May 17, 2024 - AI's Future: Exploring GPT-4 Omni and Its Impact on Creativity

Description

Chapters

Transcript

0:00 [Music]
0:10 [Applause]
0:16 [Music]
0:23 [Laughter]
0:27 oh
0:31 oh
0:32 [Music]
0:54 [Music]
0:58 [Applause]
1:01 [Music]
1:07 [Applause]
1:12 it's not simple to
1:14 say most
1:17 days I don't recognize me that these
1:21 shoes and this apron that place and its
1:25 patrons have taken more than I gave them
1:31 it's not easy
1:33 know I'm not
1:36 anything like I used to be although it's
1:40 true I was never attention sweet Center
1:45 I still remember that
1:48 girl she's imp perfect but she
1:52 tries she is good but she lies she is
2:00 hard on
2:02 herself she is broken and won't ask for
2:07 hell she is messy but she's
2:11 kind she's
2:14 lonely most of the time she's all of
2:18 this Pi up and baked in a beautiful
2:23 pie she is gone but she used to be mine
2:30 [Applause]
2:32 and it's not what I ask
2:35 for sometimes life just lifs the
2:41 backs out a person makes you believe
2:45 it's all
2:47 true and now I've got
2:50 you feel not what I asked for if I'm
2:55 honest honest I know I can give it all
2:59 up for a chance to start over rout and
3:03 ending a
3:05 two good evening good
3:08 people Friday night date night sing a
3:12 little Sarah barelis to get us started
3:14 off but I forgot that one chord in
3:17 second second half of the song Oh well
3:21 [Music]
3:32 um there's a lot going on there's a lot
3:35 simmering right
3:36 now so we'll talk about
3:39 that you can tell me about [ __ ] you're
3:41 working
3:42 on we can uh dig into stuff you rock
3:46 Kyle thank you so much I appreciate
3:48 that I thought you only knew four songs
3:51 that was good oh thank you very
3:54 much I appreciate that
4:00 that's a fine fine song
4:03 [Music]
4:33 [Applause]
4:34 Mr it yesterday I almost died in the big
4:37 storm that hit Houston yikes I hope
4:39 you're okay yeah I saw I saw footage of
4:42 people in a store like I couldn't tell
4:45 like where they were and then you can
4:46 just see crap flying at the big glass
4:48 windows I guess it was a tornado hit
4:50 downtown Houston so I'm glad you're here
4:53 I'm glad you're okay yikes
4:59 [Music]
5:11 [Music]
5:20 let us be lovers can we redo last night
5:23 we'll carry our fortunes together
5:30 I've got some real estate here in my
5:35 bag so we bought a pack of
5:39 cigarettes and Mrs Wagner's
5:42 pies and we've all all come to look for
5:47 America
5:49 [Music]
6:05 Kathy I said as we ped the gound in
6:11 Pittsburgh Michigan seems like a dream
6:14 to me
6:14 [Music]
6:18 now it took me
6:23 it I don't remember all
6:27 right we need an intro request line
6:30 Early Morning
6:33 Rain yeah I could do tabs I could get
6:36 out tabs and actually play different
6:37 songs
6:40 but that's like reading the
6:43 manual just watched a quick YouTube on
6:46 LTX it explained a lot simply great um
6:50 I'm trying to think if there's an easy
6:51 way you know what Silver Fox share that
6:54 in um why don't you share it in the
6:56 video Guild and then you can also you
6:59 could share the the link to the video
7:01 Guild Post in the Irregulars but why
7:02 don't you share that uh link to LTX in
7:06 video Guild and then double post it or
7:10 or share the link in your
7:14 regulars we're all Gilligans you're the
7:17 captain I think I'm I'm more Gilligan
7:19 than the
7:23 captain Oh man
7:26 [Music]
7:34 I had a good day today I
7:37 um so I wrote this book or chapter in a
7:42 book an
7:43 [Music]
7:47 anthology called AI
7:49 Futures put together by lab five Futures
7:53 Cindy [ __ ] she was the
7:56 instigator and today we launched
8:01 a GPT for the book and it's it's really
8:04 cool
8:07 um and it I like I've been thinking a
8:09 lot lately
8:12 about as these tools become more capable
8:15 and more integrated in what we
8:21 do you know do you do you
8:24 redefine things like what is a book
8:30 does it make sense to read a book
8:31 linearly some some it
8:35 does if you have linear narrative but if
8:38 you've got an anthology of seven
8:40 different chapters that while they
8:42 follow an
8:44 arc you could read them in any
8:48 order and then the other thing that we
8:50 did with this GPT was we did we designed
8:53 a set of structured interviews oh lab
8:56 five future Futures is here syndic is
8:59 here inst Ator
9:02 present we did we designed a set of uh
9:05 well no we designed a single interview
9:07 like 20
9:11 questions some of the questions were
9:13 about the author and some of the
9:14 questions were about Ai and the future
9:16 and thoughts on the future so we took
9:19 the transcripts of all those
9:21 interviews and we uploaded those so
9:23 we've got the the full book as a
9:25 knowledge base and then we've got a
9:28 little mini bio for from each author and
9:31 then we've got this interview from each
9:32 author all of that's in the knowledge
9:35 base and so the
9:39 the the GPT you can explore what the
9:42 book is you can
9:43 explore you can have chats with the
9:46 authors it's cool it's pretty slick I'll
9:48 show it to
9:48 [Music]
9:53 you and you know as I was building it
9:57 it's funny as as uh
10:00 because the GPT store sucked and because
10:04 only paying members could use gpts
10:07 before it kind of chilled me on using
10:12 them but now that chat GPT 4 Omni is
10:16 going to be available to
10:19 everyone and gpts are going to be
10:22 available to
10:24 everyone
10:26 um I'm kind of excited about gpts again
10:29 posted in video Guild great thanks
10:31 Silver Fox I'll go show
10:36 it I if I'm if I'm clearing my throat a
10:39 lot I was eating almonds before I came
10:41 on and you know how you sort of half
10:44 swallow something and it's sort of
10:45 caught in your gullet I had I had a half
10:48 chewed up almond half stuck in my
10:52 gullet it's like it's just like I'm sure
10:55 I've got almond shards all throughout my
10:58 my My Talking area my talkie
11:02 bits every time I see Kyle on my wallet
11:06 starts to cry and SOB well no but gp4 is
11:11 going to be free
11:15 now oh man all right
11:23 [Music]
11:31 every time I see it now get that looking
11:36 M every time I see your mouth I hear
11:41 that
11:43 smile the early M morning L I heard the
11:48 engine
11:49 turn and all F
11:53 [Music]
11:58 outside and alond skin stuck to your
12:00 throat been there you are leaving
12:05 me again
12:08 today you will convince
12:12 me again today Joy py one of gpts is
12:16 getting traction very
12:19 cool hope they aren't going the way of
12:21 ebooks which are extremely expensive for
12:23 public libraries oh interesting but
12:26 you're living this hotel looking for
12:29 someone else's Golden
12:34 Ring should have
12:37 say so long
12:39 [Music]
12:42 suzan doter doer how's Frank doing these
12:45 days Frank hasn't been around Frank is
12:48 very temp you know what we learned about
12:49 Frank [ __ ] Diva he's like hey if I'm
12:53 going to make a if I make a video I make
12:55 a video if not what are you going to do
12:56 [ __ ] boil me you do that anyway assho
13:00 so Frank is you know he's tough to work
13:04 [Music]
13:20 [Music]
13:23 with it's about game codes oh that's a
13:26 smart one Joy pie
13:31 [Music]
13:56 oh
14:00 um Genie and bottle can you show me
14:02 where gpts are now that we have GPT for
14:05 Omni yeah I can they kind of move some
14:07 stuff around in the interface gpts are
14:10 still mostly in the same place but one
14:12 thing I've noticed is that the interface
14:15 is really weird I I think someone
14:19 mentioned it might have been Joker
14:20 someone mentioned that they think that
14:23 gp4 omni's got memory
14:26 issues that it's like causing browsers
14:29 to flake
14:30 out and I've noticed my
14:33 um my i chrome didn't used to crash a
14:36 lot but it crashes a lot
14:39 now tap good people thank you jeie yeah
14:43 go ahead tap the screen uh share the
14:45 share the live it's Friday night date
14:47 night let's get some people in here if
14:48 you don't know who the hell I am my
14:49 name's Kyle Shannon this is the AI
14:51 learning lab that's a that's a image
14:53 from Danielle I'm pretty sure it's
14:55 Danielle right yeah Danielle it's
14:57 Danielle it's Danielle wasn't it oh I
15:00 didn't name it well I'm pretty sure it
15:01 was
15:02 Danielle
15:04 um little Pinball Wizard action there
15:07 which is kind of
15:08 cool I remember those days well from the
15:12 80s the
15:14 arcades steal a roll of quarters from
15:17 the from your best friend's parents head
15:21 on out to the mall come on it's not to
15:24 love
15:33 memory issues must be
15:35 [Laughter]
15:37 Jack uh Kyle has a daily Walkin talk a
15:40 nightly tick tok a business to run and a
15:43 wife's feet to rub it's true and a dog
15:45 to give chicken sticks
15:48 too but I have fidget toys
15:53 now it's a cube oh now it's not h h
15:59 look at
16:01 that if you want cool 3D printed stuff
16:04 Vicky Baptist who's in here a lot oh
16:06 there she is something caused my whole
16:08 computer to flake out it's probably it's
16:11 it could maybe it's gp4 maybe it's flaky
16:13 maybe that's why Ilia left he's like
16:15 this thing's too buggy I'm out of here
16:17 that could be
16:19 it it sounds like what's coming out
16:22 about Ilia and and
16:24 Yan leaving who those were the two leads
16:29 of the AI super alignment group within
16:33 chat GPT within open
16:35 Ai and it looks like there was a
16:40 battle and Sam won and the battle was do
16:46 we push
16:47 forward and accelerate getting to AGI or
16:50 do we prioritize
16:52 safety well the safety team just
16:56 left so
17:00 apparently we're going full steam ahead
17:02 toward
17:05 AGI Damn the Torpedoes as they
17:08 say um I don't know if that's all been
17:12 confirmed but that was sort of the the
17:14 Scuttle butt I saw
17:19 today only thing you can do is refresh
17:21 the screen as chat GPT 4 locks up a fair
17:24 a bit yeah that's what I noticed is that
17:25 it it like like the screen a load but
17:30 like gpts aren't there and then if I
17:33 dick around with it they'll show up
17:34 eventually so yeah I think I think
17:36 there's some I think they got some some
17:38 cleaning up to do on the
17:40 old on the old memory um I'm trying to
17:44 think where we want to start tonight oh
17:45 another thing that I'll talk about um
17:47 Jared pus our our fine Governor here in
17:50 the state of Colorado signed that shitty
17:53 AI bill into law tonight well into it's
17:55 it's not it doesn't take effect until
17:58 2026 and and he and he signed it with a
18:00 signing statement that basically said
18:03 this needs some work work on it over the
18:05 next two years let's make sure we get
18:07 this thing to not suck before it goes
18:10 into law or before the federal
18:12 government does something to preempt
18:14 it so that's
18:17 disappointing um because it it it really
18:21 is a steam and pile of [ __ ] Bill and it
18:23 it would have been better to start over
18:25 but whatever it's signed now so um so I
18:28 suppose when the session goes back into
18:32 session in I think it's November some
18:35 [ __ ] like that they'll probably call us
18:38 up again and go hey you guys were a
18:40 little cranky about this you got any
18:43 ideas let's go ahead and make it better
18:45 shall
18:51 we I'm too cynical for politics I don't
18:55 like it cuz I knew they were going to
18:57 sign it I knew they were going to pass
18:58 it when I talked to the guy that the the
19:01 Senate Majority Leader I got a chance to
19:02 talk to him when when we
19:05 testified look ao's wife I'm using my
19:09 coasters
19:10 um I talked to him after we testified
19:14 and he you know he basically said
19:17 yeah we kind of got our asses handed to
19:20 us with the data privacy stuff and we
19:21 were way late to the party and we still
19:24 haven't passed anything on that so we're
19:26 going to do something
19:32 so see like is it
19:34 me you shouldn't be passing
19:38 laws just cuz you want to go first right
19:42 right but
19:45 again what man does not understand he
19:48 fears what he fears he
19:51 destroys yeah well see if it you you
19:55 know what's funny wind
19:57 rider if it were that
20:00 if if he were if he were like AI is evil
20:03 and we're going to we're going to
20:05 protect the people from this evil
20:07 thing I I I would have more respect for
20:10 the
20:11 decision because at least then it's
20:13 coming from a place of a clear point of
20:16 view might be misinformed but it's at
20:18 least a clear point of view this was
20:20 really more just like yeah I know it's
20:22 not perfect but we'll we'll we'll
20:24 stretch the the effective date out a
20:26 couple of years and you can amend it
20:32 that doesn't sound like it should be a
20:34 law does
20:39 it seems like seems like it just it's
20:42 just it's just a law that says AI on it
20:45 and it says AI consumer
20:48 protection anti-bias there's like all
20:50 the buzzwords all the buzzwords are
20:52 right there in the title of it the
20:54 details of it it's not important we'll
20:56 fix that over the next two
20:57 years so it's disappointing but what are
21:00 you going to do all right make a
21:02 governor GPT P based on Kyle's knowledge
21:05 yeah yeah my my sketchy knowledge about
21:08 one bill one time in my life in 58
21:12 years 59 I'm 59
21:15 now I got 59 years and one political
21:20 politically active moment of my life in
21:22 the past month I'm [ __ ] done what
21:25 about the constituents exactly what
21:28 about them
21:33 hopefully they put together a working
21:35 group and they're actually opening open
21:37 to amending that steaman pile of poo but
21:39 what
21:41 whatever nothing we can do about it now
21:43 it's signed it's it's history as they
21:46 say it's history it's in the books it's
21:49 in the books it's in the books okay okay
21:53 okay all right Tobias hello good day how
21:57 do you like four o so far
22:02 I I don't hate
22:05 it
22:08 um my problem with 40 right now is it's
22:12 got a really shitty name and it's been
22:14 half rolled
22:15 out and what's what's weird and
22:18 frustrating is most people like even a
22:20 lot of people I know that are really
22:22 into this AI stuff and really pay
22:24 attention to this don't actually
22:26 understand what it is what's not there
22:28 why it's different why it's a big deal
22:30 we can talk about that tonight if you
22:32 want to I'm happy to talk about it um I
22:36 think it's a very very big deal and I
22:38 think almost no one knows
22:44 that because it's not a big deal yet
22:48 it's like the it's like the seeds of a
22:50 big
22:52 deal [ __ ] name yeah it needs it should
22:55 be called GPT
22:57 Omni um are there any tools for 40
23:00 rolled out yet
23:03 no do you know if any of our custom gpts
23:06 are using 40 I don't think so [ __ ] name
23:10 bad marketing still no
23:12 lvx email still on wait list ug sorry
23:16 about that Allan did you were you here
23:19 when we did the the form swarm where we
23:21 jumped on the L LTX server
23:24 Discord I think you must have missed
23:27 that if if you filled out that form you
23:30 should have
23:31 it but you but you have to have filled
23:33 out the form so so the weit list and
23:35 being in Discord when they open up a
23:38 form are two completely different
23:40 activities if you're on the wait list
23:43 that seems to just be taking forever
23:45 let's go let's go let me run over to
23:47 Discord for a second and see if they're
23:49 saying
23:52 anything three things is beyond my
23:54 memory capacity oh I was just saying
23:56 Lori just post it grab a link to the
23:59 post that you did in the in the video
24:01 Guild and just post it in irre in
24:03 Irregulars Channel just so the
24:05 Irregulars have it but it's not that big
24:07 a deal if it's posted there it's posted
24:09 that's good um let's see
24:17 announcements LTX is having hey um Cindy
24:20 [ __ ] Cindy [ __ ] paging Cindy [ __ ] lab
24:24 five futures um go look at the
24:27 announcements in the LTX Discord they're
24:30 doing a blue
24:32 contest great news we're launching the
24:34 blue contest with an incredible prize
24:36 for the winner all expenses paid trip to
24:38 London and an invitation to our special
24:41 event on June 5th capture the essence of
24:44 blue through your
24:46 lens seems like something you could
24:49 do create a video post your video submit
24:51 your entry read the
24:54 rules I am here yeah uh look wait um
25:00 there
25:02 see LTX Studio contest you can get a
25:05 free trip to London woo and you can go
25:08 to their
25:09 event and see version
25:12 2.0 where hopefully they add better
25:15 image generation tools into that piece
25:18 of
25:20 poop um Cindy [ __ ] is is like all over
25:23 LTX by the way she's loving it she's
25:26 she's a filmmaker now but you know
25:28 what's funny it's it's like she was
25:29 saying on one of the calls we were on
25:31 together today that you know she's been
25:33 wanting to do like movie making stuff
25:35 for like a lot of her life and now all
25:38 of a sudden here's this tool that lets
25:40 her do it every time she went into
25:41 something like Final Cut Pro or Premiere
25:44 or some other tool would come out she'd
25:45 just you know go apoplectic and go this
25:48 is too
25:49 complicated and so now there's a
25:55 tool oh shards of almonds still still
25:58 still clinging to my
26:01 throat applied at 300 p.m. today bro I'm
26:05 on that [ __ ] nice beautiful
26:09 congratulations all right let's see if
26:12 they mention when they're going to do
26:13 last batch closed today make sure you
26:16 share your experience on feedback and
26:20 ideas uh uh uh
26:22 uh yesterday at 11:00 a.m.
