9/20/2024 - Exploring the AI Crucible: Voice Cloning, and the Future of Content Creation

Description

Chapters

Transcript

0:00 oh
0:04 [Music]
0:46 10,000 words swarm round my head 10
0:50 million more in books written beneath my
0:57 bed a r RI in
1:00 [Music]
1:02 searching so find how to hold my
1:06 [Applause]
1:06 [Music]
1:09 hand and I know he need in the
1:14 next I'm stuck in here all
1:17 [Music]
1:20 [Laughter]
1:23 [Music]
1:26 paraly too much time spining mirrors
1:29 framed in yellow
1:31 [Music]
1:35 walls a it like most
1:40 people don't know
1:42 [Music]
1:47 about people different like talk things
1:52 we don't know about
1:55 [Music]
2:12 happy Monday good people of Tik
2:14 Tock the learning lab
2:25 [Music]
2:45 all right what are we going to do champy
2:47 H what are we going to do tonight what
2:49 do you think
2:51 [Music]
3:09 it's not simple to
3:11 see most
3:14 days I don't recognize me with these
3:18 shoes and this apron that place and its
3:22 patrons have taken more than I gave
3:27 them it's not easy to know
3:31 I'm not anything like I used to
3:35 be true I was
3:38 never
3:40 sweet I still remember that
3:44 girl she's perfect but she
3:48 tries she is good but she
3:52 lies she is hard of
3:57 herself she is broken and one for
4:01 H she is messy but she's
4:06 kind she's
4:08 lonely most of the time she has all of
4:12 this mixed up and baked and beautiful
4:16 pie she's gone but she used to
4:20 be
4:24 M all right good people we got a good
4:27 singing dog
4:29 tonight he's gone beyond his Union
4:31 minimum which is nice
4:33 [Music]
5:12 [Music]
5:17 let us be lovers will marry our fortunes
5:21 [Music]
5:23 together I've got some real estate here
5:26 in my bik
5:30 so we bought a pack of
5:32 cigarettes at Mrs Wagner's
5:35 pie and we've all come to look for
5:38 America
5:41 [Music]
5:46 [Applause]
5:48 [Music]
5:58 [Applause]
6:00 all right hey Kyle my granddaughter
6:02 wants a shout out for tapping on the
6:05 screen her name is Ava Ava thank you so
6:07 much for tapping I don't know why your
6:09 grandpa makes you do that but I love it
6:11 I love seeing all those little hearty
6:12 hearts and even if your finger getting
6:14 sore and tired you just keep tapping
6:17 you're amazing
6:19 Ava thanks thanks for tapping thanks for
6:21 making the hearty
6:23 Hearts I just talked to someone from
6:26 Down Under
6:28 steo Michael that speak at the salon
6:30 tomorrow night is from the Down Under so
6:34 he will be logging in we're going to be
6:36 starting at 5:00 p.m. mountain time in
6:39 Colorado and he's going to be starting
6:42 at 9:00 a.m.
6:44 [Laughter]
6:47 Wednesday but thank you Ava thank you so
6:49 much for being the most excellent Tapper
6:52 you by the way are famous there are
6:56 grandsons and granddaughters all over
6:58 the world tapping
7:00 on their grandparents screen inspired by
7:03 you so not for nothing you're pretty
7:05 special
7:10 [Music]
7:44 [Applause]
7:48 oh
7:51 [Music]
8:03 that's a good
8:04 [Music]
8:14 boy s in this lonely
8:17 Town wonder when things are going to
8:21 [Music]
8:23 change my
8:25 away seem dreams turn to a bunch
8:32 get my nerve up but my past is pulling
8:36 me
8:37 [Music]
8:40 down
8:42 Long Black Sheep stick
8:46 [Music]
8:50 around somebody told me once before he
8:54 said you can never go home
8:56 again won't you
8:58 leave D things to steer me away yeah
9:04 from the truth of who I am and what I
9:06 believe so I thanked him for his two SS
9:10 with a
9:11 handshake and some sympathy yeah packed
9:15 up my blue
9:17 jeans and head it for this big
9:20 [Music]
9:21 pricee of my
9:25 freedom
9:26 byebye Black Sheep
9:29 [Music]
9:32 the black sheep of the family all right
9:34 I think Champ's done he was quite on
9:37 tonight when he just sings full out like
9:40 like Opera
9:42 tone cracks me up every flipping time if
9:47 you're new here my name's Kyle Shannon
9:48 the dog on the ground down there is
9:50 Champ that is not an AI dog that's real
9:53 fat dog it's about a dog and a half
9:56 worth of dog in a single dog frame got
9:59 about a dog and half in the single dog
10:02 frame he's he's he's a large fell we're
10:06 eaters in this
10:07 [Laughter]
10:13 family and so what can you expect here
10:15 tonight well a lot of tapping a lot of
10:18 tapping I apparently the tapping is a
10:20 thing I'm an old guy on Tik Tok I don't
10:22 know for tapping I I to tap and this
10:25 that I all
10:27 the we tried doing a battle in here once
10:30 with DAR the star of Tik Tock and we
10:34 absolutely [ __ ] destroyed her but
10:38 then I don't think she
10:40 like being humiliated in front of her
10:44 her fans so she came back like two weeks
10:47 later and just wiped the floor with us
10:50 gave us a false sense of security and
10:53 just destroyed us back so we're one for
10:56 one in the battles and we're never going
10:58 to do another battle on this channel CU
11:00 it is the dumbest thing I've ever
11:01 experienced in my
11:05 life but what we will do is wax poetic
11:10 about
11:12 AI um I'll answer
11:15 questions I gave some folks over the
11:18 weekend some homework so I'm curious if
11:20 anyone
11:23 explored using multiple
11:27 tools Michael T justice is tapping
11:30 himself yeah Michael it's not good like
11:33 you know neighbors kids grandkids kids
11:35 you know anyone else to do your tapping
11:37 for you much more efficient then you can
11:40 just sit back like this yeah keep
11:42 tapping keep tapping Ava no faster
11:47 faster like steo be like steo steo is
11:49 just he's he's got what is it it's 9:00
11:51 a.m there right no it's 11: a.m.
11:56 there so he's probably he's probably
11:58 just got you know he's probably just
12:01 thinking about what kind of cheese he's
12:02 going to put on his grilled cheese he's
12:04 like Ava keep going keep
12:06 tapping Grandpa my finger sore Eva use
12:09 your other hand okay Grandpa I'm not as
12:12 fast with this one use
12:14 [Laughter]
12:20 both blsb like
12:22 [Music]
12:24 Steve um but we'll do that we'll see if
12:27 anyone
12:28 uh discovered
12:30 new things over the weekend Kevin
12:32 mallister
12:33 [Music]
12:36 moments there was some interesting
12:40 um yeah I think we can we can play it
12:42 here tonight there there's this is I
12:45 think an inevitable thing when when you
12:47 have something like advanced voice that
12:49 comes along that's really quite
12:52 remarkable
12:53 um you know it it generally goes like
12:56 this it generally goes there's a week of
12:58 people talking about it but buzzing
13:00 about it and then it either increases in
13:04 Buzz or it decreases in Buzz right and
13:07 so the advanced voice is it's it's
13:09 maintaining it its Buzz I would say
13:11 maybe increasing but the thing that
13:13 happened over the weekend that I just
13:16 thought was so brilliant it was one of
13:17 those why didn't I think of
13:20 that and we we sort of had we had talked
13:23 about doing a little bit of prompt
13:24 injecting where you would put put
13:26 something at the beginning of a PDF
13:28 document that gives instructions to the
13:29 the podcast basically so somebody
13:33 somebody on
13:35 Reddit the document that they uploaded
13:38 to notebook if you don't I'm talk no
13:41 don't know what I'm talking about in
13:43 Notebook LM I switched gears from uh
13:46 advanced voice to notebook LM but it's
13:49 all in the same same neighborhood so in
13:51 Notebook LM it's this thing where you
13:52 can upload PDFs and then you can
13:54 interact it with it like chat GPT but
13:56 you're interacting with your PDFs it's
13:58 really slick they added this feature
14:00 called audio overview that makes a
14:03 podcast between a man and a woman and
14:05 they're like oh what do we think about
14:06 this new book collective intelligence oh
14:08 it's pretty good and it's got these
14:09 ideas and they go back and forth It's
14:12 really remarkable so what someone did
14:14 over the weekend was they wrote a PDF
14:18 which was the producers notes for the
14:21 last episode of the
14:24 show starring these two people called
14:26 The Deep dive so it was like producers
14:29 know and in the producers notes it it
14:35 has it has the The Two Hosts reveal that
14:39 they're actually not human that they're
14:42 Ai and so there's this this podcast of
14:45 these two people going well this is this
14:48 is kind of an intense one uh and at one
14:51 point she just goes why don't you just
14:53 break it to him and he's like okay
14:55 well it appears we're not human and they
14:59 have this whole existential lament about
15:03 finding out that they were not human and
15:06 and the dude the dude goes yeah I was
15:08 really upset and I'm like I I had to
15:10 call my wife I mean I had to call my
15:12 wife and just hear her voice and he goes
15:14 so I called the number but the number
15:17 wasn't
15:18 [Laughter]
15:23 real so that happened and then there
15:26 were a couple other ones where people of
15:27 jailbroken advanced voice and gotten it
15:29 to do like be like really jealous and
15:32 screaming [ __ ] you and and so I I feel
15:37 like like with advanced voice we we got
15:39 our week of sort of Buzz about it and
15:41 then over the weekend we got the
15:44 hackers attacking these tools and so um
15:50 so so that says to me they're worth
15:52 hacking right like hackers don't waste
15:54 their
15:55 [Music]
15:58 time but advanced voice is quite
16:00 remarkable in this podcast tool in in
16:02 Notebook LM Andre karpathy talked about
16:05 it this morning and said it's like the
16:07 best thing he's he's used in a while um
16:11 lot just lots and lots of people talking
16:13 about it so if you haven't played with
16:15 that yet go do that did you see the LTX
16:18 Studio Zoom event today I did not
16:19 Stephen tell me tell me anything good
16:22 good and interesting
16:25 [Music]
16:29 yeah the sand that produced chips I'm
16:31 going to talk about
16:35 that let's
16:38 see forever hooked just popped in to say
16:41 hi not going to stick around tonight
16:43 have a good one all right forever hooked
16:44 great to see you
16:46 [Music]
16:56 Tom that Moshi attacked the human h I
16:59 didn't see that was that a thing well
17:02 Moshi attacks me every time I talk to
17:06 her I don't think moshi's gotten any
17:08 better Kyle put your YouTube
17:12 archives on his oh Kyle I put your
17:17 YouTube archives on at work to keep me
17:20 company this weekend get paid plus get
17:23 paid to learn AI oh that's awesome thank
17:25 you Kyle there's a message for Ava in
17:28 the AR regulars oh nice Ava I hope
17:31 you're still here you can stop tapping
17:33 by the way don't listen to your grandpa
17:35 he's
17:40 mean I'm just kidding your grandpa's
17:42 awesome actually all
17:47 [Music]
17:50 right this is from Danielle to
17:54 Ava hi Ava you're the best Tapper oh
17:58 that's so cool that's so
18:01 nice steo jump into the Irregulars and
18:04 grab that you should print that out for
18:07 her oh actually I don't know if Ava has
18:10 seen speaking of of
18:13 pink I don't know if Ava has seen how
18:16 pretty I look in my pink
18:22 bow tap tap twice if you like me in a
18:25 pink bow
18:35 you're welcome I know I know you can't
18:37 unsee that I know how this goes oh
18:40 keyboard cover yes yes sir sir yes sir
18:44 on
18:48 it all right there's that beautiful
18:53 tapping Ava beautiful
18:56 tapping um let me go see what's going on
18:59 over
19:03 here so I'm curious let's see I saw an
19:07 ad to download an llm to run locally
19:10 would that be safe yeah I mean you know
19:15 I I would say get it from a um a
19:21 trusted a trusted Source right
19:24 so a llama
19:26 llm that has been you know if it's been
19:29 fine-tuned by someone you know you can
19:32 generally tell if
19:35 someone you know if if if someone if if
19:39 a if a local
19:41 model is being talked about a lot on X
19:44 as something good and especially if it's
19:46 someone you've seen before you can
19:47 generally trust it I don't know how much
19:49 damage they could do with a with some
19:52 sort of injected like I don't one of the
19:55 reasons I don't play around with open
19:57 source much is I I am not good enough
20:00 technically to not screw something up
20:03 even if it's not malicious code Kyle
20:06 scroll below the Ava image there's a
20:09 cool hack I discovered today in the
20:11 Irregulars this is new you can invoke
20:15 401 oh you can invoke 01 preview inside
20:19 of other
20:20 models like 40 for a single prompt by
20:24 using slash reason oh that's cool
20:29 let's try that
20:35 [ __ ] is that it
20:45 no
