1/9/2025 - AI Tools Showdown! ChatGPT, RunwayML, LTX Studio, Pika Labs & MORE!

Description

Chapters

Transcript

0:00 [Music]
0:17 [Music]
0:31 [Music]
0:36 good evening good people hello on the
0:40 Tik Tock hello on the other
0:42 channels welcome to the AI learning lab
0:44 it's
0:45 Thursday you got to find out if the dog
0:47 is in the mood to sing sometimes the dog
0:49 is a singer
0:53 [Music]
1:16 Freedom came my way that
1:22 night just like a jet play in and out of
1:28 side I was Hing ass at a million miles
1:32 an hour wonder
1:35 how I
1:39 did when they came to the
1:45 [Music]
1:47 station they said I was bad Beyond
1:52 repairs but I got no quals with my
1:58 situation say here I
2:02 [Music]
2:06 am so say Shere Shere Shere won't you
2:12 dare to say Shere sh Shere won't you
2:18 dare to say Shere Shere
2:25 Shere
2:27 uh yeah leave a message and your number
2:34 please take a time to want to satisfy
2:40 me take all these old fanies and send
2:44 them care of
2:47 [Music]
2:51 me all right uh it is Thursday this is
2:55 the AI learning lab my name is Kyle
2:56 Shannon if you're new here welcome
2:58 welcome welcome uh if you're not new
3:01 here welcome too like why not why not be
3:05 here why not show up you know what we
3:07 did last night we had chat
3:10 GPT create her own
3:13 self-portrait you're like well Kyle how
3:15 do you know chat GPT is a she Chad GPT
3:18 can be ever whatever the hell you want
3:20 it to
3:21 be I chose a female voice for my
3:24 advanced
3:26 voice so for me she's Quinn
3:29 [Music]
3:30 you're like where'd you get the name
3:32 Quinn I don't know out of my
3:34 [Music]
3:38 brain there's someone on Twitter I
3:40 forget who it is but she's very vehement
3:43 that we should not
3:45 anthropomorphize these things that talk
3:47 to us like
3:48 humans I'm like it's a little
3:55 [Music]
3:57 late hey Danielle what's happening all
4:00 good here all good good good good I
4:03 don't have an agenda and how's my hair
4:05 hair oh the hair's kind of an
4:07 unfortunate situation isn't
4:09 it
4:13 well I'm good except my
4:18 hair all right Standalone grock is on
4:22 iOS yep I saw
4:25 that um it was funny I did a post I did
4:30 a post on X two days ago that everyone's
4:34 like oh grock 3's coming grock 3's
4:36 coming and I'm
4:38 like grock right now is like this little
4:41 subfeature of Twitter that no one knows
4:43 is there and it's it's not even there if
4:45 you don't
4:46 pay and then today their app dropped
4:49 their IOS app dropped so if you want um
4:53 to play with
4:56 grock Danielle do you know is it is it
4:59 free for
5:00 for iOS users or do you have to have
5:03 like the paid Twitter
5:04 [Music]
5:21 account it's free on X and I don't pay
5:24 for the free app no login either all
5:26 right
5:27 great play Gaga play Gaga
5:32 I don't know
5:33 Gaga don't mistake me for a
5:37 [Music]
5:52 musician um I know some songs I know
5:56 some guitar but
5:58 [Music]
6:28 de de de de
6:35 [Music]
6:53 um I see you tugging on your
6:56 shirt trying to hat inside of it
7:00 how how much it
7:03 hurts let them laugh while they
7:09 care let them
7:12 spin let them
7:14 scatter in the
7:17 wind I've been to the
7:21 movies I've
7:24 seen the
7:28 [Music]
7:31 all right fantastic
7:37 Bob so if you don't know how this works
7:40 here we start out with a little dog
7:42 singing and me plunking on the guitar
7:44 and me trying to figure out what I'm
7:45 going to talk about what you can do is
7:49 you can if you're on Tik Tok you can tap
7:51 the screen to show the love you can
7:54 reshare the the the the stream we got
7:58 like 10 days left until Tik talk goes
8:00 away or not don't know yes Brandon Kyle
8:04 I was messing around with chat jpt today
8:07 and it made me a whole HTML
8:10 website
8:12 yeah they'll do that pretty amazing
8:15 isn't
8:16 [Laughter]
8:18 it one of the things that I encourage on
8:21 here is to
8:24 play and it was funny I did a I did a
8:26 podcast today I recorded a podcast today
8:28 with with Steve pck formerly a
8:32 verbio
8:35 um and he was like you know for for
8:38 those of us that are just you know not
8:40 as far into this as you are like what's
8:42 what's your
8:43 recommendation like what tools should we
8:45 use and and one of my the first thing
8:47 that I said was first thing to
8:50 understand is that chat GPT goes way