26:26 yeah they might do something Sunday you
26:29 just got to pay attention to the
26:30 announcements channel in there we should
26:34 build a submission oh you know what why
26:36 can't I read my why can't I read the
26:38 letters cuz I don't have my black
26:41 construction paper over them look isn't
26:43 that better you're
26:44 [Laughter]
26:48 welcome we should build a
26:51 submission on the live oh we could do
26:54 that that's not
26:56 bad all right yeah yeah let's do that
26:59 that's a good idea Brandon all right
27:01 we'll we'll we'll make original music
27:03 for it I'm hot hang on let's open Yan
27:07 window was a little warm today it was
27:09 like 83 today in
27:13 Denver do you need to be an actor to
27:16 make movies in AI you you don't you can
27:19 you barely need to be able to form a
27:22 sentence now the definition of what's a
27:26 movie the the movies that that these two
27:28 make if you don't actually treat them
27:31 like like storytelling it just it just
27:34 creates really bad content so what's
27:37 going to happen with tools like LTX is
27:40 initially there's just going to be a lot
27:41 of bad content generated from them and
27:44 then there's going to be real
27:45 storytellers like Cindy and you know
27:47 people that get their hands on them if
27:49 we do some in here we'll do some real
27:51 storytelling that are going to be doing
27:53 good things um but it's going to be it's
27:56 it here's the deal with with all
27:58 creative tools and all creative
28:00 outputs 90% of it is
28:04 drek
28:08 90% 8% of the 10% is
28:12 decent
28:14 2% is really good well no 1.5% is really
28:19 good and like 05% of any creative output
28:22 is actually the [ __ ] that you know is
28:24 good and breaks through the
28:26 noise and that's that's how it's always
28:28 been and it's how it will always be no
28:30 matter what the tools
28:31 are so someone's going to end up being
28:34 like the best AI music producer in all
28:36 of the land and they're going to be
28:38 [ __ ] badass and they may or may not
28:41 have been a musician
28:42 before right it's just like Hip-Hop
28:45 Hip-Hop was a it's just a totally
28:47 different
28:49 discipline and you know musicians before
28:52 that were like that's not real music
28:54 you're just looping loops and samp
28:59 and they bitched and
29:02 moaned and now it's the hip hoppers that
29:05 are bitching and moaning about the AI
29:07 music it's it's all just Cycles people
29:11 it's all Cycles new technology fear rage
29:17 resistance accept denial
29:22 acceptance right so 3 four years from
29:25 now it'll all just be music again but
29:27 right now we got we got fights to fight
29:30 all right YouTube link for easy LTX to
29:34 tutorial is now in Irregulars okay so
29:36 here's what we're going to do how many
29:37 folks are in here 82 people if you're
29:39 new
29:40 here this group this
29:43 organization is a community that I
29:48 started the week after chat GPT came out
29:51 with Leah faston who's a professional
29:53 photographer out of Boston and we
29:56 started this cuzz we we we knew knew
29:58 that this [ __ ] was coming we knew it was
30:00 coming fast we knew it was going to
30:01 change things and we knew that we didn't
30:03 KN know [ __ ] anything and we also
30:06 knew that there were going to be other
30:07 people trying to figure this stuff out
30:09 that also didn't know anything and so
30:10 this
30:12 group is that it's it's a group that's
30:15 about play
30:17 playing playing focused play oh that
30:20 might be the term hang
30:23 on I got to share
30:26 something so go there and then click on
30:29 the first link that says
30:31 the
30:33 community and go read about the salon
30:36 for a
30:52 second
30:56 okay so focused play
30:59 this is about people that are like
31:00 trying to figure this stuff out and
31:02 they're doing it in this adventurous
31:04 playful kind of
31:06 way
31:08 and it's really remarkable so if you go
31:11 in
31:17 there you'll see a nice little section
31:19 about who we are and what we do why we
31:22 do it our values you should go check
31:24 those out make sure you resonate with
31:26 those
31:30 and
31:31 then there's a community area where you
31:34 can and should introduce yourself it's
31:37 called introduce
31:39 yourself so you should go do that
31:43 um if you want to know why I think
31:46 Community is probably the single most
31:48 important thing that you can commit to
31:50 over the next 3 years go watch my latest
31:53 Tik Tok video I I filmed it this morning
31:55 talking exactly about this and it'll
31:58 become Crystal Clear why you should use
32:00 your real name in here and why you
32:02 should introduce yourself to people and
32:03 get connected in this community true or
32:05 false the announcements this week were
32:08 relevant to AI beginners very true an
32:12 Murphy um very true because the the
32:16 biggest thing for AI beginners is that
32:19 open AI is making their most powerful
32:22 model free to everyone the other reason
32:26 it's relevant is in about 2 or 3 weeks
32:28 they're also going to release
32:32 um their new voice
32:35 component where you talk to these things
32:38 and it's literally like liter literally
32:42 like the movie
32:44 her so if you haven't seen the movie her
32:47 I think it came out in 2018 or something
32:48 like
32:49 that
32:52 2014 it's the one where Walkin Phoenix
32:55 falls in love with this chat bot named
32:59 Samantha Scarlett Johansson scarjo
33:04 baby and actually I saw I saw a Tik Tok
33:07 today that someone um compared the
33:11 Scarlet Johansson Samantha voice with
33:14 what what open AI demoed it's identical
33:18 it's it's identical but anyway they made
33:20 that so so what does that mean what it
33:23 means is the these tools are going to
33:26 get incredibly
33:29 intuitive this whole idea of having to
33:31 learn prompt engineering and learn
33:32 prompting and things like that I think
33:35 that that if if you want to do that
33:38 there's going to be plenty of places you
33:39 can do that um but these tools are going
33:43 to get very very intuitive and and very
33:46 fluid and there's going to be very
33:49 little friction between you and these
33:52 incredibly powerful tools so yes I think
33:54 what they announced today is incredibly
33:56 important
34:04 I'm checking
34:12 now 3D object synthesis oh
34:25 interesting what's this thing is that
34:27 the link
34:33 all
34:36 right dang it I still have to do my her
34:39 homework my Herm
34:41 work you do you really do you seriously
34:44 if you haven't seen the movie her
34:47 recently you should see it
34:51 because a the technolog is here not
34:54 coming here it's just not been launched
34:56 yet um
34:58 and Sam wman has said that you know that
35:02 movie was incredibly prophetic about the
35:05 kind of issues that it raises and the
35:07 kind of how people are using computers
35:11 in that movie he he says he sees that's
35:13 how it's going to
35:14 be um and that it's very inspirational
35:17 to the team that that built the thing
35:20 that they announced on Monday um so if
35:23 you want to see what's coming and if you
35:25 want
35:26 to it's it's weird to watch a science
35:29 fiction movie that was fullon science
35:32 fiction just a few years
35:34 ago knowing that we're living in that
35:38 reality right now it's it's weird it's
35:41 weird it's it's it's creepy okay
35:46 um so there's a couple other things
35:49 within within the AI Salon you should
35:51 introduce yourself um two little links
35:54 below introduce yourself is this thing
35:56 called AI Launchpad and tools this is a
35:58 page that I maintain that technology is
36:01 definitely not here yet not to that
36:02 degree no it's not it's you're you're
36:04 right Pate the the the conversational
36:07 part of it's here the action taking
36:08 action part of it is not here that's a
36:11 that's a really good point um this list
36:13 that I maintain the AI launch pad and
36:15 tools is um just a list of if if you're
36:19 trying to figure out what are the tools
36:21 I need to get my head around um I tried
36:24 to organize things here I start with
36:26 chat GPT that's where you should start
36:28 in my humble opinion and then here's a
36:31 bunch of other
36:32 um large language models here's a bunch
36:36 of image creation tools um video
36:39 creation tools Music Creation tools all
36:42 of them especially once you get into the
36:44 multimedia stuff any anything that's
36:46 sort of time based and and even the
36:48 images to a degree all of these tools
36:50 have very different personalities and
36:52 one of the things I've been thinking
36:54 about a lot lately is I may love a tool
36:58 like Deco
37:00 here this tool that has this thing now
37:04 called stations where you don't have to
37:06 prompt it it just every second generates
37:09 an image and it's just like this
37:11 NeverEnding Cascade of images you don't
37:13 have to prompted at
37:18 alloo bloop
37:21 blop
37:23 right I may love this you may go to this
37:26 and go well I've got no control I
37:28 [ __ ] hate this I want I want you know
37:30 like if I pick that image and I'm like o
37:33 create a station now out of that
37:35 image and now it's just going to create
37:37 a whole new set of images based on this
37:40 image as the starting
37:45 seed oh that's cool like let's do oh
37:47 steampunk lamps let's go see a bunch of
37:49 those that's super
37:51 cool someone else may be very into
37:55 something like Leonardo oh excuse me Leo
38:06 Nardo which is like the cockpit of a 747
38:11 introducing content and multi-style
38:13 reference oh
38:16 nice Leonardo is getting slick look look
38:19 here's these are all different tools
38:21 within Leonardo image generator canvas
38:23 editor realtime generation motion
38:26 Universal upscaler so to image
38:28 generation you go in here look at all
38:29 these switches and dials you can turn on
38:32 Photo re V2 turn on Alchemy V2 prompt
38:36 magic transparency public images image
38:39 Dimension like it's just it just goes on
38:41 and on and
38:43 on you can make your own models
38:47 um right and so someone else is going to
38:49 love this because they're they're going
38:50 to love the control and the ability to
38:52 do all that crazy [ __ ] so okay chat GPT
38:56 4.0 or chat
39:00 40
39:01 yes
39:05 um yeah let me talk about
39:08 that so so the tools are
39:12 getting there's a lot of them oh let me
39:14 go let me sorry I got distracted from
39:16 where I
39:18 was talking about the AI Salon okay so
39:20 within the AI salon so you should
39:22 introduce yourself check out the AI
39:24 Launchpad and
39:26 tools um there's water cooler discussion
39:28 there's a news section there's an ask
39:30 for help if you want to ask for help you
39:31 can do that there's a showand tell area
39:34 for people to show off you know specific
39:36 multimedia stuff and then there's guilds
39:38 and so the guilds are little
39:39 subcommunities
39:40 um led by different people so everyone's
39:43 doing sort of focused play in Creative
39:46 areas business areas tech tech guilds
39:50 and social impact guilds and then way
39:53 down here at the bottom is a community
39:54 club called a AI learning lab regulars
39:58 that's us so we have our own little
40:02 area and let's see where did Silver Fox
40:05 so so Silver Fox put a simple LTX
40:10 YouTube
40:12 tutorial and why we put an LTX tutorial
40:16 in there is the other
40:18 night we all went um and we crashed
40:22 their server when they were given out a
40:23 bunch of licenses and we got like 30 or
40:26 40 Irregulars LTX
40:28 licenses so so if you got one of those
40:32 LTX
40:35 licenses thank you Silver Fox for
40:37 putting together a sweet little tutorial
40:39 on how to use the tool I had the chance
40:43 to use LTX Studio as a beta tester and
40:46 I'm going to film they have different
40:48 you can down or change that's cool they
40:51 did it as a little character FM here
40:54 nice that's
40:56 cool okay scroll down in the regulars
40:59 please yes sir yes
41:03 sir GD Bell what's this spring flowers
41:07 in my backyard look actual
41:10 photography it's
41:12 heresy she put an image of the real
41:15 world in the AI Salon it's heresy run
41:18 away run away she's a
41:21 witch what other than what other than
41:26 Ducks FL
41:29 uh very small
41:30 [Laughter]
41:37 rocks burn her burn
41:43 her oh that's cool today oh politics in
41:48 the Congressional hearing about her
41:50 eyelashes that was just trash can we get
41:53 back to like you
41:56 know share a cool thing you do oh that's
41:59 cool that's look VI Vicky Baptist also
42:01 doing actual
42:04 photography very
42:06 cool Lord digital
42:14 Gods all right oh yeah if you want to um
42:18 if you like that Deco here thing this
42:22 thing not that version oh wait
42:30 if you scroll down in IR regulars this
42:32 this will take you
42:33 to that original um what's called a
42:41 station and if you go there if you click
42:43 on that right now we'll we'll we'll be
42:45 seeing the exact same art so the way
42:47 these stations work is if you go to this
42:49 same station like you'll see that thing
42:52 or you'll see you know wherever you join
42:54 join in the party you'll see the same
42:57 stuff I'm seeing which is pretty
42:59 cool all right oh wait it was AI
43:03 learning lab chat oh all right what's
43:05 over in chat what do you got in chat
43:07 dear Bri Bri Bri Bri listen I'm
43:14 hot almond
43:17 shards did you guys see that Scotty
43:20 Sheffer got [ __ ] beat up and and
43:24 taken to jail today before he start in
43:27 his second round of the PGA Championship
43:31 unbelievable world's number one golfer
43:35 and the cops didn't know him and
43:36 arrested him roughed him up and put him
43:39 in cuffs and took him to
43:43 jail it's crazy I tell you it's crazy I
43:47 tell
43:49 you Kyle Kyle F flowers bypassed Google
43:53 and Microsoft storage by
43:55 using open AI
43:58 anyone entering the LTX contest in fact
44:04 um Cindy [ __ ] lab five Futures already
44:08 submitted a submission today and we're
44:10 thinking about doing one we could build
44:12 one tonight and submit it he ran a
44:15 police
44:17 barricade oh so maybe maybe he wasn't in
44:19 the but still like even if he ran a
44:21 police barricade wouldn't wouldn't did
44:24 he not go like uh you know I'm like
44:27 Scotty Sheffer right you know I'm the
44:29 defending Champion here right
44:31 anyone maybe he didn't say that maybe he
44:34 just you know acted like a dick
44:37 and I don't know man cops these days
44:40 they're they're wound a little tight
44:48 so all right so what you should do
44:51 tonight if you generate
44:55 things um if you make music or if you
44:58 make images or if you make movies or
44:59 anything like that or if you want to
45:01 just make memes of stupid [ __ ] I said or
45:03 like I have almond shards in my
45:05 throat just share those in The Irregular
45:08 share cool things Channel and then we
45:10 can all see them so it can become kind
45:12 of like a jam session what's fascinating
45:14 about the jam sessions they used to only
45:16 be um images and now they're all sorts
45:18 of multimedia because the tools are
45:20 getting getting crazy good all right
45:23 pedestrian was hit by a car and police
45:25 were investigating but he ran through
45:28 the barricade oh well that's pretty
45:30 shitty behavior isn't
45:31 it so are the non- cops yeah I know
45:34 every everyone's so found wound so tight
45:36 aren't they cops hate that do know do I
45:39 am thing it's true it's a good point but
45:42 in this case he's at the PGA tournament
45:45 he's the defending Champion he's the
45:46 number one golfer in the world the do
45:49 you know who I am in that context is
45:51 actually
45:52 relevant but if he [ __ ] up a crash
45:55 scene then yeah [ __ ] him if he did if he
45:58 did something shitty then that's on
46:02 him all
46:07 [Music]
46:16 right um what are we going to do what
46:18 are we going to do what are we going to
46:20 do when you want to do people hey hey ky
46:24 ky Kyle yeah you can make money with ch
46:29 how do you make money with cha cow cow
46:32 cow oh someone was asking
46:34 about can we play with globe.