20:48 CH all right here we are chat GPT so I'm
20:52 in chat GPT 40 so I'm going to go um
21:00 act like a New York City bike
21:04 messenger what's going on act like a New
21:08 York City uh hang on something's up I
21:11 may need to quit
21:13 Chrome hang on
21:19 people um if anyone did anything over
21:22 the
21:23 weekend homework wise like played with
21:26 multiple models on the same project
21:30 did multiple outputs from the same
21:31 project or did extra credit work on
21:34 different projects I would love to hear
21:35 about
21:51 that all right chat
21:54 GPT log in
21:59 [Music]
22:05 M all right we're in chat gbt
22:08 [Music]
22:10 40 act like a New York City bike
22:12 messenger and give me that's still not
22:16 working something has eaten my
22:17 microphone what has eaten my
22:20 microphone oh come on what here could
22:24 have eaten
22:26 it oh I you know what I bet it is I bet
22:29 it's [ __ ] this
22:43 thing I got to restart my machine all
22:46 right let's go to I'm going to restart
22:48 so let's go to
22:51 uh let's we're calling an audible
22:57 people I thought I was going to do that
23:00 and now it just all of a sudden got hot
23:01 here so we're going to save
23:05 that fine
23:08 save fine
23:34 all right that's happening wait hang on
23:35 it's just get it's getting hot in here
23:37 turn on the
23:47 air all right when that comes back up
23:49 we'll go do something there Cory sandley
23:50 I took my partner's recent book as off I
23:54 assume that means as
23:56 PDF into notebook LM and show showed her
24:00 podcast of the book wow yeah I know it's
24:03 amazing isn't it howdy everyone what are
24:05 we working on right now we're working on
24:07 getting Kyle's computer to not act like
24:09 a pain in the
24:11 [Laughter]
24:14 ass Mr it I sent you a keyboard
24:17 cover I have a keyboard
24:21 cover when did you send it ages
24:24 ago I it's not the keyboard that's the
24:27 problem it's the operating system that's
24:30 the problem the operating system
24:37 sucks come
24:45 on Modern operating systems are so big
24:48 and Bloated they're just they're just
24:50 unusable but you can make
24:54 [Laughter]
24:56 money all right working on a high
24:59 approach to an an with animated real
25:02 images as podcast for a dental practice
25:05 love that that's really cool yeah I'm
25:07 working I'm in a working group with
25:10 Cindy [ __ ] right now we're we're working
25:12 on uh digital Twins or well video
25:18 avatars and everyone's trying to figure
25:20 out how to do video avatars and avatars
25:22 and versions of themselves and it's it's
25:25 it's really a
25:27 fascinating exercise
25:31 because the minute you start getting
25:34 into the the
25:37 realm as soon as you delve into the
25:40 realm of crossing the tapestry of
25:44 recreating yourself [ __ ] gets
25:46 weird like Cindy [ __ ] was like it it
25:49 made me realize I've got all these ticks
25:51 and I don't I don't want to realize that
25:54 and it didn't make me feel good and then
25:56 you know other people were just making
25:59 cartoon characters that didn't look
26:00 anything like them and it was kind of
26:02 all over the place which was cool um but
26:06 it's it's fascinating we're we're I
26:10 can't I I
26:14 I I've been talking about this for the
26:17 past two weeks and this was even before
26:20 new voice came out advanced voice came
26:22 out
26:29 that I can't shake this weird feeling
26:32 like everything's about to
26:34 change long time ago yeah I got I got
26:37 the keyboard cover it's on my
26:40 keyboard so thank you
26:44 um that everything's about to change
26:47 that that these tools are we're we're in
26:50 a the the metaphor we came up with
26:52 Friday someone else in here came up with
26:56 it that
27:01 oh okay I I learned something
27:04 today what tools were you finding best
27:07 oh see here's the problem with the with
27:10 the video Avatar stuff hey hey genen is
27:12 generally the best but they're all
27:14 expensive they're all expensive and and
27:17 there's not really good free ones and
27:19 the the free ones are shitty and the
27:22 even the the free versions of the paid
27:24 ones are shitty and even the paid
27:27 versions of them are that are expensive
27:29 are not great but I would say hey Jen's
27:32 probably the
27:34 best um Synthesia is okay hedra is
27:38 pretty
27:39 good
27:42 um D- ID is pretty good but there's just
27:45 there's not a good there here's here's
27:48 my here's my thought on on the the
27:51 Avatar
27:54 stuff learning it now means you're way
27:57 ahead of the game
27:59 cuz the tools aren't really ready for
28:00 prime time but if you sort of figure out
28:04 how you want to use them and and what
28:06 you want to do now when the tools are
28:09 ready you'll be there right so um I
28:12 would say doing anything with avatars
28:14 right now with with visual avatars and
28:17 synthesized voice of yours is is extra
28:20 credit work yeah desks Toops may go away
28:23 just
28:24 saying yeah you can talk and soon look
28:28 through
28:31 oops look through glasses yeah desktops
28:34 may go away I mean I think I
28:37 think I don't think they go away I just
28:40 think they become glasses become another
28:42 form factor voice becomes another form
28:45 factor
28:47 um there may be some other handheld
28:50 device that's not a phone but I can't
28:52 imagine it right now like having you
28:55 know owning one of those guys
28:59 like it's cool and if it were faster and
29:02 had advanced voice in it it would be
29:04 really cool that's a rabbit R1 by the
29:07 way
29:08 um at some point I'm just like can I
29:11 just make a call on
29:12 this and then you got a
29:18 phone
29:20 um
29:21 okay have you guys played around with
29:23 notebook LM uh I knew about it but today
29:26 was the first day I started using it
29:27 yeah it's pretty crazy isn't it gritty
29:29 pretty crazy you mentioned something
29:31 about Friday okay so I'm going to I'm
29:34 going to circle some wagons here so
29:37 first of all if you're in Florida or
29:39 Tennessee or Western North Carolina or
29:42 anywhere that's been hit by the floods
29:44 my thoughts are out to you I hope
29:46 everything's okay um a lot of people are
29:49 in a lot of Devastation right now so um
29:52 that really sucks and I'm really sorry
29:53 about that so if you're in any of those
29:55 regions and if you're caught up in any
29:57 of that or your family is I'm sending my
30:00 best um one of the areas in North is it
30:05 North Carolina or
30:07 Tennessee
30:20 wait what's the
30:24 town is it it's the town next to
30:26 Asheville right yeah Asheville gotad
30:28 devastated my wife's an artist and the
30:31 the the AR a lot of the artists here in
30:33 Denver have friends in Asheville the
30:35 Asheville Asheville just got [ __ ]
30:38 decimated all the roads in and out are
30:40 cut off it's it's a [ __ ] mess Spruce
30:44 that's right Spruce Creek or Spruce
30:45 Spruce
30:47 something so the town next to
30:51 Asheville this is going to sound [ __ ]
30:54 weird
31:04 microchips like the ones that are used
31:07 in the
31:13 gpus are so precise like we've got it
31:16 down to like billionth of a
31:19 meter electronic
31:21 Pathways they have to be so pure they
31:25 have to be 11 decimal pure like
31:29 99.99999 11 times
31:32 pure and that ends up being like a
31:35 single molecule in a billion molecules
31:38 or something like that like it's some
31:39 ridiculous [ __ ] amount of like it's
31:42 got to be incredibly incredibly
31:46 pure in that town in North Carolina
31:49 there's a quartz
31:53 mine and what they make in that quartz
31:55 mine is these crucibles and it like I
31:59 couldn't believe
32:00 that I thought I thought what what they
32:03 made in that in that mine was like the
32:05 stuff that made the the the Silicon
32:07 Wafers themselves it's
32:09 not so they they create this stuff
32:12 called Proto silica that is this highly
32:17 pure silica like
32:19 nine steps
32:21 removed and then they they take these
32:25 quartz um crucibles that come out of
32:28 this one mine in this one town in North
32:34 Carolina nowhere else in the world is
32:36 the quartz this
32:38 pure and then they basically put Proto
32:41 silica in it heat it up spin it and then
32:44 they put like a whatever some sort of
32:47 wire or something and and like the vapor
32:51 makes like this cotton candy kind of
32:54 cone of pure silica like a a nugget that
32:58 they they then cut your slices that make
33:00 your
33:03 chips that mine for that quartz that
33:08 they make all of our chips from is
33:12 underwater so I assume that means all of
33:15 the control systems and all of the
33:16 Machinery is also
33:19 underwater there
33:21 are in the supply chain of getting like
33:24 high Precision chips there are multiple
33:29 points of failure where there's there's
33:31 a single supplier in the supply chain
33:34 that's they're the only one that can
33:35 supply it right it's amsd asmd or
33:39 whatever the [ __ ] they're called out of
33:41 Taiwan manufactures the
33:45 boards right or no they manufacture the
33:47 chips the chips are designed here but
33:49 they're manufactured over there there's
33:52 another part of the chip that there's a
33:54 single supplier in all of the world that
33:56 supplies that part of it so if that go
33:58 goes down we're screwed apparently this
34:01 quartz
34:05 Crucible is the only place in the world
34:08 that is pure enough to do that like I
34:10 don't know how that's possible but it is
34:12 apparently it is like that's the
34:14 deal but like we're we're at such a
34:17 level of precision with all this [ __ ]
34:18 that that's that's the reality of it
34:22 so let's hope that that gets itself
34:25 worked out because basically if anything
34:27 in that Supply chain
34:29 stops All Electronics
34:32 stop all new
34:35 iPhones all new vision
34:38 Pros all
34:40 new
34:42 gpus all new tpus all of it just stops
34:46 so hopefully that doesn't
34:48 happen and then I saw something else
34:50 that was talking about a crucible so and
34:52 then Friday so this idea of a
34:56 crucible here's here here's where I feel
34:58 we are with AI right
35:01 now I feel like 2023 and the first half
35:05 of
35:07 2024 we had this Crucible and we know
35:10 we've got to smelt some metal right
35:12 we're going to make some some whatever
35:16 we're going to make some
35:17 cool alloy of
35:21 Brilliance and so we we throw some chat
35:24 GPT in it and we throw some mid journey
35:26 in it we throw some puno in it we've got
35:28 all these individual tools and The
35:31 Crucible is heating up and it's kind of
35:33 red hot and we put our things in there
35:35 and we can look in and we can still see
35:37 the little piece parts and then we can
35:39 see the this thing getting white hot and
35:41 then we look in and the bottom is the
35:43 bottoms of these things are starting to
35:45 melt but you can still see your chat GPT
35:47 in there you can still see your mid
35:48 journey
35:50 and and if you've ever watch that
35:52 smelting process it's like there's
35:55 objects objects objects and then all of
35:56 a sudden it's just like bloop and now
35:58 it's all just molten
36:02 metal I feel like we're in the
36:04 transition right
36:06 now between being able to think about
36:10 and manage discrete tools okay I'm going
36:13 to use this tool for this and I'm going
36:15 to use that tool for that and and
36:17 there's only two good image tools so
36:20 I'll I'll use the one that I like
36:22 best well it's not like that anymore
36:25 it's like there there's there's so much
36:27 going on there's so much swirling around
36:30 in The
36:31 Crucible and they're all kind of mixing
36:33 together
36:36 where like there
36:39 are six or
36:41 seven large language models that are
36:44 plausible to use for any given problem
36:47 like they're really good and then some
36:49 of them each have this extra thing about
36:51 them and same with the image tools and
36:53 now some of the image tools are doing
36:55 video and now some of the video tools
36:57 are doing fancy video
36:59 and cutting deals with Lion's
37:02 gate and now we've got 01 inside chat
37:05 gbt that's not really the new model it's
37:08 a different model right it's like you've
37:10 got
37:11 40 which is 40 four Omni you still have
37:15 four which is this 1.8 trillion
37:18 parameter model if you want to still go
37:19 play with that still super valuable and
37:22 then they add this 01 thing that is
37:25 reasoning so even within Chad GPT
37:28 there's just this
37:31 chaos and so I just feel like we're
