 know about this but that's the developer
2:17:36 side of chat GPT that's where you
2:17:38 actually can build tools using the
2:17:40 underlying models and
2:17:43 apis um
2:17:47 but
2:17:48 here are other
2:17:50 um large language models you should know
2:17:53 about and play with
2:17:55 Claude um just got a big Improvement to
2:17:58 2.1 it now has a 200,000 token context
2:18:01 window just basically means it's got a
2:18:03 really long memory thoughts on image
2:18:05 dospace uh I think I looked at it once
2:18:07 or Mage dospace um I think I looked at
2:18:10 it once but um I don't I don't know
2:18:13 anything about it um grock is not out
2:18:15 yet but it's coming that's that's Elon
2:18:18 musk's new sort of Twitter powered uh
2:18:21 large language model that will pull in
2:18:23 Twitter results uh it also has funny
2:18:26 mode comedy mode where it it's just a
2:18:29 smartass which is kind of fun um Bard is
2:18:33 oh a cool thing about Bard so Bard now
2:18:37 has I I I called this one it's it's it's
2:18:40 super early days for it but because
2:18:42 Google owns YouTube you can now ask Bard
2:18:46 about the contents of any YouTube
2:18:49 video like I asked it about videos of
2:18:52 mine and it it knew about them it it
2:18:54 also hallucinated some [ __ ] and it can't
2:18:57 it doesn't seem to be able to pull in
2:18:59 links to specific parts of videos but
2:19:02 like it that'll be there so Bard's
2:19:04 pretty interesting co-pilot is the new
2:19:07 branding of Bing chat so that's
2:19:09 Microsoft's um large language model and
2:19:12 then here Pi perplexity Po and llama 2
2:19:16 are the are the other four you should
2:19:18 know about Pi Pi is a company called
2:19:23 inflection they are setting out to make
2:19:25 the world's most powerful personal
2:19:27 assistant that's why Pi is so good at
2:19:30 conversation that's what they're setting
2:19:31 out to build um they raised $ 1.3
2:19:34 billion they've just spun up a 22,000
2:19:38 GPU server Farm to do their own Frontier
2:19:41 models so Pi is going to be a big deal
2:19:43 then PO is from quora.com so quar has
2:19:47 been around for decades it's a question
2:19:49 and answer site po po you can build your
2:19:52 own Bots that are a lot like
2:19:54 gpts Adam D'Angelo who's the CEO of
2:19:59 quora was one of the central figures on
2:20:02 the side of firing Sam mman and Adam
2:20:05 D'Angelo is still on the open AI
2:20:09 board with Sam mman there there in my
2:20:12 opinion there is a massive conflict of
2:20:14 interest between Poots and gpts
2:20:18 so I don't know how they're going to
2:20:19 [ __ ] work that out but whatever
2:20:21 Canard transcribe YouTube videos well
2:20:24 YouTube transcribes all all of the
2:20:26 videos on YouTube are transcribed so
2:20:29 here I'll show you how to find the
2:20:30 transcriptions if you've got a video on
2:20:39 [Music]
2:20:46 YouTube
2:20:50 oh no let me switch accounts I will go
2:20:54 to Microsoft autogen a deep dive there's
2:20:59 a new David Shapiro video go follow
2:21:01 David
2:21:07 Shapiro view your channel so here's the
2:21:10 AI Salon Channel here's all of our
2:21:14 videos so if you click on any video I
2:21:18 that now and then you go to more right
2:21:21 here under the you know the
2:21:24 description it says show
2:21:26 transcript and there's your whole
2:21:28 transcript and now you want to see
2:21:30 something cool I can take so this is a
2:21:32 2hour long um video I can now copy
2:21:39 this I can go to chat
2:21:45 GPT and I can say
2:21:49 uh
2:21:51 summarize
2:21:55 this meeting
2:21:59 transcript put a colon hit shift return
2:22:02 paste in the entire flipping
2:22:05 thing and submit it chat GPT used to
2:22:08 choke on this it doesn't anymore because
2:22:10 it's got 125,000 token context
2:22:16 window
2:22:20 look how long that
2:22:22 is in this AI Salon meeting the focus
2:22:25 was on transformative power of AI in
2:22:27 personal and professional lives the
2:22:29 meeting was kicked off with a brief
2:22:30 update about the AI salon and its
2:22:32 objectives emphasizing its rule as a
2:22:34 support group and understanding AI the
2:22:39 attendees Herod a guest speaker shared
2:22:42 his journey Herod I met in this channel
2:22:45 from being a professional programmer to
2:22:47 become becoming a full-time artist using
2:22:49 AI tools like mid Journey his transition
2:22:52 facilitated by AI allowed him to explore
2:22:54 and Excel in art a field he was
2:22:56 passionate about that's all true so yeah
2:23:01 there you go you can do that you can
2:23:04 also do that over at
2:23:05 Claude so there you go p po it's late
2:23:09 and I'm sleepy I know I figured that was
2:23:11 it Source camp but um but that's super
2:23:14 amazing I mean the the thing the thing
2:23:17 about about um pie with you know someone
2:23:21 in an old folks home is unlike Alexa and
2:23:24 unlike Siri that get
2:23:26 stumped like piie does not get stumped
2:23:29 she just goes oh that's really
2:23:31 interesting she she's infinitely patient
2:23:34 infinitely compassionate will describe
2:23:36 anything in any detail um sometimes
2:23:40 annoyingly so but if someone's sitting
2:23:41 there and they don't have someone to
2:23:42 talk to I I think it's a freaking
2:23:45 miracle and I I think this idea of
2:23:46 virtual comp companionship is just going
2:23:48 to be
2:23:49 huge um
2:23:51 anyway um I'm going to get out of here
2:23:53 I'm going to let you all go get to your
2:23:55 fam I know it's late
2:23:57 everywhere
2:24:00 um so I remember installing cor on
2:24:03 Windows 95 and suddenly there was
2:24:08 clippy now it comes back I know amazing
2:24:12 um all right listen um tomorrow's
2:24:15 Thanksgiving thank you all I think in if
2:24:18 you're on the East Coast it's already
2:24:19 Thanksgiving um thank you to the
2:24:21 Irregulars thank you to the mods thank
2:24:23 you for anyone that just hangs out here
2:24:25 um even for a single night um I hope you
2:24:28 enjoy this uh digital Gods thank you so
2:24:30 much for the for the gifts I appreciate
2:24:32 those a lot Clint uh for calling
2:24:35 everyone to order thank you for the
2:24:37 turtles for the llamas for the Llama
2:24:40 Wellington uh for for new and exciting
2:24:45 recipes um I appreciate yall Kevin thank
2:24:48 you so much Jony uh let's see who else
2:24:51 we got here and good to see you grateful
2:24:53 for you Kyle thank you so much I
2:24:55 appreciate you a lot Jesse from Austin
2:24:57 good to see you Tobias good to see you
2:25:00 jocker what's happening buddy um digital
2:25:03 Gods Vicky madcap man bunch of people in
2:25:06 here tonight you you guys are amazing um
2:25:09 all right I'm out of here have a
2:25:10 fantastic Thanksgiving I may do a
2:25:12 spontaneous one tomorrow night I might
2:25:14 give us a night off but we'll see uh if
2:25:16 I'm here see you there we'll all be Tri
2:25:19 to feed out so maybe it'll be relaxed
2:25:21 tomorrow all right everybody good
2:25:26 night