46:38 engineering what do you want to play
46:40 with there the uh this their search
46:42 engine thing there's not a ton to play
46:44 with there unless it's has it gotten
46:46 better recently cuz it's not bad it's
46:49 pretty cool but it's kind of one one
46:50 Trick
46:51 Pony
46:53 um open image wait open image text
46:58 explanation in Irregulars
47:01 what
47:04 what making videos with AI as if it gave
47:08 me some horrible sexualized women first
47:10 after three refusals and telling it to
47:12 stop doing that I got
47:19 this a filmmaker with video equipment
47:22 resting on her
47:24 shoulders that's pretty funny
47:29 she's using the latest in Mac 660 AV
47:36 technology that's
47:38 horrible
47:43 oh it's an old iMac
47:45 computer duct
47:47 [Laughter]
47:50 taped oh yeah well so so here's the
47:53 thing Corey the the uh the image model
47:58 that LTX uses it's some shitty like
48:01 non-optimized version of stable
48:03 diffusion I think it's just it's just
48:04 they just using a shitty model right now
48:06 I'm sure that that's going to get better
48:09 but the writing is also really shitty
48:11 what what Cindy [ __ ] is found with LTX
48:15 is absolutely do not let LTX make the
48:18 story make your story in chat GPT and
48:22 then copy it over to
48:24 LTX yeah that's that's really bad that's
48:27 really
48:29 bad I think that's just further evidence
48:31 for what Cindy's saying don't don't
48:33 screw around with uh don't screw around
48:38 with with
48:40 uh making the story in LTX studio
48:46 studio Silver Fox has had it with bias I
48:50 know I know I
48:54 know it's all in there like
48:56 unfortunately I mean this is
48:59 unfortunately the the what these tools
49:03 are is a reflection of what's been on
49:05 the internet for the past 40 or 50 years
49:08 which is a bunch of you know
49:12 misogynistic Anonymous you know horeshit
49:16 mixed in with decent stuff so the the
49:19 tools right now just are are you know a
49:23 a decently raw reflection of what's
49:25 already out there which is bad
49:29 so yeah I know it's frustrating thoughts
49:32 on perplexity
49:35 um I like perplexity they're well funded
49:38 they've got I think it's Google behind
49:41 them
49:45 um open continue no I did that in gp4
49:51 asked it to make that
49:52 image oh that was a dolly 3
49:55 image yeah yeah if you really want
49:58 better image control I would go to Mid
50:01 Journey but mid Journey makes a bunch of
50:03 shitty over sexualized women too so I
50:06 don't I don't
50:08 know I don't
50:10 know I mean if you really want to if you
50:13 really want to go go to
50:16 a go to models that
50:19 that you know have addressed some of
50:22 this you're probably going to have to go
50:23 to the stable diffusion
50:25 world and find luras that are that are
50:28 well balanced or find a Laura that just
50:30 makes images that you like I'm sure I'm
50:32 sure there's a ton out there but it's
50:34 just a lot more complicated uh it's a
50:37 lot more complicated an
50:47 environment 40 or 50 the inter internet
50:50 not been around that long not true The
50:53 Internet's been around
50:54 for the internet was around for
50:59 some version of it since the
51:01 60s
51:04 so 40 50 60 almost 70
51:10 years the
51:14 uh I mean Network computers the Internet
51:18 proper probably 50 years in
51:21 fact I think the 50th 50th anniversary
51:24 of the Internet isn't that coming up the
51:27 the center for people centered internet
51:30 aren't they on their 50th Anniversary
51:32 coming
51:37 up this shit's been around for a while
51:40 people
51:42 people you're thinking worldwide web
51:44 worldwide web Tim burner Lee came up
51:48 with the protocol I think in
51:50 1989 I think the first browser was in 92
51:53 or three I started an online magazine
51:58 and an online agency in '94 so '94 was
52:01 when the internet the the web was just
52:04 just getting getting off the
52:07 ground so but
52:09 still my God I
52:12 did am I that old yeah we're old
52:15 dude the big event for the 50 50th
52:18 anniversary with ie trip I Triple E this
52:22 weekend with Vince surf there you go I
52:25 was born in 1988 so please will the
52:28 majority of you feel old now yeah that's
52:31 I graduated high no I graduated college
52:34 in 87 Pate so I was I was living in New
52:38 York City pursuing a life on the life in
52:41 the Arts when before you even born
52:46 boy had a theater company I was selling
52:49 yo-yos at FAO Schwarz at Christmas time
52:52 come
52:55 on all right
52:58 you all make me feel better about my age
53:01 shut up we're not that
53:03 old [ __ ] youngsters youngsters with
53:06 your energy and your brain cells intact
53:08 and your [ __ ] eyesight with your
53:11 stupid little text on your stupid little
53:14 Tik Tock apps where you can actually
53:16 [ __ ] read
53:18 it horrible horrible of
53:24 you all right what are we gonna do um
53:27 someone asked
53:30 about gp4 I'll show you this in a second
53:33 this is the uh the GPT I did for the AI
53:37 Futures book that Cindy and I worked on
53:40 but let me let me jump over to chat GPT
53:42 so someone was asking about you know do
53:46 I like gp4 or GPT 40
53:49 better
53:52 um so here's what you need to
53:55 know well you don't you don't need to
53:57 know this here's what I'm going to tell
54:03 you newest and most advanced
54:06 model Advanced model for complex tasks
54:10 everyone's been talking about GPT 5 gp5
54:13 GPT 5 and then when this thing came out
54:15 GP people assumed it was like oh that
54:18 must be GPT
54:19 4.5 it's not it's a completely different
54:23 model it's a completely different
54:25 architecture the O here stands for Omni
54:27 or
54:28 multimodal so it can natively ingest
54:33 audio images and text where gp4 is a is
54:38 a text large language model so this is
54:40 text
54:42 only and this is multimodal now you're
54:45 like but K gp4 makes pictures so it's
54:49 not just text no Dolly 3 is a completely
54:52 separate model so when you're in chat G
54:55 pt4 and you say make me image of it
54:58 spins up a separate application or does
55:00 a call to a separate model that makes
55:02 that
55:03 image gp4
55:06 Omni can natively generate images can
55:10 natively generate audio and sound
55:13 effects and voice it can natively
55:17 understand your
55:19 voice and the inflection of your
55:22 voice but here's the problem right now
55:27 now we only have a hobbled version of
55:30 gp4
55:31 omni if you use the chat GPT app even if
55:35 it says gp4 Omni it's the old voice
55:38 model if you generate an image it's
55:41 dolly3 so they haven't released the
55:44 actual powerful components of GPT Omni
55:47 now couple of things about it it is much
55:50 faster than
55:52 gp4 it's a smaller model that's cheaper
55:56 to run and faster which I think is why
55:59 they're going to be giving it away to
56:00 for free to everyone but it's it's a
56:03 fundamentally different different model
56:06 and we don't have a lot of the stuff
56:08 that makes it really
56:11 interesting
56:14 um
56:16 gp40 won't give me free image yeah cuz
56:20 it's not well
56:23 it's yeah it's still using do right so
56:27 if you come in here and you
56:29 say
56:35 make me a picture of a 70s muscle car in
56:39 an abandoned Factory with sunlight
56:41 streaming Through the Windows
56:44 oops light streaming through the broken
56:54 windows I'm pretty sure this is
56:57 Dolly we'll see when it generates the
57:00 image we'll go look at it please look at
57:04 the exploration of
57:07 capabilities open AI blog drop down Okay
57:10 cool so if I click on this yeah this is
57:13 still see this is still
57:15 Dolly so there's your there's your
57:19 prompt you know here's your editing
57:22 right so you can
57:25 go and then I can go
57:28 so green
57:35 car let's see what it
57:42 does all right I will go look at that
57:46 Danielle thank
57:48 you see if we get a green card look at
57:50 that
57:52 boom so still not bad but this is Dolly
57:56 this is not
57:57 this is not the um if you've seen people
58:01 talking about GPT 40 does like these
58:05 really crazy
58:07 images it's going to do really crazy
58:11 images that hasn't been released yet so
58:14 it's I I'm I'm a little cranky at at
58:16 open AI for for naming it something
58:19 shitty and then being really muddy about
58:24 what's actually actually released and
58:26 what's not and when it's coming and all
58:28 that sort of
58:30 [ __ ] so it's frustrating it's all of
58:34 them are doing it right now they all
58:35 want to preempt each other
58:37 and um all right let's go look
58:42 at I'm going to go to open
58:45 AI I don't want to go to the
58:51 blog do I
58:53 no well maybe I do
59:06 is this the capabilities you're talking
59:07 about what's the drop down you're
59:09 talking about Danielle oh here explor
59:12 explorations of capabilities these yeah
59:15 so here here are things like you can put
59:17 in very specific text it'll generate
59:19 that then in the next prompt you can say
59:22 write another paragraph on the same
59:24 piece of paper and it'll do that
59:27 and then you can say the robot took that
59:29 thing and ripped it apart because he
59:30 didn't like what was
59:36 written but but none of this none of
59:38 this is there none of this is live right
59:40 now this is all just this might as well
59:43 be you know ask Pate how he how he feels
59:45 about you know marketing documents like
59:47 this we'll believe it when we see
59:50 it um
59:56 if you haven't seen some of these it's
59:58 probably worth looking at a video or two
59:59 here of of some of the voice stuff
1:00:01 that's coming so so to put the voice
1:00:03 stuff in
1:00:06 context okay right now when I use chat
1:00:10 GPT voice so I'm in the chat GPT app on
1:00:14 my
1:00:14 iPad and I put it into voice I'm
1:00:18 using I'm going to use chat GPT 4 not
1:00:22 for for Omni I'm gonna use chat gp4 and
1:00:24 I'm gonna hit my little voice thing
1:00:38 still
1:00:40 connecting when is the next Denver
1:00:42 Nuggets game oops response one
1:00:45 connection inquiry May 18th 2024 hi what
1:00:48 would you like to connect about today
1:00:50 are you looking for ideas advice or
1:00:52 something else when did the Nuggets play
1:00:54 their next game
1:00:58 2 1000 3 1000 4 1000 5 1000 6 1000 7
1:01:04 1000 8 1,000 9 1,000 10
1:01:10 1,000 response two Nuggets game schedule
1:01:14 May 18th 2024 so the current wait what
1:01:17 didn't what drop down didn't I click
1:01:27 which drop down you talking about
1:01:30 Daniela Daniel I got nothing what you
1:01:32 telling me Oh this
1:01:36 one oh all these okay I'll go do those I
1:01:41 see them now thank you um we'll go look
1:01:44 at those
1:01:48 um so we we'll come back to these I want
1:01:50 to go I want to go watch a video too
1:01:52 because I want to talk about the the
1:01:54 audio stuff that's coming
1:01:56 so the reason it takes so long for that
1:01:59 thing to answer there's actually four
1:02:01 things happening there when I talk it
1:02:03 takes my audio turns that into text it
1:02:07 then takes that text sends it to chat
1:02:11 GPT because I asked it for something
1:02:13 that's a current event it did an
1:02:15 internet search so that
1:02:19 du that means it's searching the
1:02:21 internet and then the fourth thing it
1:02:23 does is once it gets an answer it turns
1:02:25 that back into speech and plays it for
1:02:27 me so there's either if you're not using
1:02:29 internet there's three things that take
1:02:31 place there's three
1:02:33 models and then if are searching the
1:02:35 internet you add in the latency of an
1:02:36 internet search in there the average
1:02:39 response time for for a GPT 4
1:02:43 conversation from when you talk until it
1:02:45 responds the average response time is
1:02:48 5.4
1:02:50 seconds the average human being response
1:02:54 time is 250 milliseconds so a quarter of
1:02:58 a
1:02:59 second the new gp4 Omni model the
1:03:03 response the average response time is
1:03:06 320 milliseconds so a third of a a
1:03:10 second so it's within a tenth of a
1:03:13 second of human response time the other
1:03:17 thing about it is that not only does it
1:03:20 understand the words of what you're
1:03:21 saying it understands the
1:03:24 emotional the inflection of your voice
1:03:26 voice so if you're really upset and
1:03:28 excited and agitated it knows that
1:03:30 that's different than if you're super
1:03:32 relaxed and chill and it will respond in
1:03:35 kind so what's coming is very much from
1:03:39 a conversation standpoint like the movie
1:03:41 Her it does not yet have the ability to
1:03:44 take actions but let's let's go through
1:03:46 and look at a couple of these that I
1:03:49 think are kind of
1:03:51 interesting
1:03:53 um yeah that one's boring
1:03:56 what was this
1:03:59 one oh this is it understanding sarcasm
1:04:02 which is
1:04:04 interesting hey Chad are you there hey
1:04:08 yeah I'm here here quick here how quick
1:04:09 it is not SP up let's have some fun uh
1:04:12 I'd like you to be super sarcastic
1:04:14 everything you say from now on is just
1:04:16 GNA be dripping in sarcasm how does that
1:04:18 sound
1:04:20 oh that sounds just amazing being
1:04:23 sarcastic all the time is an exhausting
1:04:25 or anything I'm so excited for this nope
1:04:31 the sasm let's get this party started or
1:04:35 whatever and then this is Saul KH the
1:04:38 guy that created KH Academy with his son
1:04:40 I'll play this I played this last night
1:04:42 but I think this is worth
1:04:50 watching open AI invited myself and my
1:04:54 son Imran here uh to try out some of
1:04:56 their new technology and so we're
1:04:57 curious about how good it might be at
1:05:00 tutoring someone in math on KH Academy
1:05:03 so let's get it
1:05:05 started it's loading
1:05:08 up all right let me pause so over here
1:05:12 what you've got this is an iPad
1:05:14 screen you've got the GPT app here and
1:05:18 then the the Con Academy app or a
1:05:20 website here um this app can see this
1:05:25 screen
1:05:26 and and here right so you can copy
1:05:29 things to the clipboard it'll send it to
1:05:30 that app or in this case they're setting
1:05:32 it up so that it can see that part of
1:05:34 the
1:05:35 screen can see our
1:05:38 screen actually in in this form hello
1:05:41 there I'm here with my son and I love
1:05:45 you to tutor him on this math problem
1:05:48 but don't give him the answer you can
1:05:49 ask questions and nudge him in the right
1:05:51 direction but I really want to make sure
1:05:53 he understands it himself and he's here
1:05:55 in the room so you can can talk to him
1:05:57 directly of course I'd be happy to help
1:06:01 let's look at the problem
1:06:03 together can you first
1:06:06 identify which sides of the triangle are
1:06:09 the opposite adjacent and hypotenuse
1:06:13 relative to angle Alpha all right so I'm
1:06:16 pretty sure this is the angle Alpha
1:06:18 right here
1:06:20 right cors correct now looking at the
1:06:23 triangle which side do you think is the
1:06:25 hypotenuse
1:06:27 [Music]
1:06:29 um hypot I'm not totally sure I think I
1:06:33 think it might be this one but I really
1:06:36 am not sure this side