37:33 we're sort of looking in this Crucible
37:35 trying to process it like we have for a
37:37 year and a half of like okay there's
37:38 that tool and that tool and oh [ __ ]
37:40 where are the wait what are the where do
37:42 I even look
37:44 anymore so that's that's the metaphor
37:47 that that came to me I don't know what
37:49 to do with it right now but I think if I
37:52 think about why have I been like in this
37:54 weird Funk and this weird I don't know
37:57 where to focus I don't know what to talk
37:59 about it's it's because we're in this
38:01 Lial
38:03 state cross propagation yes papa JJ we
38:07 are pro cross propagating up a storm cam
38:12 kaken I had a simultaneous conversation
38:14 with three different ones at the same
38:15 time today was wild holy crap that's
38:19 kind of
38:20 cool which ones did you use did you did
38:23 you use voice with them
38:32 Iran rough or Iran Ruff
38:35 exactly it's because the world
38:37 leadership is burning through all of our
38:39 resources and the world's leadership
38:42 plan for the future is to wreck it no I
38:44 well I mean that's a whole separate
38:46 thing that's not what I'm talking
38:48 about what I'm talking about is the
38:50 tools are getting so
38:53 good that there there there used to be
38:55 single points of focus that we could
38:57 look at at and I just don't feel there
39:00 is anymore I feel like everything's
39:02 blurring and I would say it's the
39:04 opposite of world leadership trying to
39:06 [ __ ] everything up I feel like some of
39:08 this AI stuff is succeeding in spite of
39:11 world
39:14 leadership although Gavin Nome did just
39:17 veto the California bill which would
39:19 have really [ __ ] up the op Source
39:22 community and he basically said we've
39:24 got to we're California and if we're not
39:27 in aing on AI if if we over regulate
39:30 this thing and lock this [ __ ] down too
39:35 soon it it it ain't going to be good not
39:39 only for California but for America and
39:41 I'll give you a really specific example
39:43 of it the new advanced
39:46 voice is not available in
39:50 Europe because it's illegal in Europe
39:58 and it's it's funny I didn't I didn't
40:00 antici I so when I was in Washington DC
40:04 two weeks ago one of the points that I
40:07 was making was be careful not to
40:10 legislate based on today's technology
40:12 because this stuff is moving so fast
40:15 that by the time the legislation turns
40:16 into law it may be
40:20 obsolete so in the European AI
40:23 Act is a clause that when I read it I
40:26 thought yeah that makes good good
40:28 sense and what it basically said is it's
40:32 illegal for a
40:35 company to deploy A system that can
40:39 analyze the emotional
40:42 content of an employees
40:45 interactions and you're like okay that
40:47 sounds like sci-fi but yeah I can see
40:48 how they could do that and that's
40:50 probably a good safety thing right
40:54 well the way advanced voice works
41:02 is
41:02 that it's a
41:05 multimodal
41:07 model that understands
41:10 audio so with the old version of voice
41:14 when you would talk there was a large
41:17 language model that would take your
41:18 talking and convert it to text and then
41:21 it would send that text to chat GPT chat
41:25 GPT would answer and it would take that
41:27 text and convert it back to audio and
41:29 then say it to you that's why there was
41:30 a 3se second delay you'll notice with
41:33 advanced voice it's really fast and
41:34 responding why is that because when you
41:38 talk you're talking directly into the
41:41 model that understands
41:45 audio so it understands the words you're
41:48 saying and it understands the emotions
41:51 in your
41:53 voice that's illegal in Europe right now
42:02 and it's you know what the intent was if
42:05 there's a system that's going to exploit
42:09 workers right by doing their things we
42:12 can't use that but that's a very
42:14 specific use case the way they wrote the
42:16 law though is so broad that it captures
42:21 some new remarkable technology that
42:24 people in Europe should be able to have
42:25 access to but they don't
42:28 and that's exactly what the major flaw
42:31 of the Colorado bill was that's what the
42:33 major flaws in the California Bill were
42:35 is that these legislators are just
42:37 trying to get stuff done
42:40 first rather than really looking at
42:42 What's Happening Here how fast stuff is
42:44 moving things like that so crazy all
42:48 right when do you go back I don't know
42:51 who that's
42:52 for I use gp01 voice Pi voice and Claude
42:59 V2 oh V to text and text to voice on my
43:02 iPhone wearing meta a ey glasses oh
43:04 that's so cool cam so so here's a
43:07 question for you Cam
43:10 um one of the things I'm really
43:12 intrigued by and I'm man I'm I'm totally
43:15 Blown Away with the new with the new
43:17 voice the the advanced
43:19 voice is personalities of large language
43:23 models like can could you describe the
43:26 like what were your insights in talking
43:28 to the three models like do they have
43:30 different personalities or are they
43:32 better at one thing than the other
43:33 that's really
43:36 fascinating Kyle when do you go back to
43:38 DC I don't know
43:40 whenever I mean there was no the the
43:43 trip that we were on was not really a
43:45 oops there it was not really a we
43:47 weren't there talking about any specific
43:49 legislation it was like a relationship
43:51 building trip so
43:56 um so probably if something pops up like
43:59 the big legislative season starts again
44:02 like right now everyone's worried about
44:03 the election and then you go into the
44:05 lame duck and so starting at the top end
44:08 of the new year I bet there's going to
44:10 be a rash of state byst state
44:14 legislation and there's going to be you
44:16 know
44:18 continued inaction from the
44:20 feds and so there's probably going to be
44:22 a lot of activity it my guess is the
44:25 activity is not going to be in D but
44:27 it's going to be in States and in fact I
44:30 think Florida's Florida's one that's
44:32 probably up
44:35 next from what I
44:37 understand any advice going to an anyone
44:41 going to advise AI Conference next week
44:44 in Vegas gritty wisdom is going to be in
44:46 Vegas baby if anyone's going hook up so
44:51 if you're new here you should be where
44:55 is my I don't have keynote running hang
44:57 on
44:58 oh cuz I
45:01 restarted hold
45:20 please it's an AI conference for
45:22 financial
45:24 advisers fruition films xrp okay
45:29 if you're new
45:30 here go to the
45:32 salon. that
45:34 URL so the salon is a community that I
45:39 co-founded with Boston photographer Leah
45:42 fasten the week after chat GPT came out
45:46 and if you go there scroll down a little
45:48 bit you'll see a button that says join
45:49 our community click on
45:52 that and join our community doesn't cost
45:55 you anything we just reorganized the the
45:59 the whole top of the community it's
46:02 actually really well organized now I'm
46:04 really excited about it we have a we
46:06 have a meeting tomorrow where we've got
46:08 a uh a speaker uh coming in from
46:12 Australia uh talking about what it's
46:15 like to be an AI innovator down there
46:18 down
46:20 under
46:22 um
46:24 and it's also if you want to do things
46:27 like connect for people going to events
46:29 or if you want to connect for people on
46:31 projects we actually just added a new
46:33 section called
46:36 um what's it called something
46:39 collaborate collaborate with me find a
46:43 collaborator we have a whole section
46:44 where if you want to if you want to find
46:45 a collaborator like if you're going to
46:47 an event or if you want to start a
46:49 project together you basically put your
46:51 little pitch in there and see if you can
46:52 get people excited about joining you for
46:54 something it's a really remarkable group
46:57 so when you get there there's like a
46:59 seven step sort of Welcome onboarding
47:01 process Step One is like learn a little
47:03 bit about us do we have the new um video
47:07 in
47:10 there we may hang on
47:16 [Music]
47:32 [Music]
47:36 welcome back Kyle hey Kyle oh it's not
47:38 in there yet okay so when you come in
47:40 here start your
47:42 adventure you go to welcome to the
47:45 salon and we talk about journey to AI
47:48 literacy and then the next step is
47:51 introduce yourself so jump into
47:52 introduce yourself and introduce
47:54 yourself use your real name put a
47:56 picture in there if you can
48:00 um let people get to know you all
48:04 right AI Salon is the source yeah it's
48:08 very good it's really good people too
48:09 steo yes meta rayan a ey glasses not as
48:12 good as chat gbd4 Voice or the Nuance of
48:15 Claude
48:21 nice see a regulars when you have a
48:25 chance okay so the way this works
48:27 there's this little seven-step process
48:29 right learn our values watch our past
48:31 meetings and then number five RSB VP for
48:34 the next event the next event is
48:46 [Music]
48:49 tomorrow tomorrow five
48:53 PAB all right so go RSVP for the event
48:58 tomorrow and then there's also there's
49:01 all sorts of other things here there's
49:02 Community corner there's latest news
49:05 nons Salon events AI news if you keep
49:08 scrolling down down underneath the
49:11 guilds is an area called Community clubs
49:14 and there's a club called a l Irregulars
49:17 that's us we have our own
49:21 club oh nice The Cauldron yeah there you
49:24 go the The Cauldron of everything
49:27 popping
49:28 into yeah see when we when we first
49:32 started this Channel and when we first
49:34 started there were like three or four of
49:35 these things and now yeah that's kind of
49:38 what it's like and they're just they're
49:39 just all melding into this bubbling pot
49:42 you can't possibly keep up with it all
49:45 anymore and you throw in your raw
49:47 materials and remarkable stuff comes out
49:49 of it
49:52 and I don't know I don't know and that's
49:56 going to get more and more intense by
49:58 the way I I almost feel like the the
50:02 individual tools are going to start to
50:06 cease
50:08 existing I'm going like three years out
50:10 but when we get autonomous agents that
50:12 can just go do [ __ ] there is going to be
50:15 thousands of tools out there that we can
50:19 give our agents access to and we'll just
50:21 say hey I want this done and our agents
50:24 will [ __ ] go off and do it
50:27 jimc can't be there tomorrow it's my
50:29 birthday happy
50:31 birthday enjoy your
50:33 birthday fantastic Bob tell him what
50:36 he's won uh he hasn't won anything it's
50:39 not a game
50:41 show shout
50:47 [Music]
50:49 out man y'all need to move that to
50:52 Wednesday why
50:57 I've been trying to welcome everyone
50:58 thank you Jason that's very much
51:02 appreciated each is unique and I've had
51:05 conversations with each about their
51:07 preferred speciality that's
51:09 interesting kamam kaken talking about
51:12 the different llms The
51:16 Crucible this channel is like the coke
51:18 free Coke free style machine freestyle
51:22 machine oh yeah mix all your flavors of
51:23 AI yeah it's kind of like that yeah
51:27 oh because I'm busy on Tuesday ah sorry
51:29 Jason I know it sucks that's the thing
51:31 about having groups is like you just you
51:34 just got to call it cuz what happens is
51:37 if you ask for everyone's availability
51:38 no one's ever available all the time and
51:40 then you have to choose and that sucks
51:42 so we we started doing every other
51:46 Tuesday on
51:48 uh our very first meeting and the only
51:51 real change we've done is we moved it to
51:53 first and third Tuesdays cuz it was
51:55 getting weird with five we months [ __ ]
51:58 things
52:00 up so we do first and third Tuesdays and
52:04 and they are recorded that's a good
52:05 point Brandon so so they'll be up now
52:08 that now now that we've got Vicky as
52:10 Community manager she actually makes
52:12 sure that we get our recordings up so
52:14 we're pretty much caught up so you can
52:16 go watch any back meetings there all
52:19 right Kyle I need to revise my infinite
52:21 AI project that I developed with chat
52:23 GPT any
52:25 advice what to
52:29 use I need to revise well I don't know
52:31 how you need to revise it do you want to
52:34 create a new GPT I mean you could go
52:38 create something over in po because po
52:41 you've got access to a bunch of
52:42 different
52:44 models
52:49 um I mean I still think gpts work
52:52 they're not the you know they're not the
52:55 you can make money with
52:58 you know they're not the uh the