1:06:37 AC you're close actually side it might
1:06:41 be this one but I really am not sure
1:06:43 this side
1:06:44 AC you're close actually side yeah words
1:06:48 rise from the
1:06:50 deep sort of got
1:06:52 it I emerge from digital sleep
1:06:56 I speak in Rhythm I sing in
1:07:00 rhyme so Prett pretty pretty something
1:07:04 looks like it's looks like it's going to
1:07:06 be pretty impressive when when this
1:07:09 stuff comes out
1:07:11 multi-line rendering robot
1:07:14 texting so multi-line rendering that's
1:07:17 pretty
1:07:18 cool
1:07:20 right um tweet one tweet two or I mean
1:07:25 message one message two to meeting notes
1:07:27 with multiple
1:07:32 speakers okay good morning this our
1:07:34 first team meeting morning morning I'll
1:07:37 be your project manager for today this
1:07:40 project my name is Mark I'll be giving
1:07:41 this be to kick the project
1:07:45 off it's my that's the agenda for today
1:07:48 well of course we're new to each other
1:07:49 so get acquainted first wow this is
1:07:52 pretty slick so this is something we're
1:07:54 actually building this in Story vine
1:07:56 right now Kyle has no clue what do I
1:07:58 have no clue about what's going on you
1:08:01 were frozen for like four minutes we
1:08:03 missed everything you said really
1:08:16 [Laughter]
1:08:22 [ __ ] Kyle didn't miss us no I'm just I'm
1:08:26 just here entertaining myself
1:08:31 apparently how's that any different than
1:08:33 anything you do every other night Kyle
1:08:35 apparently this just like the Kyle makes
1:08:37 himself laugh Channel I don't even know
1:08:39 why we're here you big dumb
1:08:45 dumb prove to us that you're real Kyle
1:08:48 we watched a new Netflix series so it
1:08:50 was so good
1:08:56 oh you fell off at kid math oh okay all
1:08:59 right so let me let me go back so the
1:09:01 kid math okay so here's what I said
1:09:02 after kid math
1:09:05 Pate Pate put a comment in that said um
1:09:10 he thinks that these demos from open AI
1:09:12 were done like with a hardwired
1:09:15 connection and probably a dedicated
1:09:17 server or you know certainly a dedicated
1:09:19 high-speed internet
1:09:22 connection I think that's the case the
1:09:25 the Mo on on the live day the phone that
1:09:28 they used had a hardwired connection he
1:09:30 actually referenced that they're doing
1:09:32 it so they had
1:09:34 consistent consistent quality internet
1:09:37 connection which basically means we need
1:09:39 the 100 you know megabit
1:09:43 ethernet um so so using this stuff over
1:09:46 Wi-Fi is probably going to be
1:09:48 slower um and then so so I just
1:09:51 confirmed that yeah I think that's
1:09:52 that's probably a fair assumption on P's
1:09:54 part and then I just went down and
1:09:56 started looking at the drop down menu so
1:09:58 let me start at the top of the drop down
1:10:00 menu and and show you what what's in
1:10:04 here
1:10:09 um let's see poster creation for the
1:10:12 movie
1:10:14 detective so so these examples are wait
1:10:18 can you hear me
1:10:20 now am I here Jay Hamilton were you at
1:10:24 off office hours
1:10:26 today which is ironic considering what
1:10:28 just happened to your
1:10:31 feed shows us how often he looks down at
1:10:34 the comments well when I'm when I'm
1:10:36 demoing stuff I can't really look
1:10:39 down I watch the movie her while while I
1:10:42 was out the real Kyle abandoned us for
1:10:45 her okay you're here
1:10:48 okay um so the exploration of
1:10:52 capabilities so this is the exploration
1:10:54 of capabilities so we all have access to
1:10:57 gp4 Omni for o but we don't have any of
1:11:02 this [ __ ] what they're talking about
1:11:04 here is the
1:11:06 native image processing [ __ ] it's doing
1:11:10 and the native audio stuff it's
1:11:12 doing that's the stuff we don't have
1:11:14 access to right now we're just getting
1:11:17 to play with the text component of this
1:11:19 I don't know how they took a multimodal
1:11:22 model and made made it so that you can't
1:11:24 use those parts of it what they
1:11:27 have so here's poster creation
1:11:31 for the movie detective so let's see
1:11:34 let's design another poster with two new
1:11:37 characters this is a picture of Alex
1:11:39 nickel this is a picture of Gabrielle
1:11:42 go the final poster of the movie
1:11:44 detective features two large faces of
1:11:47 Alex and Gabe
1:11:49 prominently
1:11:51 so that's pretty slick see full sample
1:11:56 still doing weird nonsense text
1:12:00 but you know upload two
1:12:03 pictures and stick them on a poster it's
1:12:07 got they got crazy eyes you know maybe
1:12:10 they're not releasing this because it's
1:12:12 not quite ready
1:12:14 yet oh that was the other thing I meant
1:12:16 to meant to talk
1:12:18 about this Omni team is a completely
1:12:21 separate team from the gp4 GPT 5 that
1:12:25 the qar team that's doing all the
1:12:30 heading toward AGI this Omni team is it
1:12:33 sounds like it's a separate project
1:12:35 within within open AI that we doing this
1:12:38 and these guys got far enough along that
1:12:40 they decided to release it because it's
1:12:42 faster and cheaper than
1:12:44 gp4 which is why they're going to be
1:12:47 able to give this to
1:12:49 everyone character design so there's a
1:12:52 friendly looking robot
1:12:56 called giri and then it says giri Stands
1:13:00 Tall giri likes to play Frisbee so it's
1:13:03 doing character consistency giri likes
1:13:06 to program on
1:13:07 computers giri likes to ride a bike so
1:13:10 these are pretty consistent so that's
1:13:11 pretty
1:13:15 slick um poetic typography with
1:13:18 iterative editing so there's a
1:13:21 poem there it is written
1:13:24 down making it in dark mode so now it
1:13:27 does it in dark mode remove the notebook
1:13:30 paper lines so it does that it's pretty
1:13:35 slick right again what you're looking
1:13:38 what what you're looking at here is the
1:13:40 output of a model that natively can
1:13:45 understand audio images and text it's
1:13:48 not just text that's calling out to
1:13:50 other services the model itself can
1:13:53 natively process all this stuff and
1:13:55 tokenize it and deal with all
1:13:58 that so it's going to have
1:14:01 some sexy capabilities it's so sexy so
1:14:06 sexy dark mode oh yeah that's
1:14:09 sexy Kyle it showed an upscaled version
1:14:12 of the poster lowered down oh it did
1:14:15 thank
1:14:19 [Music]
1:14:23 you searching for oh yeah
1:14:26 that's
1:14:27 slick
1:14:31 wow sort of fixed is
1:14:38 aiis very
1:14:44 cool commemorative coin design for GPT
1:14:48 40 I created a vector graphics depicting
1:14:51 the open AI logo with the word gb4
1:14:54 underneath it the logo is about the same
1:14:57 width as the text the margin around the
1:15:00 border is
1:15:04 even so here's the
1:15:07 attachment oh here's the
1:15:11 attachment input now here's the
1:15:13 commemerative coin that we made for GPT
1:15:23 4 yeah that's pretty
1:15:28 lick
1:15:30 huh play the sound of coins clanging on
1:15:39 metal
1:15:44 interesting photo to caricature so
1:15:47 there's the photo there's the caricature
1:15:50 nice
1:16:00 huh it's a little [ __ ] Bonkers
1:16:06 people text to font yeah we saw that or
1:16:10 actually I saw that none of you did am I
1:16:13 still here what app is this this is GPT
1:16:16 40 Source Camp Source Camp is having her
1:16:19 mind blown couple of Kevin McAllister
1:16:22 moments going on very nice very nice
1:16:24 okay to be clear
1:16:25 what I'm showing you none of us have
1:16:28 access to yet even though we have
1:16:32 GP GPT 40
1:16:36 Omni the native image capabilities and
1:16:40 the native voice audio capabilities are
1:16:43 not released
1:16:44 yet so we're only playing with the text
1:16:48 capability of this froze again did I
1:16:57 are you still
1:16:58 talking none of us really here froze
1:17:02 again I see you Joy pry can see me and
1:17:05 that's all that matters not frozen to me
1:17:08 if if Joy's happy I'm happy okay text to
1:17:11 font um 3D object synthesis so create a
1:17:15 realistic 3D object of the open AI logo
1:17:26 oh that's interesting it has
1:17:32 it it has it create different pieces of
1:17:36 the animation and then it has it stitch
1:17:37 them together so it's not really doing
1:17:39 3D but it's
1:17:41 doing some sort of weird ass mocked
1:17:44 version of
1:17:50 it nuts so here's the thing let's see
1:17:54 brand
1:17:56 brand logo on a coaster so there's your
1:17:58 coaster there's your
1:18:00 logo oh that's
1:18:03 interesting here's the openai
1:18:06 logo here's the coaster with no
1:18:10 branding a coaster where the top part is
1:18:13 wooden the bottom part is
1:18:18 marble input we've now etched the open
1:18:21 AI logo onto the coaster the coaster
1:18:22 where the top is wooden the open AI logo
1:18:25 goes etched into the middle of the
1:18:27 wooden part the marble is the part with
1:18:30 the word open AI on
1:18:37 it poetic
1:18:46 typography okay meeting notes with
1:18:49 multiple speakers this
1:18:51 is this is what I think paid members
1:18:54 will get oh that's really interesting
1:18:55 Danielle I think I think you could be
1:18:57 right oh that that's really interesting
1:19:00 that if if free members get Dolly
1:19:03 3 and paid members get all this fancy
1:19:06 [ __ ] that's a
1:19:10 possibility this marketing AI tool is
1:19:13 what this is what we're looking at this
1:19:15 is gp4 Omni this is the stuff that's
1:19:19 been announced but not released um this
1:19:22 thing right here meeting notes from
1:19:24 multiple speaker so here's a minute long
1:19:26 audio recording of a
1:19:29 meeting okay good morning here's our
1:19:31 first meeting morning morning I'll be
1:19:35 your project manager for today this
1:19:37 project so this thing describes who the
1:19:39 four speakers are what their names are
1:19:42 and then it says can you transcribe it
1:19:44 with speaker names this is something
1:19:45 called
1:19:46 diarization I'm in the middle of
1:19:48 building this right now for my company
1:19:51 storyvine we've got a transcription
1:19:53 service and right now now my engineering
1:19:56 team is trying to figure out the
1:19:58 diarization to be able to split the
1:20:00 voices
1:20:01 up um which is a non-trivial task it
1:20:05 looks like it just became
1:20:09 trivial right Mark okay good morning
1:20:12 team good morning Mark I'll be your
1:20:14 project manager for the Day
1:20:17 dirk uh yes uh my name is Dirk Dirk meld
1:20:21 uh I will be uh during this project the
1:20:24 mark
1:20:25 expert oh why do they have to make the
1:20:28 marketing expert the one that's like uh
1:20:30 duh which way did he go boss duh which
1:20:32 way did he
1:20:35 go so anyway uh because it's got native
1:20:39 audio processing it can hear things like
1:20:42 who's speaking in a different voice and
1:20:45 it can identify that [ __ ] lecture
1:20:48 summarization here's a video of a
1:20:50 presentation on techniques for
1:20:52 maximizing llm performance 45 minute
1:20:56 video start but depending on the
1:20:58 complexity of your time they are
1:21:01 really
1:21:04 okay the presentation focused on
1:21:06 techniques for blah blah
1:21:09 blah
1:21:15 challenges all right it just it watched
1:21:17 the video for you and summarized
1:21:20 it variable binding Cube stacking
1:21:29 an image depicting three cubes stacked
1:21:32 on a table the top cube is red and has a
1:21:34 g on
1:21:36 it oh that's
1:21:38 interesting variable binding so variable
1:21:42 variable variable and then binding them
1:21:44 together I guess making them predictable
1:21:46 I mean predictability in AI is one of
1:21:49 the big challenges right
1:21:53 now huh
1:21:55 all right and the last one concrete
1:22:00 poetry open AI logo oh that's cool a
1:22:05 concrete poem in the outer shape of the
1:22:07 open AI logo composed of the word
1:22:13 Omni and that's called what
1:22:18 concrete huh well there you
1:22:23 go wild
1:22:27 yeah it's possible Danielle that maybe
1:22:29 some of those features are the ones that
1:22:31 that go to paid users I don't know I do
1:22:33 not know I like I got no I got no Spidey
1:22:37 Sense on who's getting what right
1:22:43 now it looks like gpts right now are
1:22:46 still using chat GPT 4 not the new
1:22:50 one where are you seeing all this so if
1:22:53 you go to open.com
1:23:00 go to spring
1:23:02 update and then the first link hello GPT
1:23:08 40 click on
1:23:11 that and that'll take you to the site we
1:23:13 were just
1:23:16 on all videos on this page are at 1X
1:23:20 real time all videos on this page are at
1:23:23 1X real time
1:23:26 there is no Oz behind the curtain this
1:23:28 [ __ ] actually works that's what they're
1:23:31 claiming we'll believe it when we see it
1:23:33 if we have to have a 100 you know
1:23:35 megabit hardwired connection for this
1:23:37 [ __ ] to
1:23:38 work I assume I assume it's still going
1:23:40 to be significantly better than what we
1:23:42 have now even on Wi-Fi
1:23:45 but but just be skeptical about these
1:23:47 demos because they are marketing people
1:23:49 this is funny this is an example of some
1:23:52 of the limitations of the of the model
1:23:54 this is pretty funny
1:23:56 funny so
1:23:59 let's for the
1:24:03 home what was
1:24:06 that sorry guys I got carried
1:24:10 away guilty is charg what can I say
1:24:13 sometimes I just can't help
1:24:15 myself ready for another
1:24:18 round Francisco
1:24:31 feeling I'm very
1:24:34 wrong hello uh my name is NCH I'm here
1:24:37 with my coworker hi I'm sh I'm trying to
1:24:39 teach my coworker how to speeding maning
1:24:41 we want to start from simple words like
1:24:43 niow can you teach him how to pronounce
1:24:47 that of course hey Nao nice to meet you
1:24:50 niow is pretty straightforward to
1:24:52 pronounce it sounds like me how just
1:24:55 make sure to keep the knee part high and
1:24:58 then go down and pitch on the how give
1:25:00 it a try KNE
1:25:03 how that was great really good first try
1:25:06 on you you're
1:25:08 natural
1:25:09 so he didn't follow directions and it
1:25:12 didn't catch him so
1:25:14 anyway so that that is the [ __ ] that's
1:25:20 coming when we don't
1:25:22 know is it going to be any good good we
1:25:25 don't
1:25:26 know do I think it's going to be good I
1:25:29 think it's going to be good I think it's
1:25:31 going to be powerful but until we have
1:25:34 it we don't have
1:25:36 it I bet they share this after the
1:25:39 Microsoft announcement possibly a year
1:25:42 from now Kyle look up the Limitless
1:25:46 pin which one's that the the
1:25:49 uh the Humane
1:25:51 pin Limitless
1:25:54 pin oh by the way the rabbit um the
1:25:58 rabbit R1 got a software update today I
1:26:01 didn't install it
1:26:03 yet that is um they're using gp4 Omni
1:26:08 for for uh translation services on the
1:26:11 rabbit right now so rabbit is using gp4
1:26:15 Omni in some
1:26:22 ways oh yeah I saw this I saw this
1:26:25 around this is for those who hate good
1:26:27 enough the okay projects the one we have
1:26:31 another brand new AI native device to
1:26:34 talk about today this is fresh off the
1:26:37 heels of the AI
1:26:39 [Music]
1:26:44 pin okay every single product launch now
1:26:47 is basically just a copy of Apple's
1:26:50 product launches they have defined an
1:26:52 entire generation of prodct launch
1:26:55 videos and I mean look at this this is
1:26:57 exactly what every Apple device looks
1:26:59 like when they launch but that's okay it
1:27:01 looks
1:27:04 good that's so true
1:27:07 [Music]
1:27:27 today we're introducing Limitless it is
1:27:29 a web app a Mac app a Windows app and a
1:27:33 wearable Limitless is a personalized AI
1:27:35 powered by what you've seen said or
1:27:37 heard Limitless augments not replaces
1:27:39 human intelligence with artificial
1:27:41 intelligence all right so right away the
1:27:43 form factor is It's recording all the
1:27:46 time and that yeah yeah we're going to
1:27:49 see a ton of these I ordered one of
1:27:52 these the the open source one that came
1:27:54 out like a month after the the rabbit
1:27:58 said they were shipping so maybe in
1:28:01 March I ordered that one I haven't
1:28:03 received it yet there's an open source
1:28:05 one and this is this is very similar to
1:28:07 that open source one that I ordered
1:28:09 wonder if it's using the same uh
1:28:11 Hardware spec as that that's
1:28:15 interesting
1:28:20 um the gp4 versus gp4 Omni is more than
1:28:25 somewhat confusing Firefly it is
1:28:28 unacceptably confusing it is like it the
1:28:32 the naming is a
1:28:34 disaster and the fact that they didn't
1:28:36 launch it completely is kind of a
1:28:38 disaster right and the fact that like
1:28:41 they're like it's going to be free for
1:28:43 everyone but it's not really free for
1:28:44 anyone right now but in the coming weeks
1:28:46 it will be and then you'll have this
1:28:47 voice thing but we're not really sure
1:28:48 about the security yet so maybe you'll
1:28:51 have it maybe you won't oh and by the
1:28:52 way it does all this other cool image
1:28:53 [ __ ] we didn't even talk talk about but
1:28:56 we got no no Clarity on when that's
1:28:58 coming or who gets it right so it's it's
1:29:00 a mess right now it'll look better than
1:29:03 the Alexa I wear as a hat right now
1:29:06 that's
1:29:09 funny that's
1:29:12 good
1:29:15 uh unacceptably confusing
1:29:18 yes it is I tagged uh Greg Brockman and
1:29:23 Sam Altman to today in another in
1:29:26 another tweet uh suggesting the name GPT
1:29:29 Omni cuz GPT Omni sound Omni sounds kind
1:29:32 of cool it's GPT Omni it's clearly
1:29:36 different than chat GPT
1:29:40 4 I don't know whatever guy looks like
1:29:44 Bill Burr he does a little
1:29:51 bit Firefly supports my GPT on
1:29:55 suggestion um Greg Brockman and Sam
1:29:58 Alman have not adopted by naming
1:29:59 convention because or my my brand name
1:30:02 suggestion because they don't know who
1:30:04 the [ __ ] I
1:30:06 am maybe Frank should send the message I
1:30:10 I I don't think Frank is as appreciated
1:30:12 out in the real world as he is in this
1:30:16 channel let's just call it that done
1:30:19 yeah GPT Omni I I think we should I like
1:30:22 Omni yeah well that it's from them it's
1:30:24 their name it's just they went with GPT
1:30:27 40 because someone was like Hey wouldn't
1:30:30 it be cool if we just used an O it' be
1:30:31 kind of like a zero but it's an o yeah
1:30:34 except our previous product is called
1:30:36 GPT 4.0 you [ __ ]
1:30:41 [ __ ] was was no one in the room willing
1:30:45 to say
1:30:47 that
1:30:52 apparently Kyle is infam come on yeah
1:30:56 yeah I'm I'm infamous in my own
1:30:59 mind spicy Kyle but seriously like no
1:31:04 one there no one there had the guts to
1:31:07 go uh hey uh Sam uh Greg uh whoever
1:31:11 whoever the [ __ ] in the room
1:31:15 uh do don't you think that might be
1:31:18 confusing no one no
1:31:21 one or someone said hey don't you think
1:31:23 that might be confusing and then they
1:31:25 were like hey you know what I think it
1:31:27 would be good to cons confuse our our
1:31:29 consumers let's do it GPT Overlord GP
1:31:34 sounds like zero Omni has a great
1:31:36 presence Omni presence if you will
1:31:39 exactly GPT Omni that's what we're going
1:31:42 with people all right that's what we're
1:31:43 calling it in
1:31:45 here calling Mr Herman calling Mr
1:31:48 Herman if you can't convince with your
1:31:52 ey orb scan then what
1:31:55 exactly oh back to calling them muddy AI
1:31:59 they are totally muddy AI right now they
1:32:03 get close to Sam St Sam and then they St
1:32:05 struck it's probably true and listen I
1:32:07 mean maybe he's a douchebag to work with
1:32:09 who
1:32:10 knows and it might not even be him like
1:32:13 it might be this might be a kind of
1:32:14 thing where you know people are making
1:32:16 decisions that
1:32:20 are he here's the thing to remember
1:32:22 about about open AI
1:32:25 they're an 800 person company which
1:32:27 sounds like a lot but Google is 200,000
1:32:30 200,000 versus 800 um open AI started as
1:32:36 and is still primarily a research
1:32:39 company it's it's a company full of
1:32:41 scientists doing technical
1:32:43 research that happened to have launched
1:32:47 an app that was the fastest adoption of
1:32:50 technology in history from 0 to 100
1:32:53 million users in about 6 weeks that's
1:32:56 what chat GPT did so overnight chat GPT
1:33:00 became a consumer company they're not a
1:33:02 consumer company so I give them a little
1:33:05 bit of Grace that that you know they're
1:33:09 not knocking the marketing and and
1:33:11 product marketing out of the park but
1:33:15 seriously GPT
1:33:17 4.0 and
1:33:19 GP no one no one uh boss don't you think
1:33:23 maybe that's uh [ __ ]
1:33:27 stupid no no no
1:33:34 one not 800 797 now right I don't know
1:33:38 what that's in reference
1:33:41 to omnid dextrous they need a marketing
1:33:45 research division they need a they just
1:33:47 need a marketing
1:33:49 person how does one keep up with happen
1:33:52 chance
1:33:56 chat gbt is like the wheel truck company
1:33:59 what' you buy truck what kind wheel
1:34:03 exactly H well you know what it probably
1:34:07 was was someone probably some engineer
1:34:09 like well who uh you know clearly uh uh
1:34:12 the the letter O is is not a numeral and
1:34:15 and from a perspective on the ASI U on
1:34:18 the asky table uh the letter O is 27
1:34:22 positions away from the the the numeral
1:34:25 zero so clearly it's different I mean I
1:34:28 who would confuse that it's the the O is
1:34:31 small and zero is is oblong and taller
1:34:34 so uh I I think it makes perfect sense
1:34:36 and it's it's perfectly clear to me that
1:34:38 it implies I'm Omni Channel it says it
1:34:42 right right you know it's if you extend
1:34:45 the logic of the man you know maybe
1:34:50 maybe stands for oyle help us
1:34:55 they need a common sense person as as as
1:34:58 my buddy Aaron sugarm he president at
1:35:01 agency.com said to me once I was I was
1:35:04 similarly apoplectic as I am now about
1:35:07 some thing that one of our employees did
1:35:11 and I said Aaron it's common sense and
1:35:15 his reply to me was Kyle um Common Sense
1:35:19 isn't common I was like oh [ __ ] you're
1:35:23 right
1:35:26 [Applause]
1:35:28 that was that was a that was like just a
1:35:30 moment I'll never forget of like just
1:35:32 that moment of like thinking that isn't
1:35:34 this obvious and like no it's not like
1:35:37 until you see something you can't see
1:35:41 something which I know that's profound
1:35:43 we should flood open AI with marketing
1:35:45 resumés saying it appears that they're
1:35:48 hiring
1:35:52 exactly it's like mov 43 with the
1:35:55 cooling
1:36:00 chamber common sense is not common I
1:36:03 know that was really funny I sort of
1:36:05 laughed and cried in the same moment
1:36:07 when he said that cuz I just I realized
1:36:10 he was he was right and it and it just
1:36:12 like what hit me was like oh God I'm
1:36:14 going to
1:36:17 like this is a thing when when you end
1:36:20 up working with people that get you and
1:36:24 get who you are and kind of get what
1:36:26 your worldview is it's a really rare
1:36:29 thing a lot of times a lot of times you
1:36:32 just assume that people around you get
1:36:35 the way your brain works and the way
1:36:37 your worldview is and most of the time
1:36:40 it's not the case most of the time
1:36:42 people are just like I don't know what
1:36:43 the [ __ ] Kyle's talking
1:36:45 about so you
1:36:49 know if you find people you know that
1:36:52 get who you are and how you work
1:36:54 hang on to
1:37:00 them I don't know why we're still
1:37:02 looking at that
1:37:03 dude
1:37:05 dude go
1:37:08 there Kyle your hair is on fleek tonight
1:37:11 I think that's a good thing hang
1:37:19 on hey someone said that my hair is on
1:37:22 fleek and I'm a gen xer can can you tell
1:37:24 me what the [ __ ] they
1:37:30 meant response three slang explanation
1:37:33 May 18th 2024 they're giving you a
1:37:35 compliment oh yeah is a slang term that
1:37:38 means something is perfectly styled or
1:37:40 looks really good so they were saying
1:37:42 your hair looks
1:37:43 great yeah yeah chat GPT hey when are we
1:37:47 getting the new version of you CU you're
1:37:49 kind of slow and tired
1:37:56 response for new version expectations
1:37:58 May 18th 2024 I understand your
1:38:01 frustration open aai often updates its
1:38:03 models to improve speed accuracy and
1:38:06 features but I don't have a specific
1:38:08 timeline for when a new version of me
1:38:10 might be released you can always check
1:38:12 open AI announcements for any news on
1:38:16 updates what's a gen Z way of saying
1:38:19 you're a loser
1:38:20 [Laughter]
1:38:27 response five modern slang for
1:38:29 disapproval May 18th 2024 in gen Z slang
1:38:34 a somewhat playful way to call someone a
1:38:35 loser might be to say they're basic or
1:38:38 to jokingly use the term local both of
1:38:40 these are mild ways to suggest someone
1:38:42 is unoriginal or typical you know you
1:38:46 are basically
1:38:51 local Response six friendly banter May
1:38:54 18th 2024 ha nice one I guess I walked
1:38:57 right into that what's next on you did
1:38:59 you dummy um side hustle Mimi did you
1:39:01 tell it to add the date and number your
1:39:03 chats yes I did my
1:39:05 Jarvis um yes I named him that doesn't
1:39:07 do that yeah so in my oh that was
1:39:10 something I actually wanted to to show
1:39:12 you guys today I saw a cool tweet that
1:39:15 someone copied and pasted something from
1:39:16 open AI that I thought was a good idea
1:39:19 um showing the new memory
1:39:22 feature um
1:39:24 I wrote in my custom instructions so if
1:39:27 you go to chat GPT couple of things have
1:39:30 changed you used to have your name down
1:39:33 here they've now moved your account
1:39:35 information to the upper right like most
1:39:37 other software does it um but if I go to
1:39:40 my personal
1:39:42 account and then I go to um customize
1:39:46 chat
1:39:47 GPT I've
1:39:50 got um under how would you like chat GPT
1:39:52 to respond I say for every response give
1:39:55 it a sequential response count followed
1:39:57 by a response title that summarizes the
1:39:59 response in in conversation mode it's
1:40:02 annoying but in text mode it's actually
1:40:05 really good and then I also this is the
1:40:07 thing I'm really excited about I added
1:40:10 avoid using these over word overused
1:40:12 words or phrases delve tapestry vibrant
1:40:15 landscape realm Embark vital moreover
1:40:19 arguably dive into it's important to
1:40:22 note important to consider based on the
1:40:25 information provided remember that
1:40:27 navigating the landscape or tapestry a
1:40:30 testament to so it my my uh chat GPT
1:40:33 doesn't say any of that annoying [ __ ]
1:40:35 anymore um let me go let me go grab
1:40:38 something
1:40:40 [Music]
1:40:44 here so i s i I pulled this off a tweet
1:40:47 I forgot who I who who I got it from
1:40:50 but if you don't know it chat
1:40:56 GPT now has a thing called memory and
1:40:59 everyone's got this I think even free
1:41:01 accounts have this so if you go to
1:41:03 settings and then you go to
1:41:06 personalization it's got this thing
1:41:07 called memory which you can turn on and
1:41:09 off and then if you Ma manage your
1:41:13 memory it's got all these things like
1:41:17 I'm 59 and have 35,000 followers on a AI
1:41:21 education Channel I prefer images in
1:41:25 16x9 wide format um I created a
1:41:29 framework called play solve share um
1:41:32 what else so so you can just as you're
1:41:35 writing in chat gbt you can say hey
1:41:37 remember this and it writes it to your
1:41:40 memory so this tweet basically said you
1:41:46 can basically make chat GPT behave in a
1:41:49 different way remember to use a friendly
1:41:52 and conversational tone remember to
1:41:54 address me by my first name remember to
1:41:57 use professional language in our
1:41:59 conversations remember to ask me about
1:42:01 my hobbies or interests remember to
1:42:04 provide motivational quotes to start our
1:42:06 chats so you can grab all of
1:42:08 these these these were given by chat GPT
1:42:11 so let's go over to chat
1:42:13 GPT chat J and so what I'm going to do
1:42:17 is I'm just going to go into a regular
1:42:20 chat and I'm going to go
1:42:24 hey chat
1:42:28 GPT hey I want you to remember these
1:42:33 things and I'm going to I'm going to go
1:42:36 change and kill a couple of them hang on
1:42:38 use a friendly and conversational tone I
1:42:40 don't care about
1:42:42 that address me by my first name I'll
1:42:44 put my first name
1:42:49 Kyle remember to use professional
1:42:51 language in our conversations I don't
1:42:53 care about that ask me about my hobbies
1:42:56 that's kind of cool provide motivational
1:42:58 quotes I like that at the start of our
1:43:01 chats avoid using technical jargon
1:43:03 unless I ask for it I like that remember
1:43:05 to remind me about my daily goals and
1:43:08 tasks no because I don't I have I have a
1:43:12 zero relationship with to-do lists
1:43:15 incorporate humor into our conversations
1:43:20 um I like snark
1:43:25 and
1:43:27 smart
1:43:33 comedy I
1:43:36 also
1:43:39 like when you call it like you see it
1:43:46 don't be
1:43:49 afraid to criticize
1:43:55 I won't be
1:43:57 triggered I
1:44:02 promise I promise here we go I promise
1:44:06 uh I'm Gen X and we don't give a [ __ ]
1:44:15 okay remember to use concise and to the
1:44:18 point responses okay that's good um so
1:44:21 I'm going to do that
1:44:26 and so it says memory
1:44:28 updated remember to address them by the
1:44:30 first name so just added all this [ __ ]
1:44:32 so if I go to ma manage memories now so
1:44:34 here's all the things that I just
1:44:42 added all right so let's let's start a
1:44:44 new conversation
1:44:52 [Music]
1:44:56 you know the phrase there are no stupid
1:44:58 questions
1:45:00 period I think that's
1:45:06 stupid I don't know I get it Kyle so it
1:45:10 remembered my name it it's never done
1:45:12 that before the phrase there are no
1:45:14 stupid questions can feel a bit P
1:45:16 patronizing while it aims to encourage
1:45:18 curiosity you can sometimes ignore the
1:45:20 reality that some questions may seem
1:45:22 ill-informed
1:45:25 trivial instead of dismissing them as