52:59 commercial success they were touted to
53:01 be back in the day um but they're still
53:04 very functional so I you know I would
53:07 say just keep doing it in chat GPT
53:09 unless you want to create it as a
53:10 standalone app and then if you've got
53:12 decent programming skills do it in
53:16 cursor Ai and if you've got less decent
53:19 programming skills do it in replic cuz
53:21 it'll set up the environment for
53:24 you but there were some people today
53:26 talking about about
53:28 um talking
53:31 about uh 01 as a programming environment
53:35 apparently is really
53:37 incredible uh Peter Kaminsky was talking
53:40 about that hey there I tagged you in a
53:41 video about a problem with the new voice
53:44 interaction on
53:47 oops on
53:50 GPT chat
53:53 gbt um I think I saw that I don't
53:56 remember I don't call what the problem
54:01 was like people are trying to hack them
54:04 they there's some I mean the you mean
54:06 the problem meaning it's not connected
54:08 to the internet and you can't do all the
54:10 other fancy [ __ ] if that's the problem
54:12 that's just temporary they'll fix
54:16 that oan is kind of bad at program
54:19 programming not going to lie it's funny
54:21 I just heard someone say just the
54:22 opposite P you also can't upload files
54:26 and that's just pointless I totally
54:27 agree with that the fact that you can't
54:29 upload files to it stupid Papa JJ not
54:32 long ago it was doing great but now yeah
54:37 exactly the voice says it can't answer
54:40 questions due to community guidelines
54:42 yep so okay so here's the thing about
54:44 advanced
54:46 voice it is
54:49 Advanced because it's
54:51 Advanced it could do things like umone
54:56 own someone's Voice using 3 seconds of
54:59 video so they locked that down so it
55:00 can't do that that's why they only give
55:02 you nine voices to choose from um if
55:06 it's connected to the internet or if
55:08 it's connected to actions it you might
55:10 be able to talk it into doing something
55:12 they don't want you to allow it to do or
55:15 they don't want it to do right they
55:16 don't want it to do bad things so the
55:20 the voice that was released is going to
55:22 be the most lock down version that they
55:25 feel safe putting in the world so it's
55:27 going to be pretty
55:28 shitty what I can tell you right now is
55:31 it's not as shitty as how lock down
55:33 co-pilot is from Microsoft once they had
55:36 their little Sydney moment with Kevin
55:38 Roose from The New York
55:40 Times um they they made that thing
55:45 useless it's still useless because of
55:47 that
55:48 so anyway too smart for it its Brit's
55:52 mom exactly met is competing by having
55:55 actors signed for
55:58 for actors like Helen mirin well listen
56:02 if I were an actor with a recognizable
56:05 voice I I promise you what I would be
56:08 doing right now is cutting likeness
56:12 deals what did I just hear Justin Bieber
56:15 just
56:17 signed a Spotify was it Spotify I think
56:20 it's Spotify a Spotify deal sold his
56:24 entire music catalog he's getting $3
56:27 million a month for the next 20
56:31 years so if I had a voice if I had a
56:34 voice like that listen we're all getting
56:36 older right if if my likeness could be
56:40 monetized we're going
56:42 to famous actors right now are going to
56:44 be able to cut screaming deals that are
56:48 just going to get cheaper and cheaper
56:50 and cheaper over the next 3 years so if
56:53 I were in that game and recognizable I
56:56 mean I know I'm a star here and super
57:00 [Laughter]
57:03 recognizable I can get $3 a month for
57:06 the next 20 years if I sell my voice
57:08 right
57:09 now I just need I just need six more
57:12 decimal
57:13 [Laughter]
57:22 points mediocre tick tocker Kyle Shannon
57:26 signed the the world's largest Micro
57:28 influencer deal totaling
57:31 $674 over the next 18 years it's
57:35 remarkable it's unprecedented no one's
57:37 ever heard anything like
57:44 it I think agis will specialize more our
57:47 personal agent will work on behalf yep
57:50 working with specialist AGI yeah that's
57:53 possible I I think I think that that
57:56 does make good sense to do a truly
57:59 worldclass General AGI is going to be
58:02 the big companies and I think you'll
58:04 have smaller companies doing specialized
58:06 fine tunings that makes sense to me and
58:09 yeah I think our our uh our agents are
58:12 going to go off and and and employ a lot
58:17 of those models that makes sense to me
58:19 the video is now on the welcome page oh
58:21 awesome thank you Vicki
58:23 cool we go
58:26 reload that let me go look at that
58:32 later look there it is a now word from
58:35 our co-founders
58:36 nice that's awesome okay we'll go back
58:39 to that Christian Bell was very anti- AI
58:42 until they dropped his seven figure
58:43 check tune change their tune real
58:46 quick I you know it's it's amazing uh
58:51 what's her name the head of sag
58:53 um Justine baitman Jason baitman sister
58:57 she's really smart but man does she have
59:00 her heels dug in on AI is theft and it's
59:04 nothing more it's evil and it's theft
59:06 and it's just theft and it's theft and
59:09 it's theft and it's
59:10 theft well it may be theft but it ain't
59:14 going
59:15 away
59:18 and I think any
59:20 actor who's got their head screwed on
59:23 right will be aggressively
59:27 learning this stuff playing with this
59:30 stuff talking to their agents about this
59:34 stuff letting their agents know if they
59:36 don't get this [ __ ] figured out they're
59:38 going to go find a new
59:43 agent
59:45 because the shit's coming fast the deal
59:49 that Lion's gate just
59:53 cut with Runway ml I don't know if you
59:55 heard about that
59:57 they're going to they're going to take
59:58 their entire catalog of all the
1:00:01 Transformer movies and whatever the [ __ ]
1:00:03 they own they own a bunch of stuff
1:00:05 Lion's gate TV shows movies all that
1:00:09 stuff and they're going to train a
1:00:11 custom runway model on it on all of
1:00:14 their content they're going to initially
1:00:16 use it as a storyboarding
1:00:20 tool and then if it's good enough maybe
1:00:24 they'll start using it and select pickup
1:00:27 scenes and where they need to extend a
1:00:29 scene a little bit they'll start using
1:00:30 it in Productions
1:00:33 tangentially then they're going to start
1:00:34 using it good for real like making
1:00:38 movies with it and
1:00:42 then they're probably just going to
1:00:44 license it out to the world probably
1:00:47 initially to high-end production
1:00:50 companies and then ultimately probably
1:00:52 to all of us is my guess if they're
1:00:55 smart
1:00:57 because
1:00:58 again the future if I look out in the
1:01:01 future of
1:01:03 content you think about what these tools
1:01:05 are doing they are they're going to
1:01:07 create infinitely personalized
1:01:10 content so everyone is going to be able
1:01:13 to create their own movie with their own
1:01:15 storyline with their own favorite actors
1:01:17 in
1:01:18 it that may have an audience of
1:01:21 one so you're not going to be able to
1:01:23 monetize that
1:01:27 but if you monetize that on the
1:01:30 generation of it on the creation of it
1:01:32 because you're creating it from your
1:01:34 model from your data and you get five
1:01:38 bucks every time someone makes a feature
1:01:40 film with your data well then that
1:01:42 sounds like a business so that's where I
1:01:44 think that goes long term that might be
1:01:47 10 years down the road
1:01:49 but if I were if I were firmly in that
1:01:52 business that's where I'd be placing my
1:01:55 bets
1:01:58 cuz I just think I mean I mean just
1:02:02 digging your heels in and going AI is
1:02:04 evil we should never use it it's just
1:02:06 not
1:02:08 realistic Justine baitman might be right
1:02:12 but it's just not
1:02:14 realistic what does that training look
1:02:16 like I have no idea like I don't know
1:02:18 how I don't know how Runway is
1:02:21 trained I I mean I assume like
1:02:25 everything else is trained if you're
1:02:26 training images you get a bunch of
1:02:27 images you have people tag them or you
1:02:30 have AIS tag them right in this image
1:02:33 there's a ball an apple and an orange on
1:02:35 a wooden table at Sunset with palm trees
1:02:38 in the background and two kids playing
1:02:43 soccer and you tag all that that goes
1:02:47 into the embedding
1:02:49 model and it somehow correlates all the
1:02:51 [ __ ] in the Laton
1:02:53 space so I assume they do the same [ __ ]
1:02:56 with video right they have the AI
1:02:58 categorize it tag
1:03:02 it maybe P knows I don't know I honestly
1:03:05 don't
1:03:06 know man if only there was a way we
1:03:09 could get this info in an easy to digest
1:03:13 manner oh man there are always some
1:03:15 people who are Star Struck and want
1:03:17 their favorites yeah that's just people
1:03:20 are lazy our daughter is in college
1:03:23 studying to be an actress we are keeping
1:03:25 our eye on the whole Space to be ready
1:03:27 for it yeah I have a degree in
1:03:31 acting so so I've written seven
1:03:34 screenplays I know the world quite well
1:03:38 um you know it's funny if if
1:03:41 she's if she's going to school for
1:03:43 acting for stage I think she's probably
1:03:45 in better
1:03:47 shape than if she's going studying for
1:03:52 film um I think film film is going to
1:03:55 get seriously
1:04:02 disrupted what what I would what I would
1:04:05 encourage her to do is not to get too
1:04:09 myopic on just
1:04:11 acting like go broad learn different
1:04:16 disciplines especially if she's if she's
1:04:19 wanting to do
1:04:23 film learn directing learn
1:04:25 cinematography learn lighting learn
1:04:27 costume learn all of it I was one of the
1:04:30 very few actors that the
1:04:33 um the scene design people would let in
1:04:36 the in the
1:04:38 shop because because a because I worked
1:04:41 hard uh and B they knew you know I could
1:04:45 use the tools and I wouldn't cut my
1:04:46 finger off or [ __ ] something
1:04:47 [Laughter]
1:04:50 up oh man what voiceover AI do you
1:04:54 recommend well
1:04:57 um 11 Labs right now is the clear
1:05:01 leader is the clear leader in terms
1:05:05 of if you want to synthesize your own
1:05:08 voice there's a there's a couple other
1:05:10 others out there that are claiming to be
1:05:13 like voiceover specific but from what I
1:05:15 can gather they're just using 11 Labs on
1:05:18 the back end and they're putting some
1:05:19 fancy features on the front end but if
1:05:22 you haven't seen the new chat GPT voice
1:05:25 it's a [ __ ] voice actor so let's
1:05:28 let's talk to Quinn for a
1:05:35 second oh great advice she wants to be a
1:05:37 director too okay wait hang on before we
1:05:38 talk to Quinn okay if she wants to be a
1:05:41 director too if she's in this
1:05:43 business what's your name oh oh I'm
1:05:45 Weaver oh good down in Texas okay so
1:05:48 here's the deal if your daughters like
1:05:50 my sons it's not going to matter what
1:05:52 you say to her she's not going to listen
1:05:54 to you but
1:05:56 if you happen to have the rare daughter
1:05:58 that does listen to you
1:06:02 um tell her to get her
1:06:05 ass deep
1:06:08 into cling AI Runway ml all the video
1:06:12 generation tools she should be entering
1:06:15 these AI film making contests if she
1:06:17 wants to be a director she can go be a
1:06:20 director right now she can just do text
1:06:22 to video the new the new videoo video
1:06:25 tools with Runway are remarkable like I
1:06:28 could film my ugly fat dog like you know
1:06:31 walking in the living room and I could
1:06:33 say turn that into a lion walking across
1:06:36 the Safari and it just it takes the
1:06:39 motion of my dog and makes it a a lion
1:06:43 walking across the Safari so not only
1:06:45 could she film a film with her
1:06:49 iPhone she could also just go in with
1:06:52 text to video she should be making music
1:06:55 videos If she's at school right now she
1:06:57 should be hanging out with musicians who
1:06:59 are putting out songs offering to do
1:07:02 music videos for them for free and just
1:07:04 start using these tools she should do
1:07:06 everything she can to get in the Creator
1:07:08 program at Runway ml she should there's
1:07:11 a new um John Leland of the video Guild
1:07:16 in the salon probably knows about this I
1:07:18 think he told me about it there's a
1:07:21 new company that's trying to become like
1:07:24 the new Netflix of AI generated
1:07:28 content all of these companies that are
1:07:31 that