1:45:27 stupid it might be more
1:45:30 productive
1:45:32 uh but hey some questions do make you me
1:45:35 remember what the person was thinking so
1:45:37 it's it's trying to do a little comedy
1:45:38 there so I now have you know i' I've now
1:45:42 got a bit more personality and in fact I
1:45:44 think what I'm going to do is I'm going
1:45:46 to go grab my custom GPT
1:45:51 stuff I'm going to leave in the the the
1:45:54 sequencing and the title and the date
1:45:56 but I'm going to pull out this
1:45:59 thing with all the all the phrases I'm
1:46:02 going to pull that out of custom
1:46:03 instructions and put it in my memory so
1:46:05 I'm going to save that and then I'm just
1:46:07 going to go I'm going to go I'm going to
1:46:11 go um I want you to
1:46:16 remember this
1:46:29 memory
1:46:30 updated um are
1:46:34 there other
1:46:37 overused words
1:46:40 or
1:46:42 things
1:46:45 that are OB that that uh make your
1:46:51 output
1:46:57 obviously AI question
1:47:02 mark there are a few other
1:47:06 okay
1:47:09 um go search the
1:47:13 internet for
1:47:18 more and more let's see more
1:47:23 and more
1:47:25 subtle
1:47:37 ones my hair's on
1:47:43 fleek and thank you by the way for the
1:47:46 lovely compliment I just didn't know
1:47:48 what the hell it meant okay bearing in
1:47:51 mind paved the way for delin div in into
1:47:53 the world of okay great avoiding
1:47:56 these
1:47:57 great
1:48:00 remember everything
1:48:04 we've talked about
1:48:09 here memory updated okay my chat GPT is
1:48:13 going to be better than your chat GPT
1:48:14 because it's not going to be as shitty
1:48:16 got it Kyle I've made note to avoid all
1:48:18 those additional overused words and
1:48:20 phrases we discussed if there's anything
1:48:22 else you'd like add or change feel free
1:48:24 to let me
1:48:33 know I think you should also remember
1:48:36 that using the pop poop emoji is
1:48:50 funny has anyone tried to share a GPT
1:48:53 was someone what was that
1:48:57 Fleek send me that list Kyle please oh I
1:49:00 don't I just go how I got that list was
1:49:03 I told chat GPT to give me the list and
1:49:05 then I told it to search the internet to
1:49:06 find me the rest of the list it doesn't
1:49:09 matter like there's ones that bug me
1:49:12 like delve and and and Del Del is the
1:49:16 one noted Kyle I'll make sure to sprink
1:49:20 in some humor with the poop emoji when
1:49:22 appropriate if there's any anything else
1:49:23 you'd like to tweet just let let me know
1:49:26 poop
1:49:29 emoji people are going to be like Kyle
1:49:31 why is it always using poop emojis I'm
1:49:33 gonna be in some some corporate demo I'm
1:49:37 just gon to keep doing poop
1:49:38 [Laughter]
1:49:44 Emojis all right the other word I would
1:49:47 love to leave out is robust I had a
1:49:49 cooworker that would say robust
1:49:57 does anyone tried to share a GPT with
1:49:59 someone who doesn't have an account or
1:50:00 paid account that's so again the problem
1:50:04 Danielle right now is um all of the
1:50:07 stuff that we got all EX excited about
1:50:09 that non gp4 people are going to have
1:50:12 GPT 40 they don't have it yet none of
1:50:15 this shit's been rolled out so um here
1:50:19 here's what I did here's how you can
1:50:20 test it so if you want to test um
1:50:24 if you want to test if free chat GPT has
1:50:26 been rolled out go to Chrome and go new
1:50:29 incognito
1:50:30 window and go to chat gp.com
1:50:34 chat
1:50:36 gp.com
1:50:40 and then just look in this pull down
1:50:42 menu see where it says GPT 3.5 and then
1:50:45 it says GPT 4 pay for it so long as
1:50:49 you've got that this per this account
1:50:52 cannot see
1:50:55 um cannot see your
1:50:57 GPT so they haven't rolled it out yet
1:51:00 because they're poo poo
1:51:05 heads La the little details chat GPT is
1:51:08 based off the old school Court courtroom
1:51:10 typewriter
1:51:11 operators I did try to share GPT with
1:51:14 non plus users I write menus and social
1:51:17 media
1:51:18 for a something and I put just about
1:51:22 every Adit in my custom instructions Oh
1:51:25 that's oh that's super cool oh that you
1:51:27 don't want oh yeah you writing menus you
1:51:31 don't want the awesome
1:51:37 tortilla the tapestry of the salad
1:51:45 bar I hate sending [ __ ] before I'm done
1:51:48 I hate that too it drives me [ __ ]
1:51:49 crazy
1:51:53 what did I miss uh we just added so I
1:51:57 just went in I found a cool um list of
1:52:01 things this was something that open AI
1:52:04 put
1:52:07 out which is a list of things that you
1:52:10 can add to your memory so remember to
1:52:13 use a friendly tone use my first name
1:52:16 professional language so it's just a
1:52:18 it's just a bunch
1:52:19 of like rules of for how you want chat
1:52:22 GP to to behave and you can just go into
1:52:25 chat GPT itself and say hey remember all
1:52:27 this [ __ ] and so I just took this list
1:52:30 that I got off Twitter and I just
1:52:32 modified it got rid of some [ __ ] I
1:52:34 didn't want added some [ __ ] I
1:52:36 wanted and then I told it I also wanted
1:52:38 it to use poop emojis
1:52:42 more there's a rule in my company that
1:52:44 we can't make poop jokes because I made
1:52:46 too many of them my founder was like
1:52:49 listen we got to knock it off with the
1:52:50 poop jokes
1:52:56 see I'm not just annoying here I'm
1:52:58 annoying professionally
1:53:04 too where did you find that list here
1:53:06 I'll you know what I'll do I'll put
1:53:08 this
1:53:10 um I'll put this in uh the Irregulars
1:53:17 channel so in share cool things I'm just
1:53:20 going to pop it in there
1:53:28 um I'll
1:53:34 put some things to add to chat GPT
1:53:39 memory oops
1:54:02 [Music]
1:54:04 personalize to your
1:54:14 taste
1:54:17 boom all right there you go where is it
1:54:22 new activity
1:54:24 some things to add to chat GPT memory
1:54:26 all
1:54:33 [Laughter]
1:54:42 right everyone's recording every every
1:54:45 meeting now aren't they it's it's it
1:54:47 really
1:54:49 is
1:54:51 we it's get it's going to get surreal
1:54:53 here people it's going to get
1:54:56 surreal excellent custom guard rails for
1:54:59 GPT designers yep very
1:55:02 cool you're welcome cam katkin cam
1:55:06 katkin from
1:55:08 Cleveland hey how's it going there I'm
1:55:10 cam Kin I'm from Cleveland cam K from
1:55:12 Cleveland nice to meet you hey let me
1:55:15 talk to you about
1:55:17 insurance you
1:55:19 know sometimes you think life's just
1:55:21 swimming along and everything's going to
1:55:22 be fine but you're not
1:55:27 protected wait you mean this isn't
1:55:29 professionally this is my professional
1:55:33 Corey I think I think I think you
1:55:36 understand my level of professionalism
1:55:38 here's the deal I'm professional as [ __ ]
1:55:41 when it comes to like getting [ __ ] done
1:55:43 and you know
1:55:46 delivering um high quality
1:55:49 work um but when it comes to doing the
1:55:51 stupid [ __ ] of corporate America it
1:55:54 drives me batshit crazy I think that's
1:55:56 why all the politics stuff drove me so
1:55:58 nuts it's just they're [ __ ] weird man
1:56:01 they're just you could make money with j
1:56:04 a lot of these people are people just I
1:56:05 wouldn't want to spend a [ __ ] you
1:56:07 know minute with if they weren't paying
1:56:09 me
1:56:12 money and if you're one of my clients I
1:56:15 will never admit who it is I'm talking
1:56:19 about but it's most of them
1:56:28 Kyle is a Rainman in this
1:56:32 area I I am I'm I'm professional Rainman
1:56:36 yeah yeah yeah time time time for
1:56:39 contracts time for contracts yeah yeah
1:56:42 I'm an excellent negotiator I'm an
1:56:44 excellent negotiator time for contracts
1:56:46 yeah yeah three 362 362 3 362,000 that's
1:56:51 we're going to charge you 362 ,000 yeah
1:56:53 yeah yeah per per per per sentence yeah
1:56:57 yeah
1:56:59 [Laughter]
1:57:04 yeah H and I kind of lose my [ __ ] like
1:57:08 grain man when people do stupid stuff
1:57:11 you know he did it when when Charlie
1:57:13 babbett hugged him I do it when when
1:57:16 clients do stupid [ __ ] it makes me
1:57:19 apoplectic
1:57:24 ah TK met cuno last night couldn't call
1:57:28 her by her first name it was cuno it's
1:57:31 my Kyle my best Kyle's voice gruno
1:57:34 what's
1:57:38 happening yeah it's it it is it's a
1:57:41 little it's a little [ __ ] up because
1:57:43 there's a couple of uh Tik Tock names
1:57:46 that slightly overlap and then I can't
1:57:48 remember the first names of people it's
1:57:50 a
1:57:51 mess that's really funny Joker in the
1:57:54 house I'm Rainman but without the math
1:57:57 skills that's good that's solid I'm I'm
1:58:01 uh I was I was once actually my um the
1:58:04 the chapter in the AI Futures book my my
1:58:07 chapter opens up with um
1:58:11 um clay shery once called me the Forest
1:58:14 Gump of the
1:58:15 internet he was like there's there's no
1:58:19 reason someone with your kind of
1:58:21 ridiculousness should be
1:58:23 having as much success as you're having
1:58:25 with running a business and so he's just
1:58:28 like you just seem to stumble into all
1:58:30 this cool [ __ ] that you're
1:58:33 doing and that's you know it's funny we
1:58:35 we were talking about that we've been
1:58:36 talking a lot about the salon and the
1:58:38 future of the salon and what's the
1:58:39 purpose of it and who do we want in it
1:58:41 and how do we want to support them and
1:58:43 all that sort of stuff and it's
1:58:45 like Leah who I founded the salon with
1:58:48 she and I both have kind of made a
1:58:51 career of
1:58:53 play and you know having a good time but
1:58:58 in a focused way and and I I think this
1:59:01 idea of focused play might be something
1:59:04 we really lean into with the salon that
1:59:06 one of the things that makes the salon
1:59:09 magical is that there's a lot of people
1:59:12 in there doing quite remarkable work but
1:59:14 they're not taking it too seriously
1:59:15 they're not taking themselves too
1:59:17 seriously and what that does is it makes
1:59:21 it safe for other
1:59:24 people to not be
1:59:26 perfect right if you've got a [ __ ]
1:59:29 bunch of you know Masters of the
1:59:32 Universe in there going oh my you know
1:59:34 just blow hards
1:59:36 like well let me tell you about my
1:59:38 prompt engineering approach and My
1:59:42 Philosophy nobody wants to be around
1:59:45 that nobody gives a [ __ ] about your
1:59:46 stupid prompt Mastery you douche
1:59:55 nobody cares people are just trying to
2:00:00 learn pleus you cool thank you I
2:00:03 appreciate that L My Philosophy Kyle
2:00:06 Gump life is like a box of
2:00:08 AI you know what it we are definitely in
2:00:13 I I'll tell you what having a Forest
2:00:15 Gump attitude right now about this AI
2:00:17 stuff is where it's at I think it's a
2:00:20 superpower
2:00:22 because Forest
2:00:24 Gump he ran and he ran and he ran and he
2:00:27 ran and he ran and then one day he just
2:00:29 stopped running people like Forest
2:00:31 what's going on he's like I'm done now
2:00:35 or whatever it was I'm tired whatever he
2:00:37 said right he's just done now that's
2:00:40 what it is with these tools we we're
2:00:42 we're in the tools we're playing with
2:00:43 the tools we're making
2:00:45 gpts open AI [ __ ] up the GPT store and
2:00:48 they don't make them available to
2:00:49 everyone and then we're all like all
2:00:51 right [ __ ] it moving on next
2:00:54 right we didn't get too wrapped up in it
2:00:56 nobody well some people might have cried
2:00:58 but you know not for
2:01:01 long and I think that's I think that
2:01:04 that mentality of just oh new thing
2:01:06 showed up okay I'll go this way now oh
2:01:08 another let me take oh this one has that
2:01:11 cream [ __ ] in it that's not a good
2:01:13 chocolate I'll throw that out oh this
2:01:15 one's good this one's got peanut butter
2:01:17 I like peanut butter peanut butter is
2:01:18 tasty Charlie babbit Charlie babot time
2:01:22 for wner I'm an excellent
2:01:24 driver we moved on exactly how important
2:01:28 is storage for AI massively important if
2:01:31 you're open
2:01:33 AI I don't think any of us really give a
2:01:36 [ __ ] we're just like hey open AI
2:01:39 Microsoft go figure that server [ __ ] out
2:01:42 I think Pate worries about storage a
2:01:45 lot um where I think you have to worry
2:01:47 about storage is if you want to save any
2:01:49 of the [ __ ] that you make cuz we're all
2:01:51 going to be making so much more [ __ ]
2:01:53 that how where where do you put it
2:01:57 so ask Jim oh how important is storage
2:02:01 ask Jim that's who who said that Silver
2:02:04 Fox solid solid comedy right there you
2:02:08 could follow that that up with a poop
2:02:10 emoji slam dunk uh Jim Jim runs storage
2:02:16 facilities and see someone asked about
2:02:18 storage for AI get it see you see how it
2:02:21 see how it works here is comedy it's
2:02:23 comedy
2:02:25 comedy all
2:02:29 right I was going to show something else
2:02:31 wasn't I was I going to show something
2:02:33 else does anyone remember anything I've
2:02:35 said tonight video is robbing mine got
2:02:37 an external hard drive I
2:02:39 know I um I went to I went to grab a
2:02:44 picture out of my photos app today and
2:02:47 it said iCloud syncing is paused because
2:02:50 your local storage is has been depleted
2:02:53 or some [ __ ] like that and I was like oh
2:02:56 come on just figure it out put in the
2:02:59 cloud stop bugging
2:03:03 me we remember nothing but we like the
2:03:06 list that was
2:03:08 [Laughter]
2:03:13 nice you can make
2:03:17 money which time zone are you in I am in
2:03:21 uh Mountain time I'm in Denver Denver
2:03:24 Colorado what's your live schedule well
2:03:27 I generally go live five nights a week
2:03:30 starting around 8:00 P.M mountain time
2:03:34 sometimes occasionally I'm doing
2:03:35 something and then I start a little bit
2:03:37 later sometimes it could be as late as 9
2:03:41 o'l I used to do seven nights a week and
2:03:44 I did that for a year and so I've backed
2:03:46 it off
2:03:48 to um five nights a week and I got to
2:03:51 tell you um
2:03:53 it has given me some space to figure
2:03:55 some new stuff out I'm I'm kind of I'm
2:03:57 I'm newly energized this is actually
2:03:59 this has been good
2:04:02 um cuz the timing's good because as when
2:04:07 when the rest of this open AI [ __ ] drops
2:04:09 when when we get voice and when we get
2:04:11 these new modes and models and [ __ ] like
2:04:14 that there's going to be some work for
2:04:16 us to do to figure out what these tools
2:04:17 are
2:04:19 so so anyway and yeah I started a new
2:04:22 series recorded series on Tik Tok where
2:04:24 I go out and I take a nice little walk
2:04:26 in the
2:04:28 morning so I've got more content up
2:04:30 lately had to upgrade my C drive to 4
2:04:33 terabytes for
2:04:35 [Laughter]
2:04:38 iCloud we're happy for you thank you Dr
2:04:41 J appreciate that for sure lots of
2:04:44 figuring what sector has open AI
2:04:47 predominantly influenced
2:04:52 I
2:04:54 think I mean maybe it's because I'm
2:04:56 closest to it but I think I think the
2:04:59 world of marketing is being hit pretty
2:05:02 hard by by chat GPT and and and and the
2:05:09 like
2:05:13 um I think clients
2:05:16 are it's funny I think clients are not
2:05:19 really willing to go heavy in on AI
2:05:23 but they're saying to their agencies hey
2:05:26 uh question this time and materials
2:05:29 business model of yours where you charge
2:05:31 me more money for like a a pen of like
2:05:34 50 copywriters couldn't that be like
2:05:38 two or or instead of it taking you three
2:05:42 months to turn around social media copy
2:05:45 could we have it in like six weeks or a
2:05:48 month and hey I have an idea what if it
2:05:50 cost us half as much for the same amount
2:05:53 so I think clients are starting to do
2:05:55 that to agencies and so agencies are
2:05:57 starting to have have to scramble um I
2:06:00 think Tech support's going to be Absolut
2:06:02 or not tech support I think customer
2:06:04 service call centers are going to be
2:06:06 absolutely decimated if you look at
2:06:09 here's the
2:06:11 deal the the chat Bots that with the
2:06:14 current technology where there's like
2:06:17 two to five second latency between when
2:06:19 you ask a question and when it gets
2:06:21 answered that's not acceptable for
2:06:23 customer service the [ __ ] that open AI
2:06:26 is demoing with this third of a second
2:06:28 response time that's plenty fast like
2:06:30 that's
2:06:32 fine and so I think I think customer
2:06:35 service call centers are going to be
2:06:37 effectively obliterated
2:06:40 if if the
2:06:44 uh the data to answer the questions if
2:06:48 they can get it to work where it's using
2:06:51 like rag retrieval augmented generation
2:06:54 where it's pulling actual answers from
2:06:56 an answer database but it's using you
2:06:59 know large language models to make it
2:07:01 sound good um I I think they're just
2:07:04 going to obliterate that whole industry
2:07:06 but it's it's not there yet they're not
2:07:08 the the tech isn't quite there yet but
2:07:10 that that one's due for a massive
2:07:12 shakeup I like the call center video
2:07:15 where the guy asked the AI for
2:07:16 historical essay and it delivered yeah
2:07:18 exactly that that's the thing is they've
2:07:20 got to they've got to figure out how to
2:07:22 navigate um there's a there's a person
2:07:25 in the AI
2:07:28 Salon um created Rachel created
2:07:33 um a system for an insurance company
2:07:37 where their employees could ask about
2:07:39 their their personal health insurance so
2:07:43 they could call up this system or no it
2:07:45 was it was it's a chatbot it's a chatbot
2:07:47 system so it wasn't voice but an
2:07:50 employee could log in say you know
2:07:52 what's my healthc care coverage is this
2:07:54 covered right and so it had to figure
2:07:57 out what coverage they were on
2:07:59 understand what all the you know what
2:08:01 was covered what
2:08:03 wasn't have stock answers in it and not
2:08:06 get anything wrong like it can't get it
2:08:08 wrong and so she created this whole
2:08:11 sophisticated system that had you know
2:08:13 stock answers and then the large
2:08:17 language model would make it
2:08:19 conversational and then um there were
2:08:22 all sorts of rules about if the employee
2:08:24 goes off the rails here's how you handle
2:08:27 that so there's a lot of work to do
2:08:30 there but I think it's solvable and man
2:08:32 the minute we can just talk to a bot
2:08:35 where you call up customer service the
2:08:37 phone gets answered like that and the
2:08:40 person is decently conversational and
2:08:42 actually is knowledgeable and can take
2:08:45 action holy [ __ ] I'm I'm there for it
2:08:48 all day long can you ask GPT 40 to
2:08:51 change CH your GPT background color and
2:08:53 GPT logo size
2:08:56 no no you can't it's there how it's how
2:09:01 it looks is how it
2:09:06 [Music]
2:09:07 looks actual answer
2:09:12 database um people building movie
2:09:14 studios well yeah it certainly affected
2:09:17 that Tyler Perry cancelling his uh 800 m
2:09:21 ion dollar Studio project that I would
2:09:24 say I would say it impacted the film
2:09:26 industry pretty
2:09:27 significantly and and Sora isn't even
2:09:30 out yet like the video model from open
2:09:33 AI isn't even out yet and you know it's
2:09:36 having an
2:09:37 impact and then
2:09:39 music Sony they sent um cease and desist
2:09:44 letters to like 70 or 700 AI companies
2:09:49 saying you trained you trained your ship
2:09:51 on our music you can't do that we're
2:09:54 suing your ass so we'll see we'll see
2:09:56 how that's going to play out it's
2:10:00 fascinating what ever became of the New
2:10:02 York City AI
2:10:05 chatbot I don't
2:10:07 know how do you suppose Security will be
2:10:09 handled with 4.0 Etc
2:10:12 blockchain I don't know yet I have no
2:10:15 idea and and I think Source Camp I
2:10:19 think if you look at the W on the wall
2:10:23 that Ilia and Yan left and the super
2:10:27 alignment team has basically been
2:10:29 dissolved I think Altman doesn't really
2:10:32 give a [ __ ] about security and and
2:10:36 privacy right now I think he's like I
2:10:38 want AGI I I want AGI out as as soon as
2:10:41 we can do it I I think they are I I
2:10:44 think open AI is
2:10:46 accelerating accelerating their efforts
2:10:48 I don't know if they're I don't I don't
2:10:49 know how close they are but I would
2:10:51 imagine they're a lot closer than we
2:10:52 think they are and I don't think they
2:10:56 are they're not going slow and being
2:10:59 careful I think that's I think that's
2:11:01 why those guys
2:11:04 left I need a bot to respond to all the
2:11:07 widow or friend requests on Facebook oh
2:11:12 God my wife gets those all the time
2:11:15 she's like what is it with like doctors
2:11:17 it's like all these doctors just like
2:11:19 hit her up on Instagram and followers so
2:11:22 she's like what is it with these people
2:11:24 can you explain again where the base
2:11:27 data is pulled from in these AI tools
2:11:30 large language models okay Firefly
2:11:32 you're going to get a good a good one
2:11:34 here and then and then I got to go but
2:11:36 okay so so what you're asking is is how
2:11:39 do these things work right okay let's
2:11:44 see I think what I'm going to do
2:11:52 oh I've got it up
2:11:55 here
2:12:00 yeah I'm G to show you how chat GPT
2:12:02 works if you don't know how chat GPT
2:12:05 Works here's how it
2:12:11 works couple of things you need to
2:12:15 know GPT if you don't know what GPT
2:12:18 stands for stands for generative
2:12:20 pre-trained Transformer
2:12:22 the generative means it's
2:12:25 generating original work it's not
2:12:29 copying and pasting it's one of the
2:12:31 biggest misconceptions about all of
2:12:33 these tools whether they're music ones
2:12:35 or word ones is that they are copying
2:12:39 and pasting from existing documents it's
2:12:41 not how it works these systems were
2:12:44 absolutely trained on all sorts of
2:12:46 information including copywritten
2:12:48 information absolutely 100% yes
2:12:52 the actual documents themselves once
2:12:55 they get embedded into these systems
2:12:58 actually cease to
2:13:00 exist they they the individual words
2:13:03 within the documents that they were
2:13:04 trained on turn into what are called
2:13:06 vectors mathematical points in
2:13:09 space in math they're they're points in
2:13:12 mathematical space thousand dimensional
2:13:14 mathematical space so anyway the the G
2:13:18 the generative part means it's
2:13:20 generating so everything that's
2:13:22 generated is actually
2:13:24 original um sort of word by word and
2:13:27 I'll I'll show you how it works in a
2:13:29 second when you understand how these
2:13:31 things work it is actually [ __ ]
2:13:33 remarkable that they work at all much
2:13:36 less what they do
2:13:38 pre-trained means they've been
2:13:41 pre-trained in this case in the case of
2:13:43 open Ai and Claude and all of the big
2:13:46 large language models they've been
2:13:48 pre-trained on as much of the public
2:13:51 internet as these companies could get
2:13:53 their hands on so they went out and they
2:13:55 would scrape as much of the internet as
2:13:57 they could and then they would get human
2:13:58 beings to tag all these documents and
2:14:01 you know get them into formats that that
2:14:03 these things could understand and then
2:14:06 they would embed them and then the T in
2:14:08 GPT stands for Transformer that's the
2:14:11 technology so 2017 Google invents this
2:14:16 thing called the transformer in a paper
2:14:18 called attention is all you need so 2017
2:14:21 is the seminal moment where the core
2:14:23 technology that all of this magical [ __ ]
2:14:26 we're
2:14:27 using was sort of born from it sprung
2:14:30 from that and what the Transformer
2:14:32 allowed is that you could you could
2:14:34 throw massive amounts of data and
2:14:37 massive amounts of compute power at
2:14:39 these models and they would work better
2:14:42 the more crap you threw at them um and
2:14:46 so the Transformer is just the
2:14:48 technology so generative it's making
2:14:50 [ __ ] from pre-trained data you don't
2:14:52 have to add your own data it's got as
2:14:54 much data as it can handle and then it's
2:14:57 doing it in this Transformer
2:14:58 architecture so that's GPT um you might
2:15:01 hear large language model llm that's
2:15:03 just basically what these things are
2:15:05 what they're called and then the latent
2:15:07 space Laten Space is really important it
2:15:10 it it's not important for you to know it
2:15:11 but it's just conceptually Laten space
2:15:14 is that mathematical space where all
2:15:17 these things called tokens exist and
2:15:20 tokens this is a a
2:15:23 two-dimensional um representation of
2:15:26 latent space right so X and Y and then
2:15:29 you've got these clusters and these
2:15:30 clusters are you know points in
2:15:33 mathematical space these
2:15:36 vectors and each one of these points
2:15:39 represents a token that's tied to a word
2:15:42 or a comma or a space or whatever it
2:15:44 might be and they're clustered by
2:15:46 semantic meaning so you might have the
2:15:49 word dog down in this black section here
2:15:52 and that means like the fuzzy thing that
2:15:54 sings When you play guitar and you like
2:15:55 to pet and then up in the green area you
2:15:58 might have the word dog and it means he
2:16:00 did his girlfriend wrong and and he's a
2:16:02 dog right and so so you can have the
2:16:05 same words that live in different
2:16:07 semantic
2:16:08 clusters this is two-dimensional very
2:16:11 simplified Vector space the real stuff
2:16:14 or latent space the the real latent
2:16:17 space is like thousand
2:16:19 dimensional like billions and billions
2:16:22 if not trillions of points in
2:16:25 mathematical space okay so how does this
2:16:29 [ __ ]
2:16:30 work when you create a
2:16:33 prompt that prompt gets converted into a
2:16:36 mathematical
2:16:38 probability for the next
2:16:41 token or the next series of tokens think
2:16:44 of it like the next word so so basically
2:16:47 what happens is you write a prompt that
2:16:48 says the quick brown fox jumps over the
2:16:50 lazy blank we know it's dog cuz we've
2:16:53 got large language models in our brain
2:16:55 and somehow our brains dig into the gray
2:16:58 matter and pull out the word dog here's
2:17:01 how chat GPT does it you write a prompt
2:17:06 and it creates a probability and what it
2:17:07 basically says is there's a there's a
2:17:10 very high probability that that one of
2:17:14 these light blue tokens is is the one
2:17:16 that has the highest probability of
2:17:18 being the right one and within this
2:17:20 cluster of light blue tokens there's one
2:17:23 of them that has the highest probability
2:17:25 and for the most part what it does is it
2:17:27 takes that vector or that token right
2:17:31 with with all of these uh mathematical
2:17:35 numbers points in space those are
2:17:38 associated with the word dog and then it
2:17:40 takes the word dog and sticks that in
2:17:42 your
2:17:44 sentence and it does that over and over
2:17:46 and over again so when it writes a whole
2:17:50 paragraph for you a whole article it's
2:17:53 doing it one word at a time it has zero
2:17:57 idea what it actually wrote It's just
2:18:01 saying mathematical point this
2:18:03 associated with dog Stick Dog here
2:18:05 mathematical point that it means cat
2:18:07 stick cat here
2:18:12 right and this is why they hallucinate
2:18:15 this is why when you ask it to give you
2:18:17 something you're like hey uh you know
2:18:19 you made a mistake there could you fix
2:18:20 that it's oh sorry about that let me fix
2:18:22 that for you and then it lies again
2:18:25 because it what it's not actually doing
2:18:28 it's not going back and saying did I get
2:18:30 it
2:18:31 right
2:18:33 qar that you you may have heard of when
2:18:35 they fired Sam Waltman there's this
2:18:37 thing called qar which is a reasoning
2:18:39 engine that does something more akin to
2:18:42 did I get it right like am I on the
2:18:44 right path to a right an actual correct
2:18:48 answer not just a high probability
2:18:50 answer but actually the correct answer
2:18:52 and that's reasoning and once these
2:18:54 things get reasoning then they're going
2:18:56 to get really really smart but anyway
2:18:57 that's how they work crazy now now let
2:19:00 me let me I'm going to show you one
2:19:02 other thing okay so where's my
2:19:09 box
2:19:13 okay okay you got your box so this is
2:19:16 this box
2:19:21 this is why chat GPT needs a marketing
2:19:23 person okay so you got you got GPT
2:19:28 4 you have chat
2:19:31 GPT they also call it chat GPT
2:19:35 4 and now we have chat
2:19:40 gp40 I'm G to need another box okay this
2:19:44 is
2:19:45 gp4 so gp4 is that Laten space gp4 is a
2:19:50 box that you take all of the internet
2:19:53 you embed it in this special way and you
2:19:56 dump all of those tokens into the box
2:19:58 all these points and mathematical space
2:20:00 get dumped into the
2:20:01 box this is kind of this hermetically
2:20:04 sealed data
2:20:05 set this this is why when you when with
2:20:09 with old chat GPT you'd say you know
2:20:11 tell me something new and it would say
2:20:13 my latest information was September 2021
2:20:16 cuz the box was kind of sealed in
2:20:19 September 2021 they've updated it since
2:20:21 but it's still this is this this is is
2:20:24 is called sort of the Frontier Model
2:20:26 this is the the GPT
2:20:29 model and it's called right now it's
2:20:32 let's just go with GPT 4 is
2:20:35 this this is chat
2:20:38 GPT it's an application it's when you
2:20:41 log into your web browser and you see
2:20:45 it's got here's the four questions you
2:20:46 could ask and it's got all the your
2:20:48 prompt history and it's got the little
2:20:49 box that's just a a web app that some
2:20:52 web developers wrote that says let
2:20:55 people type down here and when they hit
2:20:57 the submit button what it does is it
2:20:59 does an API call down into the model so
2:21:02 it asks for the model this is why this
2:21:06 is chat
2:21:07 GPT this is Microsoft co-pilot which is
2:21:12 also
2:21:14 gp4 but it's but they're why are they so