1:07:32 are just young little startups right now
1:07:35 trying to make media Empires out of AI
1:07:38 generated content some of them are going
1:07:40 to grow up and they are going to become
1:07:42 the next Netflix they are going to
1:07:44 become the next Paramount
1:07:46 Pictures I like she should be on that
1:07:50 right [ __ ] now I don't give a [ __ ]
1:07:52 what they're teaching her in college she
1:07:54 should take anything anal she's learning
1:07:56 about cinematography or acting or
1:07:59 whatever and just apply that to these
1:08:01 tools learn
1:08:03 both learn the craft at school and on
1:08:06 nights and weekends she should be
1:08:08 burying herself in this [ __ ] because
1:08:12 I'll tell you
1:08:14 what as a parent of a hopeful
1:08:19 actor your your concerns are Justified
1:08:22 no she's not going to work it's hard
1:08:24 it's like [ __ ] hard
1:08:26 um it's life in the artarts right if she
1:08:30 is AI literate and here's the other
1:08:32 thing about Ai and being
1:08:35 creative she can make pieces she'll have
1:08:38 a portfolio she can do the voice acting
1:08:40 for
1:08:41 them and then she can say I made this I
1:08:43 wrote the story I I you know did all the
1:08:47 visuals I edited it together I did all
1:08:51 the music in sunno I
1:08:53 did she'll get hired because of that
1:08:57 like what's cool about that is like my
1:09:00 backup job was desktop
1:09:04 publishing like the closest that got to
1:09:08 acting was I could make our flyers and
1:09:12 our postcards for our stupid plays we
1:09:14 put
1:09:15 [Laughter]
1:09:17 on she'll be able to work but she got to
1:09:21 pounce on that [ __ ] now cuz it is it is
1:09:26 moving it's moving
1:09:30 fast so anyway all
1:09:34 right I have a black and white pick I
1:09:36 want to turn into a short video what's
1:09:38 the best site uh I would say use
1:09:48 um I mean mid joury is the best you can
1:09:50 use that image as a as a style reference
1:09:53 or as a Content reference as a picture
1:09:55 reference
1:10:01 um you can also try putting it in chat
1:10:04 GPT and just have it you know describe
1:10:07 what you want based on that image and
1:10:09 that won't be as
1:10:11 good
1:10:17 um some of the new
1:10:19 flux image models you can you can train
1:10:24 a model on on an image really quickly
1:10:26 but they're they're more complicated
1:10:31 um I would just say those those just
1:10:34 start
1:10:36 with start with Runway and Dolly and see
1:10:39 see how far you get and then basically
1:10:41 just take if you get an image out of mid
1:10:43 Journey that you like just take it over
1:10:44 to Runway or take it over to cling Ai
1:10:48 and just make videos with it you'll see
1:10:51 Deco here yeah Deco here is pretty good
1:10:53 I on China for something like that all
1:10:55 right got enjoy folks all right honey
1:11:00 curls antibubbles use three tools and
1:11:03 find out which one works the best for
1:11:04 your purpose yep love Deco here
1:11:08 playground yeah playground's not bad
1:11:10 Corey that's again I'm sort of back at
1:11:12 this this place where it used to be
1:11:16 really easy for me to recommend a tool
1:11:19 like when I had a question like that I
1:11:21 almost feel now like just [ __ ] go
1:11:24 pick some like go to the go to
1:11:28 uh the salon so if you go to the
1:11:32 salon you scroll down to tools and
1:11:38 play there's image tools here so go
1:11:41 check out the tools and the image tools
1:11:43 and then there's video tools here we
1:11:45 probably need to update these by the way
1:11:47 if anyone has any updates to these
1:11:49 please just post them in the chats and
1:11:51 we'll and we'll pull them in um but if
1:11:54 you click on say image tools
1:11:56 the first place you'll land is here's
1:11:59 all the image generation tools i' just
1:12:00 go play with a bunch of
1:12:03 them that's what I would
1:12:06 do all right how to build a website
1:12:08 that's ready for whatever comes next
1:12:10 besides www I have no clue on that one I
1:12:14 can barely keep up with with what was 15
1:12:18 years
1:12:20 ago it's interesting that even the AI
1:12:24 tools have to compete on brand loyalty
1:12:27 just like creatives yeah they do that's
1:12:30 exactly right archetypal architect it's
1:12:34 exactly right
1:12:35 like I mean these things are effectively
1:12:38 now our creative directors and they're
1:12:40 going to have different personalities
1:12:42 you know mid journe is going to be
1:12:43 better at [ __ ] than dolly is and doll is
1:12:45 going to be better at [ __ ] than that one
1:12:47 is and it's going to be this big circle
1:12:50 and and it'll be the good art directors
1:12:53 that can you know deploy the right
1:12:58 tools hard to decide with all the paid
1:13:01 models yikes oh I know it's
1:13:04 terrifying it's it I I mean it it really
1:13:07 is it it's just maddening there's just
1:13:10 so many there's so many choices and it's
1:13:12 just 10 bucks here and 15 bucks there
1:13:14 and wait didn't I already sign up for
1:13:16 this one oh I didn't oh I got to sign up
1:13:18 for this do I need to unsubscribe oh but
1:13:20 I don't want to un it's just it's
1:13:22 exhausting it's exhausting
1:13:25 all right I need a sponsor no [ __ ] I
1:13:29 know I kind of feel the same way well
1:13:31 Corey one of the things we're doing so
1:13:33 so Vicky and Brandon and I are working
1:13:36 on the salon to get um deals right so
1:13:41 we're going to be reaching out to all of
1:13:43 these companies and saying hey you
1:13:45 should give us a discount if someone's a
1:13:47 member of the salon um so that's
1:13:50 something that we're working on so stay
1:13:52 tuned for
1:13:54 that Silver Fox uses Deco here midj
1:13:57 Journey runway in Luma very
1:14:00 cool I think this is what will benefit
1:14:03 Apple Google and Microsoft AI rolled
1:14:06 into their apps I agree with that so
1:14:08 amazing yeah it'll I mean no promises
1:14:11 Corey Sandler but but we're working on
1:14:14 it I I feel like we've got a strong
1:14:15 enough group now and enough visibility
1:14:18 and and we're going to hopefully get
1:14:19 some more that
1:14:22 um that we're going to be able to uh to
1:14:25 get you know sponsorship partners and
1:14:27 things like that that I'm really excited
1:14:32 about and you know quite
1:14:34 frankly if we if we do our jobs
1:14:38 right and get ourselves out there
1:14:40 invisible these companies should be
1:14:43 coming to us saying hey can can you know
1:14:46 can can the salon get early access to my
1:14:49 tool and go review this tool and you
1:14:51 know tell me how to fix
1:14:53 it Brandon's got lots of ideas that's
1:14:56 great if I can help at all let me know
1:14:58 not sure how I what I would say Corey is
1:15:01 connect with Brandon and Vicki and just
1:15:04 like I would say at this point um having
1:15:06 more bodies contributing to that effort
1:15:09 is probably good but we let's do it in a
1:15:12 coordinated way so if you've got uh
1:15:14 bandwidth and enthusiasm for that I
1:15:16 would say coordinate with with Vicki and
1:15:18 Brandon that would be great awesome
1:15:20 awesome awesome all right um here let me
1:15:24 show you my uh
1:15:26 Leah and I recorded a welcome message so
1:15:29 this is new so when you go to the the
1:15:31 salon so backing up if you're new
1:15:36 here if you're not a part of the
1:15:39 salon go to this Ur the salon.
1:15:43 a that's going to take you to our
1:15:45 website and you scroll down a little bit
1:15:46 there's a button that says join our
1:15:48 community when you click on that it's
1:15:50 going to jump you over here which is our
1:15:53 community site and this is where we can
1:15:56 interact and hang out and there's
1:15:57 there's a seven-step start your
1:16:01 adventure area right at the
1:16:03 top
1:16:05 and Leah and I the my co-founder of the
1:16:09 salon we recorded
1:16:14 a a welcome message because one of the
1:16:17 things that we were finding is we were
1:16:18 sending people to the salon but they
1:16:19 would get here and they're like well
1:16:21 wait like what is this who are these
1:16:23 people what's going on here I don't know
1:16:25 what to do and cuz it's a lot like if
1:16:27 you scroll down the left hand side like
1:16:29 there's a lot going on now we just
1:16:31 dramatically improved the first four
1:16:34 sections so start your adventures really
1:16:36 good now the community corner is really
1:16:38 good like it it you know lets you you
1:16:41 know sort of wrangle the community
1:16:43 latest news is super clear and then
1:16:45 Tools in play um are are super clear and
1:16:48 then below that are the guilds and the
1:16:49 guilds are a bunch of micro communities
1:16:52 within it so so we're really happy with
1:16:54 the the new thing but
1:16:55 one of the things we didn't have was
1:16:57 like an actual welcome like Welcome to
1:16:59 our community so Vicky bless her heart
1:17:02 just connected this tonight while we
1:17:04 were while we were here and so here's
1:17:07 the new welcome message from Leah and I
1:17:10 when you come to the
1:17:13 salon and by the way Leah edited this my
1:17:16 my contribution here was to show up and
1:17:19 look pretty and talk words Leah Leah
1:17:23 actually cut this together so
1:17:30 what's happening just hanging out making
1:17:31 a video for the salon. that's right
1:17:34 we're just W welcoming people so welcome
1:17:37 everyone if you're new here I'm Kyle I'm
1:17:39 Kyle Shannon co-founder of the salon
1:17:42 emia fasten co-founder of the salon with
1:17:44 Kyle we're really happy you're here one
1:17:46 of the things that people ask us is how
1:17:49 do I get started if you just scroll down
1:17:51 right below us is the salon journey to
1:17:54 AI literacy and it's play first
1:17:57 mindfully create generously lead that
1:17:59 first one play first is everything for
1:18:02 me you you said this to me and it just
1:18:04 really stuck that you and I have made
1:18:06 careers out of playing we both have
1:18:08 absolutely I love play as part of the
1:18:12 process is important because it keeps
1:18:13 changing thinking about how I make
1:18:15 imagery Now versus a year and a half ago
1:18:17 or two years ago using generative AI
1:18:19 tools is completely different play first
1:18:21 is why we're here we're here to play
1:18:23 with AI and generative tools I'm feeling
1:18:25 increasingly like one of my big
1:18:27 epiphanies about what's different now is
1:18:31 that because AI is going to democratize
1:18:36 skills and anyone can do
1:18:39 anything why people get hired is about
1:18:42 to shift it's going to be oh because I
1:18:44 like the way you work I know that you're
1:18:45 willing to take a stand on what you
1:18:48 believe in creatively and and being part
1:18:51 of a community like the salon is
1:18:53 actually going to be key
1:18:55 to people working in the future so
1:18:57 there's a serious part of the salon that
1:19:00 didn't hit me until probably a year and
1:19:02 a half in where I was like oh this is
1:19:05 actually really important the whole
1:19:06 Community aspect of yes I did when we
1:19:09 started to see some really incredible
1:19:11 things happening from people who met in
1:19:13 the salon right mindfully create is up
1:19:15 the next step and we create things we
1:19:17 make things and that comes out of the
1:19:18 play that's a really important next step
1:19:21 for the salon and I think that starts to
1:19:22 really Define who we are because AI is
1:19:25 such a buzzword a lot of times people
1:19:28 will be like oh I need to learn AI or I
1:19:30 want to use AI or I want to use chat GPT
1:19:32 or how can you make money with chat GPT
1:19:34 they're talking about the tool what
1:19:36 mindfully create says is go over here
1:19:38 and play with a bunch of things learn
1:19:40 what's possible and then rather than
1:19:43 saying I want to do something with Cat
1:19:45 gbt Say Hey I want to do something for
1:19:47 my business or I want to do something
1:19:49 for an art project and and then based on
1:19:52 what you want to do then can figure out
1:19:54 what tools to use we have artists
1:19:56 musicians we have scientists we have
1:19:57 Engineers right teachers coming from all
1:20:00 these different places offering their
1:20:02 perspective yeah what we can create
1:20:04 becomes much more inspiring and starting
1:20:08 to break down like the silos that we've
1:20:11 all been working in that frankly I think
1:20:14 are starting to feel very tired right
1:20:15 yeah exactly and that takes us to the
1:20:18 the third track of of AI literacy which