2:21:17 different cuz they're different
2:21:18 applications the open AI chaty PT is
2:21:22 just better than Microsoft's co-pilot
2:21:25 cuz they put too many um safety rules in
2:21:28 this and so it's a steaming piece of
2:21:30 [ __ ] but they're both calling the same
2:21:32 underlying model
2:21:35 now
2:21:38 um ah
2:21:41 okay this is
2:21:45 gp4 this is GPT
2:21:48 Omni GPT 40 it's a completely different
2:21:52 model and it's fancier it's got pictures
2:21:56 it's got It's got audio and so so so
2:22:00 what cat GPT has done we got to get rid
2:22:03 of Microsoft
2:22:05 here is you're still
2:22:08 using the chat GPT
2:22:12 app and you can switch to use this model
2:22:15 or this model when you switch to use
2:22:18 this model it kind of does this except
2:22:21 when you ask for pictures it goes out to
2:22:24 yet another model to make images so it's
2:22:27 not using its full capability all right
2:22:29 that's my box explanation there's more
2:22:32 boxes
2:22:36 now you had to ask right does that
2:22:39 answer your
2:22:42 question that was a lot I don't you know
2:22:45 I I really just kind of wanted to know
2:22:47 like is it plagiarism
2:22:51 you gave us a little much there but it's
2:22:53 fantastic Danielle I love the Box
2:22:56 demo I like it Picasso Firefly sort of
2:23:00 yeah so yeah so
2:23:04 basically
2:23:07 it's here here's a way to think about
2:23:11 it and
2:23:14 and the new voice stuff with the new
2:23:16 model actually changes this a bit it
2:23:19 doesn't mathematically but it it kind of
2:23:21 changes it a
2:23:23 bit when people when people ask
2:23:25 questions like well are these are these
2:23:27 tools
2:23:29 sentient like like you know do they have
2:23:32 feelings or do they know what they're
2:23:34 saying no they're just probability Eng
2:23:37 they're literally calculators they just
2:23:39 happen to be calculators that are
2:23:40 spitting out
2:23:42 words um and they're doing it in a in a
2:23:44 really sophisticated way but but you
2:23:47 know technically what they're doing is
2:23:49 just probability token probability token
2:23:52 probability token the new model when you
2:23:55 talk to it it not only understands the
2:24:00 words you're saying it understands how
2:24:02 you're saying them because it
2:24:04 understands
2:24:06 audio so it could understand a dog
2:24:08 barking or two different people talking
2:24:10 or you're pissed off or you're laughing
2:24:13 or you're
2:24:14 crying and it's going to be able to
2:24:16 respond in kind so they're they're
2:24:19 they're getting they're they're going to
2:24:21 start the experience of talking to them
2:24:23 is going to get increasingly human even
2:24:26 though technically they might not be
2:24:27 super
2:24:29 human a bot in a
2:24:33 box where does the data come from so the
2:24:35 data comes
2:24:37 from the public internet as much of the
2:24:40 public internet as they can get they
2:24:41 send out scrapers web scrapers so every
2:24:44 blog post all the Wikipedia there's
2:24:47 probably a fair amount of Nefarious data
2:24:50 in there that was pulled off the dark
2:24:52 web that they're not supposed to have
2:24:53 access to they probably used web
2:24:56 scrapers to pull [ __ ] off YouTube and
2:24:59 Instagram and I I I promise you I
2:25:02 promise you including
2:25:04 Adobe Adobe was taking the high horse on
2:25:07 this for a while but apparently Adobe
2:25:10 has now trained Firefly on Mid Journey
2:25:14 images which were trained on copywritten
2:25:16 information so I don't think anyone is
2:25:19 is pristine in and you know clean here
2:25:23 they basically went out and just got as
2:25:24 much information as they could embedded
2:25:27 it and put it in these tools for us to
2:25:28 go play with here's the
2:25:31 deal
2:25:34 if if any one of these companies if open
2:25:37 aai had ever gone to the New York
2:25:40 Times four years ago and said hey can we
2:25:43 have your
2:25:44 permission to use your data to train our
2:25:49 model the answer would have been no
2:25:50 every single person they would have
2:25:52 asked would have said no every single
2:25:55 one promise guaranteed you do that we'll
2:25:58 sue you or you pay us $100
2:26:00 million well there's a million companies
2:26:03 I need to talk to well you pay us $100
2:26:05 million and then you go figure out the
2:26:07 rest on your own that's what it would
2:26:08 have been the
2:26:10 conversation so this is one of those
2:26:12 things where you know they're asking for
2:26:16 forgiveness and they're trying to cut
2:26:17 deals after the fact because none of
2:26:21 this technology would exist if they
2:26:23 hadn't taken that approach in my
2:26:25 opinion
2:26:28 so you
2:26:29 know and and you know there there's
2:26:32 there's a part of me that's like oh well
2:26:34 it's you know they're taking you know
2:26:36 data and they're you know it's it's you
2:26:39 know they're they're profiting off of
2:26:41 other people's
2:26:44 work it it is it's very different than
2:26:47 say sampling from the 80s and '90s when
2:26:50 when hip hoppers started sampling music
2:26:53 and making hip-hop there they're
2:26:55 literally copying and pasting parts of
2:26:57 songs that's not the way these tools are
2:27:02 working and it's such a vast amount of
2:27:04 data that for me this kind of goes into
2:27:07 the into the neighborhood of it's kind
2:27:11 of like all of the information in a
2:27:13 library all of the information in a
2:27:15 museum why libraries and museums exist
2:27:19 is is so so that Humanity can benefit
2:27:23 from the output of humanity so there's
2:27:26 absolutely an argument to be said you
2:27:29 can you can sort of go down two paths
2:27:30 here you can say all work should be
2:27:33 protected and no one should be able to
2:27:35 use any information without paying
2:27:39 people or
2:27:42 and this is the output of humanity and
2:27:45 being put in these tools that are this
2:27:47 powerful benefits all of humanity
2:27:51 which is something necessary in the
2:27:54 world which one's right I don't
2:27:57 know these things exist
2:28:05 now the the the only thing that I know
2:28:08 that I that that is crystal clear to me
2:28:10 is that the intellectual property
2:28:12 lawyers are going to make a [ __ ]
2:28:16 fortune on both sides everyone's going
2:28:19 to be suing everyone for the next 20 30
2:28:21 40 years promise you that that I know
2:28:24 for
2:28:25 sure what's right here's here here's my
2:28:29 My Philosophy on this stuff is this I I
2:28:32 have a degree in acting I've written
2:28:33 seven screenplays like I know I know the
2:28:37 arguments for why this [ __ ] is evil and
2:28:40 awful and should not exist I get the
2:28:44 arguments I understand them
2:28:46 conceptually
2:28:48 and the tech technology exists and it's
2:28:52 not going away so this channel is is is
2:28:56 not about having the argument about is
2:28:58 it good or bad whether it's good or bad
2:29:02 is inde it's an independent
2:29:04 conversation from it's here what are we
2:29:07 going to do about
2:29:10 it the metaphor I like to use is arguing
2:29:14 about whether AI is good or bad is like
2:29:17 the Florida
2:29:18 legislature going into session and
2:29:21 passing a law making hurricanes
2:29:25 illegal like okay sounds good like let's
2:29:29 go into legislation let's debate are
2:29:31 hurricanes good or bad let's pass a law
2:29:35 and then we can tell the people good
2:29:37 news we've made hurricanes illegal
2:29:40 you're going to be
2:29:43 fine and the people on the beach are
2:29:45 like um what are those clouds oh that's
2:29:48 a hurricane
2:29:51 isn't it coming here
2:29:53 yeah but what are we gonna do what it's
2:29:56 okay we passed the law you're going to
2:29:59 be all
2:30:00 right
2:30:02 like that's what we're in the middle of
2:30:05 so this channel is about grab a [ __ ]
2:30:08 surfboard and let's start paddling out
2:30:12 to see if if we can catch a Sweet
2:30:16 Wave cuz I'd much rather be on the
2:30:19 surfboard than would in the Cozy beach
2:30:23 hut and and sort of all of the
2:30:26 consternation about is it excited
2:30:29 that people are having those
2:30:32 conversations and they're independent of
2:30:34 the fact that the shit's
2:30:36 here
2:30:37 so that's my two cents on the
2:30:41 matter grab a [ __ ]
2:30:44 [Laughter]
2:30:48 surfboard and and listen and you can
2:30:51 still go fight the fights go help pass
2:30:53 laws go sue people let them sue you like
2:30:56 it go do all that [ __ ] and this shit's
2:31:00 still coming it's here super glitchy
2:31:05 tonight thank you thank you Mr AI
2:31:09 laughing my ass off could you provide
2:31:13 more input about previous chat
2:31:18 GPT oh yeah
2:31:21 um too late to evacuate yeah it's too
2:31:24 late to evacuate smart man thank you
2:31:26 very much I don't think so it uses the
2:31:29 ho it doesn't just use a
2:31:35 part mid 90s the internet was basically
2:31:38 opened up to the general public via home
2:31:40 computers right yeah exactly yeah the
2:31:43 the parallels to the early days of the
2:31:45 worldwide web and the early days of AI
2:31:47 are like they're they're identical kind
2:31:50 of trajectories there's a lot more um
2:31:54 resistance and anger and fear about AI
2:31:56 cuz it's it's way more intense and it's
2:31:58 moving faster but back in the internet I
2:32:01 remember the the head of the music
2:32:04 recording Academy I think it was 1997
2:32:07 got up on stage and he said see those
2:32:09 two kids in the third row they're
2:32:12 stealing music on the Napster and you
2:32:15 shouldn't steal music kids cuz if you
2:32:18 steal music you're stealing from right
2:32:20 you know like I don't know if you
2:32:21 remember that speech but it was like in
2:32:23 the middle of the Grammys he was like
2:32:25 finger wagging at some young kids that
2:32:28 used Napster was like Jesus dude uh it's
2:32:34 still coming like you can be pissed off
2:32:36 about it but shit's going to
2:32:39 change and [ __ ] changed did it change
2:32:41 for the better I don't
2:32:44 know certainly not for music
2:32:47 Publishers I don't think it's helped
2:32:49 musicians but but musicians
2:32:51 now don't necessarily need a music
2:32:55 publisher to get out there they have
2:32:57 things like Tik
2:33:00 Tock
2:33:01 right just different it's just different
2:33:04 but there's going to be every single
2:33:07 industry is going to have that kind of
2:33:09 disruption you know the tsunami is
2:33:12 coming for every every creative industry
2:33:15 every knowledge worker anything that's
2:33:17 got any kind of computer mediation is
2:33:19 about to be completely
2:33:21 upended over the next 3 to 5
2:33:24 years
2:33:26 so why we're here why I'm here every
2:33:29 [ __ ]
2:33:32 night is
2:33:35 because being in this conversation
2:33:38 actually the questions you ask what I'm
2:33:40 trying to explain and answer and talk
2:33:44 you know around being in that
2:33:46 conversation over the next 3 to 5 years
2:33:49 is actually super important important if
2:33:50 you're in the conversation I think
2:33:52 you're going to be okay if you're
2:33:54 sitting on the sidelines like AI is evil
2:33:57 it's just evil I don't know who these
2:33:58 people think they are you're going to
2:34:00 get your [ __ ] ass handed to
2:34:03 you cuz it's independent of what you
2:34:07 think of it it's still coming so you
2:34:09 might as well
2:34:11 learn you can still resent the [ __ ] out
2:34:14 of
2:34:15 it I go through these bizarre swings of
2:34:19 like holy [ __ ] this is is the best thing
2:34:20 ever and holy [ __ ] what the [ __ ] am I
2:34:22 going to do what's this mean for my
2:34:24 company I got a 12-year-old company that
2:34:27 does automated video
2:34:32 production Sora
2:34:36 anyone hey Jen
2:34:44 avatars it doesn't matter what I think
2:34:47 it's still
2:34:48 coming so my job is to figure out okay
2:34:52 here's what my company's good
2:34:54 at here's where the value of that goes
2:34:58 up in the future here's where the value
2:34:59 of that goes down in the future let me
2:35:02 do more that you know goes up in
2:35:05 value and that's what we're going to
2:35:07 have to all confront with ourselves
2:35:09 personally with our
2:35:11 businesses so anyway all right I'm tired
2:35:14 I'm I'm Crispy I'm crusty
2:35:20 ah I'm prepping my kids for the future
2:35:23 by encouraging them to get law degrees
2:35:25 there you go intellectual property law
2:35:27 seriously I think I think the big skill
2:35:30 going into the future is is critical
2:35:32 thinking critical thinking is the big
2:35:34 thing so anything that you can do to
2:35:36 prepare your kids to like use your
2:35:38 [ __ ] brains learn what taste is learn
2:35:41 about judgment learn why something's
2:35:44 better than something else get good at
2:35:46 that stuff cuz the AI is going to do do
2:35:50 all the work and we're going to need
2:35:51 humans to figure out what's
2:35:55 good lime Wire yeah that's what it
2:35:59 was lime Wire yeah
2:36:04 exactly and pottery yeah learn Pottery
2:36:08 cuz Potter Pottery is going to be in
2:36:10 demand people are going to be like we
2:36:12 need more cups and
2:36:16 plates and Cory will make that for you
2:36:18 she's down the street she she's the one
2:36:20 with the she'll be muddy she'll be a
2:36:22 little covered in mud just that's she's
2:36:24 super nice she's super sweet she'll sing
2:36:26 you a song
2:36:29 too get some rest all right I'm out of
2:36:32 here follow my [ __ ] Channel subscribe
2:36:35 to this thing if you want to support if
2:36:37 you want to support programming like
2:36:40 this don't donate to the Evelyn and
2:36:43 Harry Woods Foundation or whatever the
2:36:46 [ __ ] going on subscribe to my lives
2:36:48 we'll miss you this weekend I know I'll
2:36:50 miss you guys too um thanks your
2:36:53 regulars thank you subscribers thanks
2:36:55 mods thank you all for hanging out
2:36:57 tonight happy date
2:36:59 night and uh yeah have a good weekend I
2:37:03 don't know what I'm going to work on
2:37:04 this weekend something I'll figure
2:37:05 something out I'm working on some Salon
2:37:08 strategy stuff that I'm pretty excited
2:37:10 about so I think we'll do that so all
2:37:12 right who's in here tonight dragonfly
2:37:14 Alchemy Joey Corey Sandler Joker TK good
2:37:18 night Vicky good night Dan
2:37:20 Joy py good to see you good night
2:37:22 everybody laughing Thunder Silver Fox
2:37:26 Shane saw side hustle Mimi Dr J Man a
2:37:30 big pile tonight the mods are fantastic
2:37:33 they are thank you Firefly good question
2:37:35 by the way thank you I don't know if I
2:37:37 answered it but I think we got made some
2:37:39 progress maybe joy in the
2:37:42 house
2:37:46 uh all right Tobias good night I think I
2:37:50 got everyone Teton skier all right peace
2:37:53 out I will see you Monday but I'll
2:37:57 probably put out some videos over the
2:37:59 weekend I'm trying to do these morning
2:38:01 Walkin talks so hopefully I'll get a
2:38:03 couple of those done this weekend
2:38:05 another great night thank you ton skier
2:38:07 I appreciate that we need to meet in
2:38:11 2024 all right peace out everybody bye