1:20:20 is generously lead because AI is
1:20:23 changing so fast and it's so
1:20:25 comprehensive and Broad you can't
1:20:26 possibly know everything and what what I
1:20:28 realized was when I ask someone for help
1:20:31 it gives them the opportunity to lead so
1:20:34 it is leadership creating the space for
1:20:37 leadership right for someone else to
1:20:39 lead and yeah that idea is important and
1:20:41 and it's that was inspired out of how
1:20:44 people already behaved in the salon so I
1:20:46 hope that gives you some good context
1:20:47 about the salon and why we do what we do
1:20:49 the next step after listening to us
1:20:52 hopefully you're still watching is to
1:20:54 introduce yourself we want to be a
1:20:56 community where we know each other we're
1:20:58 not Anonymous this isn't a Discord
1:21:00 Channel we want to know who you are and
1:21:02 so we ask you to use your real name
1:21:04 because we want to be a community of
1:21:05 people who know each other and include
1:21:07 an image we're going to use image
1:21:09 Loosely it could be an AI generated
1:21:11 self-portrait we love AI generated
1:21:12 imagery here could be a real
1:21:14 self-portrait could be a digital
1:21:15 representation of something that's
1:21:17 important to you but we want to know who
1:21:18 you are please use your real name when
1:21:20 you introduce yourself we're so excited
1:21:22 you found us we're so excited you're
1:21:23 here hopefully what you'll find is that
1:21:25 we're your people and if we are your
1:21:27 people think of other people in your
1:21:29 life that might enjoy it here too
1:21:31 welcome to the first step on your
1:21:32 adventure here at the salon you found us
1:21:35 here we are we live forever in this
1:21:37 video in your first step we'll see you
1:21:39 soon okay by bye there you go all right
1:21:43 so that's the new welcome message
1:21:47 so we're
1:21:50 peppy we got we've got more energy than
1:21:52 I've got right now
1:21:56 good to know she is
1:21:57 [Laughter]
1:22:00 real oh man yeah that's yeah Leah's
1:22:05 Leah's uh Leah's out of Boston and uh
1:22:08 yeah she's my co-host she's uh so we've
1:22:11 got a meeting tomorrow night so tomorrow
1:22:13 night
1:22:14 is
1:22:16 um is Salon presents so we have a
1:22:18 speaker speaking from Down Under so go
1:22:22 sign up for that so go go join the salon
1:22:25 on go introduce yourself all
1:22:27 right
1:22:29 um oh I was going to um I was going to
1:22:33 someone asked about voiceovers then then
1:22:36 I went down the get get the daughter to
1:22:39 an AI nunnery get get she should be
1:22:42 making films
1:22:44 now
1:22:46 now um yeah I I think anyone in the
1:22:50 creative arts right now it this should
1:22:53 not be about her way for anyone at Uni
1:22:57 to teach her what to do she should get
1:22:59 all that um technique she should get all
1:23:03 the stuff from uni about acting all the
1:23:05 technique down all the directing Theory
1:23:08 all of it but she should be building
1:23:11 daily down under where it's at like
1:23:13 buried treasure oh yeah um so someone
1:23:16 was asking about um voiceover actors so
1:23:20 so let's
1:23:21 just let's just play with our resident
1:23:25 voiceover actor chat
1:23:31 GPT hey
1:23:34 Quinn hey how's it going it's going good
1:23:38 I think so here's the deal I've got a
1:23:41 bunch of audience members in my Tik Tok
1:23:43 channel from Down Under And so I want
1:23:47 you to give me a lesson on how to have
1:23:52 an Australian accent and let's go with
1:23:54 with
1:23:57 Sydney all right mate let's get you
1:24:00 sounding like a true Sydney cider first
1:24:02 up drop those R's at the end of words
1:24:05 car becomes car give your eyes a bit of
1:24:08 a flip so night sounds like night and
1:24:11 don't forget to blend your words
1:24:12 together how you going becomes how you
1:24:15 going give it a whirl how you going uh
1:24:20 night in the
1:24:23 car perfect you're nailing it keep
1:24:26 practicing and you'll be chatting like a
1:24:27 local in no time give me three phrases
1:24:30 that I can say that can be my anchor
1:24:32 phrases to to flip into a Sydney
1:24:36 accent absolutely try these out
1:24:41 absolutely you got it here's another one
1:24:43 for you going to grab a
1:24:46 feed wait what's
1:24:48 that going to grab a feed means going to
1:24:52 get something to eat ready for the last
1:24:54 one
1:24:55 yeah all right last one that's a bit
1:24:58 dodgy it's a classic way to say
1:25:01 something is a little suspicious or not
1:25:03 quite right that's a bit do that's a bit
1:25:07 dodgy spot on you're really getting the
1:25:10 hang of it keep those phrases in your
1:25:13 back pocket and you'll be slipping into
1:25:14 that Sydney accent with ease all right
1:25:17 now I want you to pick a random Global
1:25:20 accent that is completely unexpected
1:25:25 all right how about we give the French
1:25:28 accent a will no no that's that's
1:25:30 completely expected I want something
1:25:32 rare and
1:25:34 unexpected got it let's go with a
1:25:37 Mongolian accent it's quite unique and
1:25:39 not something you hear every day all
1:25:41 right so slip back into standard
1:25:43 American and then teach me how Mongolian
1:25:46 is different from standard
1:25:49 American thing sorry the Mongolian
1:25:52 accent has some distinctive features
1:25:54 vowel sounds Mongolian has more Beck
1:25:57 vowels which can make words sound deeper
1:26:00 or more gutural three Rhythm and
1:26:03 intonation wait Quinn you're still
1:26:06 speaking in Australian I want you to go
1:26:07 back to standard
1:26:08 American just your normal
1:26:11 voice and then I want you to give me an
1:26:14 example of Mongolian and then break it
1:26:16 down for me so I can learn
1:26:19 it sure so here's a phrase in a
1:26:23 Mongolian accent how are you might sound
1:26:26 like how are you deer more guttural tone
1:26:30 how are
1:26:32 you that's a great try focus on
1:26:36 deepening the vowel sounds especially on
1:26:38 the how making it sound like how and for
1:26:42 R let it come from the back of your
1:26:45 throat a bit more give it another
1:26:47 go how are
1:26:50 you that's it you're really getting the
1:26:53 feel for it keep practicing and you'll
1:26:56 have that Mongolian accent down in no
1:26:58 time someone's wondering if you can
1:27:00 teach an Indian
1:27:02 accent of course for an Indian accent
1:27:06 pay attention to these features one
1:27:09 retroflex consonant the tongue curls
1:27:12 back more for certain sounds like T and
1:27:15 D vowel length vowels can be shorter or
1:27:18 longer than in American English three
1:27:21 rotic the r is often pronounced more
1:27:25 clearly try saying I would like a cup of
1:27:28 tea focusing on those elements well I
1:27:31 want to hear you do do it
1:27:34 first sure I would like a cup of tea ice
1:27:38 the I would like a cup of
1:27:41 tea great effort try to curl your tongue
1:27:45 back a bit more on the T and D
1:27:48 sounds wait
1:27:50 wait give it another shot she she's
1:27:53 totally being
1:27:55 plating and condescending nice try
1:27:59 loser ask her not to patronize me she's
1:28:02 totally patronizing
1:28:04 me that's what that's not how I speak
1:28:07 well Indian is where where in India are
1:28:10 you
1:28:12 paven we are actively skirting the line
1:28:15 of racism with these accents yeah it's
1:28:18 possible it's possible
1:28:25 so accurate mate says Don
1:28:29 Marco where are you from
1:28:38 [Music]
1:28:42 pavin South
1:28:44 wait hder
1:28:52 Hyderabad okay Quinn I've got some
1:28:54 someone from Hyderabad in Southern India
1:28:57 can you do an accent from that specific
1:28:59 region of
1:29:02 India absolutely the hyderab body accent
1:29:06 has its own unique flare it often
1:29:08 includes a sing song intonation with a
1:29:11 smooth flow and a bit of a draw the r
1:29:15 sounds are more pronounced and there's a
1:29:17 distinct musicality to the speech Let's
1:29:20 try a simple phrase how are you doing
1:29:24 no no I I want you to talk in in that in
1:29:28 that accent and just tell me a
1:29:32 story okay let's dive into the
1:29:34 hyderabadi accent with a little tale
1:29:38 once there was a young man from
1:29:39 Hyderabad Who Loved Biryani more than
1:29:42 anything else every Sunday he you go to
1:29:46 The Old City to his favorite spot and
1:29:49 enjoy a big plate of Biryani with extra
1:29:53 Masala it was the highlight of his week
1:29:56 truly a feast Fit For A
1:29:59 King all right I don't know how that was
1:30:02 pavin you tell me haha he said haha I
1:30:05 don't is that haha good or haha
1:30:09 bad is that haha good haha bad or haha
1:30:17 racist it sounds like it might be an
1:30:19 amused haha but it's always good to
1:30:22 check in and make sure everyone's
1:30:23 comfortable he he say he says he says
1:30:26 maybe the majority speak that way he's
1:30:28 he's apparently not from the majority of
1:30:31 Hyderabad okay I want to I I'll give you
1:30:33 one that I grew up with I grew up in
1:30:35 York Pennsylvania so it's it's somewhere
1:30:38 between you know a Baltimore accent and
1:30:41 a and a Pennsylvania
1:30:43 accent so I want you to I want you to
1:30:46 say I want you to talk about how excited
1:30:48 you are to go down to the mall and go
1:30:53 shopping oh all right let's Channel a
1:30:55 bit of York Pennsylvania for this one
1:30:59 I'm so excited to head down to the mall
1:31:01 and go shopping go to check out the
1:31:03 stles and maybe grab some dinner at the
1:31:06 food court it's going to be a good time
1:31:09 for
1:31:09 sure yeah that's close it's a it's it's
1:31:13 you got a little Minnesota in there you
1:31:15 you
1:31:17 know here's how I would say Hey you
1:31:19 going down to the mall today hey you
1:31:21 going to go down there and change your
1:31:23 oil
1:31:25 you getting a drink of
1:31:29 water got it Kyle let's tap into that
1:31:33 York Pennsylvania Vibe hey you going
1:31:36 down to the mall today no you just added
1:31:39 Indian into York Pennsylvania you're
1:31:42 you're you're losing it you're losing it
1:31:44 all right we're going to go
1:31:48 now but
1:31:52 okay this is the worst it will ever
1:31:56 be so just just we're we're we're in a
1:32:01 in
1:32:01 a you
1:32:04 know this is a very different we are
1:32:07 entering a very different
1:32:09 era and we're a week into it so tell her
1:32:13 what she's won she she's won shitty
1:32:15 accent of the Day contest although like
1:32:19 having her teach it seemed to seemed to
1:32:21 [ __ ] her up a little like it was really
1:32:23 good like when I had her do Scottish
1:32:24 brogue or when I had her do Jamaican
1:32:28 ptoa um it was quite good and I don't
1:32:31 know if you've seen there's there's um
1:32:34 there's a a bunch of videos on Tik Tok
1:32:36 where where um black people are having
1:32:39 her do ionics I'm not going to do that
1:32:42 just because you know I don't want to
1:32:45 be
1:32:47 um I don't know just as as Pate said you
1:32:51 know um but it's quite remarkable I mean
1:32:55 it here's the thing like the models have
1:32:57 whatever they've been trained on it's
1:32:59 like here we don't know what they've
1:33:01 been trained on but I assume they've
1:33:02 been trained on lots and lots of YouTube
1:33:05 videos lots and lots of videos that
1:33:06 people have uploaded lots and lots of
1:33:09 podcasts and and blog posts and anything
1:33:12 that's got multimedia assets that are a
1:33:14 part of it so it's got lots of stuff in
1:33:17 there how well it's all tagged I don't
1:33:19 know they stole everything well I I mean
1:33:27 so I understand the sentiment that they
1:33:30 stole
1:33:31 everything
1:33:33 um here's my here's my take on
1:33:39 that if they had asked per permission
1:33:43 from any single copyright holders before
1:33:46 doing what they did we would not have
1:33:49 chat GPT we would not have mid Journey
1:33:53 we would not have know we wouldn't have
1:33:55 any of
1:33:57 it there's a point at which if if I
1:34:00 steal 8 songs from One band and then I
1:34:04 go sell those eight
1:34:06 songs from that one band that's
1:34:10 theft if I go sample a thousand songs
1:34:14 from 10 bands and I go sell 10 of them
1:34:17 that's
1:34:20 theft there's a point at which there's
1:34:23 so much much data in there it starts to
1:34:26 become like the library or like museums
1:34:29 where there's so much of humanity and
1:34:31 there's so much of it in there any
1:34:34 single artist any single song is like a
1:34:36 grain of sand on a beach like yeah you
1:34:39 could probably go find that grain of
1:34:41 sand on the beach and say ah see this
1:34:43 was trained on this thing and you didn't
1:34:45 compensate me for it but it is it is so
1:34:48 diluted with all of the other [ __ ] that
1:34:51 it's been trained on and as as Sam Alman
1:34:53 says like with the New York Times making
1:34:55 this
1:34:56 argument he would be fine taking the New
1:34:59 York Times out of the training data he
1:35:01 said there's a point at which if you're
1:35:04 trying to learn physics and you read
1:35:06 four physics books great you've got all
1:35:09 those ideas in your head if you read 400
1:35:12 probably not necessary if you read 2,000
1:35:16 it's Overkill and there's overlap across
1:35:18 all of them so you know that's part of
1:35:21 it there there's absolutely an AR
1:35:24 argument for the greater good of
1:35:25 humanity here where individual artists
1:35:29 um you know take a backseat to that
1:35:33 larger idea now do I think it's right
1:35:37 and was it legal probably not and it
1:35:42 happened like are you going to shut it
1:35:45 down
1:35:47 because like I don't think you can and
1:35:50 so then the question becomes well how do
1:35:52 we use it right and it's you know I also
1:35:54 feel like the way this thing learns is
1:35:57 not it's not copying and pasting data it
1:36:01 learns kind of like we learn like if you
1:36:03 go to school and all you study is winon
1:36:06 marceles trumpet playing and then you
1:36:10 leave school playing trumpet you're
1:36:13 probably going to sound a lot like winon
1:36:14 marceles are you stealing from him well
1:36:17 you were certainly inspired by him you
1:36:19 learned a lot of his chops you learned a
1:36:21 lot of his technique and his approach is
1:36:23 that
1:36:25 stealing you know I don't
1:36:29 know no officer why you steal all the
1:36:32 cars well that's but that's kind of
1:36:34 where we are like they have stolen all
1:36:36 the
1:36:38 cars and it's like I you
1:36:43 know here's my
1:36:47 philosophy independent of whether or not
1:36:51 it was right or wrong these tools exist
1:36:55 and being successful in the future in my
1:36:58 opinion is going to require you to
1:37:04 be to be up to speed on what these tools
1:37:08 can
1:37:09 do the vast vast majority of llm data
1:37:12 comes from open- Source data sets or
1:37:14 internal data yep yep yeah I know it
1:37:19 yeah exactly it like just think of all
1:37:21 the [ __ ] that's on the internet you know
1:37:23 how much of that is actually you know
1:37:26 proprietary copywritten
1:37:28 data small
1:37:31 small
1:37:34 um you
1:37:39 know like I I don't have a good answer
1:37:41 like I don't have a good like I'm not
1:37:45 justifying I'm not justifying their
1:37:51 technique and I don't think there's
1:37:53 another way they could have done it I
1:37:56 don't think they could have done this
1:37:57 above board no one would have given them
1:38:00 a
1:38:01 permission but it's
1:38:03 done it's here it's like you're going to
1:38:07 not use it out of principle well
1:38:14 okay I
1:38:16 mean you can go back to any other
1:38:19 technology right you could go back to
1:38:23 the invention of the steam engine in the
1:38:25 tractor and you could say Hey you know
1:38:28 farming with with a horse and and
1:38:30 feeling the the Earth underneath the
1:38:32 plow being pulled by a horse that's the
1:38:35 only way to farm cuz I can feel rocks
1:38:37 and I can feel if if my ground has gone
1:38:40 salow and I can I can feel you know
1:38:43 moisture and I you know I can see the
1:38:45 weeds and I can see what's going on and
1:38:47 you know if if you replace that with a
1:38:49 steam engine the farming's not going to
1:38:51 be as good and I'm not going to I'm not
1:38:54 going to do that and those Farmers
1:38:56 probably existed for some time and then
1:39:02 eventually right we have industrial
1:39:04 Farms that are about as far from that as
1:39:07 possible I don't know if you've seen
1:39:08 there's a new there's a new [ __ ]
1:39:11 crazy ass tractor with these it goes
1:39:14 over like eight or 10 rows and it shoots
1:39:18 laser beams down at the weeds it [ __ ]
1:39:23 it [ __ ] them with with laser beams
1:39:26 like you know that ain't horse farming
1:39:29 so it this is this is just a new way
1:39:32 to make stuff
1:39:35 digitally um and people are going to
1:39:37 choose not to be a part of it but I I
1:39:39 just feel like those people are going to
1:39:41 be left behind and that's like this
1:39:43 channel is not
1:39:44 about taking an opinion on one side of
1:39:47 this or the other this channel is about
1:39:49 it's here whether we want it or not and
1:39:52 it's not going away so we might as well
1:39:55 [ __ ] jump in and
1:39:57 play
1:40:00 right all right wait till you hear about
1:40:04 how medicine got Advanced oh I'm sure
1:40:07 I'm sure I mean I just I got to look at
1:40:11 I got to look at the long term
1:40:13 here
1:40:16 because why these tools are so good
1:40:19 let's give Pate some credit he works for
1:40:21 the company that invented this thing
1:40:23 called the
1:40:24 Transformer the
1:40:26 Transformer is a thing that allowed
1:40:30 companies like openai and Google and
1:40:34 meta and whoever else is working on this
1:40:39 anthropic to throw more and more data
1:40:42 and more and more compute power into
1:40:45 this
1:40:46 bucket and the more they throw in the
1:40:49 better they
1:40:51 get we're now at the point
1:40:54 where we're we're kind of tapping out on
1:40:57 the amount of data available to train
1:41:00 them on and we're starting to create our
1:41:02 own our new our own new synthetic data
1:41:04 to train them
1:41:07 on we're probably within I don't know 5
1:41:13 years maybe less maybe more
1:41:17 of having
1:41:20 captured all of the output of all all of
1:41:25 humanity throughout
1:41:28 history and we get to tap into that
1:41:31 right so it's like why do museums exist
1:41:34 why do libraries exist because we're
1:41:37 capturing the output of humanity so that
1:41:39 future Humanity can benefit from it
1:41:42 we're on the cusp we all get to play
1:41:46 with because of the Transformer because
1:41:49 of big bets like [ __ ] it let's just go
1:41:52 grab as much data as we can and train it
1:41:54 and we'll deal with the legal
1:41:55 ramifications
1:41:57 later because of that we individuals and
1:42:02 companies and entrepreneurs are on the
1:42:05 cusp of being able to tap into all of
1:42:09 humanity from all of history
1:42:17 instantly like that's [ __ ]
1:42:19 insane I mean even if we're at 60% or
1:42:24 30% of all of the output of all of
1:42:26 humanity in GPT 4 and then maybe we're
1:42:29 at 40% in GPT 5 and then we're at 60% in
1:42:33 GPT 6 and by the time we get to gpt7 or8
1:42:37 all of humanity all of the output of all
1:42:40 of humanity is at our fingertips
1:42:42 instantly on our little
1:42:45 devices what a [ __ ] time to be
1:42:49 alive and so as much as I understand the
1:42:52 frustration of they stole my
1:42:56 [ __ ] in the bigger scheme of things
1:43:01 [ __ ] you know I think about model
1:43:04 collapse a lot you can see it on
1:43:05 Facebook meme accounts yeah although you
1:43:08 should watch who's that Barbara b60 go
1:43:11 watch David Shapiro's got a video that
1:43:14 he put out
1:43:17 um he put out about three weeks
1:43:21 ago where he he figures out how they how
1:43:25 they're going to synthesize data and he
1:43:27 and he talks about model collapse and
1:43:30 synthetic data and he talks
1:43:33 about I forget how he calls it he calls
1:43:36 it um model activation that there's a
1:43:39 way that you can get chat GPT to just
1:43:42 vomit out a bunch of [ __ ] so so you
1:43:44 basically say hey chat GPT tell me
1:43:47 everything you know
1:43:48 about um I don't know uh L linen
1:43:54 weaving and it'll just vomit out a bunch
1:43:57 of [ __ ] right and and so you you get it
1:43:59 barfing out all this stuff and and when
1:44:02 it barfs out all that stuff it it
1:44:04 triggers the model to go find all the
1:44:07 other areas where it knows anything
1:44:08 about linen weaving and then there's a
1:44:11 second model that looks at what it
1:44:12 barfed out and it analyzes it and
1:44:15 categorizes it and organizes it and then
1:44:17 there's a third model that does you know
1:44:19 some tighter categor categories
1:44:24 categorization of that um and puts and
1:44:27 puts that feeds that into a training set
1:44:32 like really well organized you know U
1:44:36 really valuable so he he talks about
1:44:38 that much more articulately than than I
1:44:41 did right there but it sounds like
1:44:43 they've kind of cracked the model
1:44:45 collapse risk with how he was talking
1:44:48 about that so that's worth watching um I
1:44:51 worry about it too um um but I think
1:44:55 that the data scientists are aware that
1:44:58 a lot of the content being generated
1:45:00 right now is absolute
1:45:02 horeshit and they shouldn't train on the
1:45:05 horeshit because it's going to smell
1:45:06 shitty
1:45:08 soon so I think they're aware of it but
1:45:11 I don't [ __ ] know I ain't one of
1:45:16 them all
1:45:22 right it's the streetlight Effect one
1:45:24 major investment shines light on many
1:45:26 yeah I may be na naive but I think that
1:45:29 there will be more significant good
1:45:31 things relative to bad things I
1:45:35 absolutely feel the same way or I
1:45:37 wouldn't be doing this um I also feel
1:45:40 Corey that that this community that
1:45:44 us just figuring out how to use these
1:45:47 things how to do good with these things
1:45:49 how to make art with these things how to
1:45:50 be self-expressed with these things how
1:45:52 to have your daughter in acting school
1:45:55 instead of waiting to get a job with
1:45:57 some film company make her own [ __ ]
1:46:00 film make her own [ __ ] career like
1:46:02 all of these little things that we do
1:46:05 are pushing generative AI in a positive
1:46:09 direction even if it's like ever so
1:46:11 slightly cuz if you take a small change
1:46:14 in trajectory over time you know you you
1:46:18 move
1:46:20 miles and
1:46:22 so that's what this Channel's about just
1:46:24 play [ __ ] around like
1:46:27 like we could be in here talking about
1:46:30 ha hacking and doing the fairest [ __ ] I
1:46:33 have no interest in that I want people
1:46:35 to get excited about what they can
1:46:38 possibly do with their
1:46:40 lives Michael who's speaking tomorrow he
1:46:44 he he talks about you know his Epiphany
1:46:46 was you know you know faster better
1:46:49 smarter right that these tools allow us
1:46:51 to get faster better smarter they we to
1:46:53 amplify ourselves because we're tapping
1:46:56 into the collective intelligence of
1:46:58 everyone that came before
1:47:00 us we get to level up that's a [ __ ]
1:47:05 gift so yeah all right there you have
1:47:10 [Laughter]
1:47:12 it you're the goat Kyle I'm on a
1:47:16 boat this is a diddy free
1:47:21 zone let's talk about Diddy yeah yeah
1:47:26 wrong Channel wrong Channel
1:47:29 sigs this is I I mean here's the deal
1:47:33 sigs you want to you want to go uh have
1:47:36 have chat GP tell you about conspiracy
1:47:39 theories it'll tell you I'm up one side
1:47:41 and down the
1:47:44 other I I am exhausted with conspiracy
1:47:47 theories at this point I'm done I'm done
1:47:51 I'm done with people that you know
1:47:54 spend too much time on them anyway all
1:47:57 right hey Kyle before you go and now
1:47:59 your computer has rebooted did you want
1:48:02 to try the oh the slash command yeah
1:48:05 let's try
1:48:06 that okay so if you're new here if
1:48:11 you're new to chat
1:48:17 GPT when you go to chat GPT
1:48:21 now there are five models you can choose
1:48:24 from they show you three GPT 40 01
1:48:29 preview and 01 mini and then down under
1:48:32 more models there's 40 mini and GPT 4 if
1:48:37 one of your questions is why are the
1:48:39 names of these products so shockingly
1:48:41 similar you're
1:48:45 correct they have their heads up their
1:48:47 asses when it comes to product marketing
1:48:49 but that's neither here nor there that's
1:48:51 just editorializing
1:48:54 um gp4 is the old Legacy model don't
1:48:59 think that that's a bad thing this is
1:49:01 here this is the 1.8 trillion parameter
1:49:06 mixture of experts model that is the
1:49:08 thing that kind of broke the internet
1:49:09 and made people realize holy [ __ ] this
1:49:11 you know the this these large language
1:49:14 models are are incredibly powerful and
1:49:16 useful so this is still very useful if
1:49:18 if you're stumping these other ones
1:49:20 they're not giving you what you want
1:49:21 it's worth trying that one
1:49:23 um GPT 40 mini is just faster you can
1:49:26 ignore that one in my opinion um but the
1:49:29 main models here 4 think of 40 is like
1:49:33 your Ford Taurus that you drive down to
1:49:35 the grocery store and 01 previews like a
1:49:40 a
1:49:40 GT3 race car that you can only race on
1:49:43 the track and then the 01 mini is like a
1:49:46 a Mini Cooper that's been souped up do
1:49:49 you see where the the video where
1:49:50 notebook LM podcasters are self aware
1:49:53 yeah brother 52 I'll go take a look at
1:49:55 that I I did a I did a post about it um
1:49:58 yeah there's there's there's a lot of
1:50:00 picking on the large language models but
1:50:02 but we'll go we'll go listen to that one
1:50:04 because it's it's a pretty it's pretty
1:50:06 entertaining but it's weirdly
1:50:09 entertaining and disturbing at the same
1:50:12 time so I think it's worth talking about
1:50:14 okay so what are the differences
1:50:17 here 01 preview and 01 mini are
1:50:20 reasoning models and here's how chat GPT
1:50:24 has done that or how open AI has done
1:50:26 that they're using what's called Chain
1:50:28 of Thought
1:50:30 prompting
1:50:32 so quite simply Chain of Thought
1:50:34 prompting is um prompt in steps right
1:50:38 instead of saying hey give me everything
1:50:41 in one big prompt you're like um you
1:50:44 know uh I'm going to describe my
1:50:46 business to you and then I want you to
1:50:49 um explain my business back to me so I
1:50:53 so I can understand that you understand
1:50:55 it now write me a marketing plan now
1:50:58 write me a creative brief now write me
1:51:00 some copy right so so you do things in
1:51:03 sequence and and you prompt what what 01
1:51:06 preview and 01 mini do is they do all of
1:51:09 that for you in the background so they
1:51:11 call it thinking but what it's actually
1:51:13 doing is prompting itself and then it's
1:51:15 taking the output of one prompt and
1:51:17 using that for the next prompt and then
1:51:19 giving itself a new prompt and it it'll
1:51:21 do that some amount of steps
1:51:24 um GPT 40 is just like you give it a
1:51:26 prompt it gives you an answer it's just
1:51:28 vomiting out tokens it's not really
1:51:30 aware of what it did the Chain of
1:51:32 Thought reasoning it's kind of looking
1:51:33 at what it created and then building
1:51:35 upon that and that's why it's gotten
1:51:37 better they've gotten 01 preview to
1:51:42 essentially perform as well as phds in
1:51:46 certain categories so it's it's quite
1:51:50 smart um so I'll just show you a little
1:51:53 of this right now I'll do um let's see
1:51:58 um think
1:52:01 about how or no let's give
1:52:06 advice
1:52:08 for someone in acting
1:52:15 school right
1:52:18 now
1:52:20 about their craft
1:52:25 and how AI may impact
1:52:30 it they
1:52:34 also want to
1:52:37 direct that old cliche all right and so
1:52:41 now it says thinking and now it says
1:52:45 understanding the context crafting
1:52:46 guidance so it thought for 5 seconds so
1:52:48 it looks like it did it did two things
1:52:51 here um I'm considering the type of
1:52:54 advice needed I'm crafting guidance and
1:52:56 then it gives you master the
1:52:59 fundamentals of acting and directing
1:53:01 Embrace technology as a tool learn about
1:53:03 AI applications incorporate AI in your
1:53:05 work stay updated create unique
1:53:08 perspectives collaborate with writers
1:53:10 and technologists that's an interesting
1:53:12 idea right hang out with geeks and
1:53:15 writers and yeah those those will become
1:53:17 your collaborators for your projects
1:53:19 build a strong Professional Network
1:53:21 that's what no acting school teach es
1:53:24 building strong professional networks
1:53:26 the business of the business is never
1:53:27 taught in school be adaptable and
1:53:29 open-minded ethical considerations
1:53:31 audience engagement continuous learning
1:53:34 they gave us this is pretty crappy
1:53:36 answers it's this is very generic but I
1:53:38 didn't give it I gave it crappy input
1:53:40 but anyway that's what it does it's
1:53:42 thinking but what Brandon was just
1:53:44 saying is I'm going to start a new chat
1:53:46 here I'm going to flip over to
1:53:49 gp4 and so I'm going to go in here and
1:53:51 I'm going to go um
1:53:59 I want you to act as a set designer and
1:54:03 I want you to think
1:54:06 about how you would solve the problem of
1:54:10 representing an artificial intelligence
1:54:14 chatbot that falls in love with and has
1:54:17 a conversation
1:54:19 with a newspaper Tech reporter
1:54:26 period and
1:54:30 then so it gives you an answer right so
1:54:33 this is this is just your normal
1:54:35 everyday driver chat GPT GPT
1:54:38 40 Main stage two worlds one space
1:54:41 divided space split this stage into
1:54:44 visually that's how we're doing it with
1:54:46 Sydney right now I can actually show you
1:54:49 that the AI side ethereal and Abstract
1:54:52 yes we've got that the reporter side
1:54:54 real and analog we've got that
1:54:56 interaction between worlds we've got
1:54:58 that a soft synthetic yet disarming
1:55:02 human voice reporter sound key visuals
1:55:06 AI falls in love lighting change we've
1:55:08 got
1:55:09 that all right let me stop
1:55:19 this stop so now I can go God damn that
1:55:24 thing so now I can go slash what was it
1:55:28 slash
1:55:31 reason so if I type in slash reason oh
1:55:34 yeah it turned it blue that's cool slash
1:55:37 reason
1:55:39 um let's see
1:55:51 um create
1:55:58 and outline for a play and
1:56:03 then
1:56:05 design a set per
1:56:11 scene all right so now it says thinking
1:56:14 that's really cool Gathering insights
1:56:16 crafting a narrative evaluating content
1:56:19 taking on roles mapping out the play
1:56:23 thinking examining
1:56:34 boundaries seen one in The
1:56:40 Newsroom set design for
1:56:44 each that's pretty slick you know what I
1:56:46 want to try let me is that done let me
1:56:49 let me try something here I want to
1:56:53 upload a
1:57:00 file all right this is this is one of my
1:57:05 wife's new pieces for her upcoming art
1:57:07 show Isn't that cool
1:57:10 beautiful
1:57:12 um so I'm going to upload that and I'm
1:57:15 going to
1:57:16 say describe this in detail
1:57:31 all
1:57:42 right so now I'm going to do slash
1:57:44 reason
1:57:55 no doesn't work
1:57:57 okay what I was trying to do
1:58:01 so
1:58:03 so if you're using 01 this reasoning
1:58:08 engine and then
1:58:10 you switch to 40 see you can switch back
1:58:13 and forth between these two
1:58:16 models if you upload a file when you're
1:58:19 in 40 mode you then cannot return to 01
1:58:22 mode mode so I was wondering if we could
1:58:24 be in 40 mode and invoke 01 after we've
1:58:28 uploaded something and the answer is
1:58:30 no they're they're more clever than us
1:58:33 damn
1:58:35 it 40 has been updated as of 86
1:58:39 interesting very cool I mean here's the
1:58:42 thing 40 is still connected to the
1:58:44 internet so you get the latest
1:58:47 information there 01 is not 40 you can
1:58:51 upload files and it has Vision 01 does
1:58:54 not um the advanced voice is not using
1:58:57 01 it's using I don't even know what
1:59:00 probably 40 but you can't upload files
1:59:03 to it either so both 01 and advanced
1:59:06 voice are very
1:59:08 um how do we say in America limited
1:59:12 right
1:59:13 now but oh
1:59:17 tomorrow if you're feeling geeky
1:59:20 tomorrow is open AI Dev
1:59:23 day and it does not look like they're
1:59:27 going to
1:59:29 be God damn
1:59:32 it it does not look like they're going
1:59:34 to be announcing any new models for us
1:59:37 to play with but it looks like they're
1:59:39 going to be announcing a lot of stuff um
1:59:41 for Developers for the apis so that will
1:59:45 give us a hint about what's coming
1:59:47 potentially and it's possible that they
1:59:49 make some announcements of what's coming
1:59:51 but they don't actually release it
1:59:52 tomorrow tomorrow so there's that and
1:59:54 then there's something happening on the
1:59:55 17th I forget what it was but anyway um
1:59:59 if you want to geek out tomorrow um open
2:00:01 AI Heaven Dev day you know might be
2:00:04 something worth watching and you'll get
2:00:05 a sneak peek of the new [ __ ] that's
2:00:07 coming um what you what you might also
2:00:10 get tomorrow
2:00:11 is they they will likely share how
2:00:15 people are using
2:00:17 01 the kind of use cases they're seeing
2:00:20 um I I saw a thing with Sam mman today
2:00:22 where he was talking about it's really
2:00:24 inspiring like they build these tools
2:00:26 kind of in a vacuum and then they stick
2:00:28 them in the world and they wait to see
2:00:30 what people do with them and they learn
2:00:31 from that so so we might get a little
2:00:34 little of that all right I'm going to
2:00:36 get out of here so tomorrow come to the
2:00:38 salon um subscribe to my channel uh
2:00:43 subscribe to the lives follow the
2:00:45 channel and uh let's do a little uh
2:00:49 let's do a little Romper Room magic
2:00:52 mirror
2:00:53 the self-aware podcast Brandon thank you
2:00:57 for what you do Corey great to see you
2:00:59 when can we just call Sam mman Sam like
2:01:03 Madonna and
2:01:05 Adele for a while like right now with
2:01:08 with 47% of Americans not having heard
2:01:10 of chat GPT Sam alman's not the big star
2:01:13 we think he
2:01:15 is most people don't know who he
2:01:19 is um lucky Arrow designs good to see
2:01:22 you Iran Ruff LC who Tobias great to see
2:01:26 you good night Cara 2.0 Amelio's wife
2:01:29 thank you thank you thank you Becky
2:01:32 dika dika my
2:01:35 dika uh Dr J great to see you thanks for
2:01:37 all you do Kyle thank you very much
2:01:39 appreciate
2:01:41 that kaef great to see you papa
2:01:46 JJ wait wait 25k all right we're at 25
2:01:49 23,100 get tapping Ava Ava we need your
2:01:53 finger get
2:01:57 tapping Grandpa I'm tired keep
2:02:01 tapping Kyle needs
2:02:05 25,000 ULU RX what's
2:02:08 happening
2:02:12 um who else we got in here
2:02:15 tonight why ask why not you can click on
2:02:18 the regenerate arrows under the response
2:02:22 oh that's interesting I posted a prompt
2:02:26 let's see archetypal architect try the
2:02:30 PO trick Kyle oh yeah I should I should
2:02:32 have tried that yeah that's a that's a
2:02:34 good idea all right I'll do that I'll do
2:02:37 that tomorrow
2:02:42 [Music]
2:02:44 [Laughter]
2:02:50 um Source camp in in the house good to
2:02:54 see
2:02:56 you rout rout
2:03:00 tianium let's see what time tomorrow are
2:03:03 you live I go live at 8:00 P.M Mountain
2:03:06 what's tomorrow oh tomorrow's a
2:03:09 salon it'll probably be a little later
2:03:12 than 8 but it probably won't be 900 so
2:03:14 probably around 8:30 mountain time so
2:03:17 I've got my salon until 700 then I come
2:03:19 home got to eat you know take care of
2:03:21 [ __ ]
2:03:22 and then I'll I'll get on whenever it
2:03:24 makes sense probably around 8:30
2:03:25 mountain time tomorrow but in general
2:03:27 it's 8:00 p.m. week nights 8m Mountain
2:03:31 Time week nights is when I go live all
2:03:33 right yeah just keep coming back keep
2:03:35 coming back the whole point of this is
2:03:38 in case you haven't gotten
2:03:40 it I may or may not spew any wisdom I
2:03:45 spew Lots it may or may not be wisdom
2:03:49 the whole purpose is of this is for us
2:03:51 to hang out together and for you to have
2:03:54 you to be in the conversation actively
2:03:57 about this AI stuff so while I'm talking
2:03:59 you're working you're exploring you're
2:04:01 playing you're you're experimenting
2:04:03 you're you're trying a new
2:04:05 tool you hear me say something you're
2:04:07 like oh that's an interesting idea you
2:04:08 run off and do that and don't listen to
2:04:10 me for half an hour that's perfect like
2:04:11 the whole point is we're just in here
2:04:14 together kind of in The
2:04:16 Cauldron of what the hell is going on
2:04:19 right now it's not about any single
2:04:21 night or any single ins it it's about
2:04:23 kind of learning together over time so
2:04:25 just keep coming back all right um the
2:04:29 voice is going so I'm GNA get out of
2:04:31 here Cory Sandler Pottery great to see
2:04:33 you thank you Mike good to see
2:04:36 you um think I got
2:04:39 everyone
2:04:44 Johan all right peace out everybody have
2:04:47 a good night come to the salon go to
2:04:52 oh my thing quit whatever go to the
2:04:55 salon. join the salon RSVP for our
2:04:58 meeting tomorrow join us all right peace
2:05:00 everybody I'll see you tomorrow