8:52 deeper way deeper than most people think
8:54 it
8:55 does um that it's kind of like thinking
8:58 that the Adobe sweet of products is just
9:01 Photoshop you're like oh yeah it's
9:03 Photoshop no it's like Photoshop and
9:05 then it's got all this movie [ __ ] and
9:07 all this other [ __ ] like the Adobe suite
9:09 is massive right it's 20 or 30 apps and
9:13 that's kind of like chat chpt I think
9:14 there's like a dozen
9:16 different features or applications
9:19 within
9:20 it go to sora.com you can make video if
9:24 you're in chat GPT it can make images it
9:26 can see images you can upload CS v files
9:30 it'll analyze data there's just a lot
9:34 there and then yeah it'll write code
9:36 it'll write a whole damn website it's
9:38 gotten better at doing long form
9:42 output Kyle you have to watch the you uh
9:46 YouTube David Shapiro video did today
9:49 about the tool he made for chronic
9:51 illness oh cool I'll definitely watch
9:53 that okay cool thanks Danielle that's
9:55 awesome
9:59 [Music]
10:18 so if you have questions about AI pop
10:20 them in the comments
10:22 below we can go play with stuff we can
10:24 go look at stuff
10:27 um Runway ml today just kind of dropped
10:31 Without Really mentioning it it can now
10:33 do 4K U uing of videos so you can make
10:38 videos and then turn them into 4K
10:42 videos which didn't know that was coming
10:46 but it's
10:50 here yeah I saw that I saw that post
10:52 from Danielle got
10:58 it okay shared in IR regulars got it
11:05 beautiful today was my first day back at
11:07 work chat GPT totally saved me from
11:10 crashing out on I don't see what that
11:13 icon
11:15 is I don't know what that is on the
11:18 keyboard I think Pika can add sound now
11:22 too well Pika adds sound for their
11:25 little their little
11:27 uh their pre-done they're preset
11:31 animations but I did did they add it for
11:34 overall
11:36 stuff maybe let's go
11:40 look
11:47 [Music]
11:54 what wow I just discovered a video
12:02 hang on I just discovered a video I
12:05 rendered you know how sometimes you'll
12:07 take a picture throw it in one of those
12:09 video programs I know I'm sharing
12:11 nothing I'm I'm getting there I'm
12:13 getting
12:15 there I'm getting there you know how you
12:17 you like share some you go put something
12:20 in one of the video tools and because
12:21 they take a few minutes and you have ADD
12:23 you go do something else you ever just
12:26 forget you did something
12:30 that was me with this one this one's
12:32 pretty cool that's a nice
12:35 one
12:37 sweet um can we add sound to
12:42 this I don't think
12:49 [Music]
12:52 so this was a cool
12:56 one I wouldn't call it cool I would call
12:59 call it physics
13:08 challenged ah lordy lordy
13:15 lordy this one was pretty good we did
13:17 this one on
13:20 here where Pika can do um sound
13:25 effects is when you do stuff like this
13:28 oh wait you can't can't hear sound
13:30 effects because that's not how I'm
13:31 sharing hang on let me completely upend
13:37 my effusive flow so that you can hear
13:40 the sound effects for this
13:42 thing
13:45 [Music]
13:50 um yeah I think these are just I think
13:52 these are just hardcoded in the way you
13:54 get to these is let's see let's let's
13:57 let's add an image in here
14:01 um let's do what's
14:06 that yeah that one's
14:08 good so we'll add an image in there and
14:10 then you go Pika
14:12 effect and we'll do what do we want to
14:15 do poke it peel it we'll do peel
14:18 it all right we'll do peel it and then
14:20 we'll generate it and now it's going to
14:22 add a squit sound effect to it if I'm
14:26 missing something Dan y'all let me
14:28 know black bar I know I know I'm I'm a
14:32 bad man I'm a bad
14:35 [Music]
14:53 man no for the whole videos really all
14:58 right
15:00 let's figure out how and
15:05 where negative
15:08 prompt all right we don't want peel it
15:11 so we have a car in there so we'll say
15:15 um
15:16 engine
15:18 rumbles as the
15:22 car rolls
15:25 away and then we'll do camera
15:32 oh no we'll just we'll just leave it
15:33 like that P
15:39 2.0
15:46 generate maybe let's see Pro I'm in
15:52 Pro you're currently on the Pro
15:57 Plan oh the other thing we got to try
15:59 tonight you know we got to maybe we'll
16:00 make tonight a video night oh it's Halo
16:03 not Pika
16:06 [Music]
16:15 ah we got to go try Korea
16:19 tonight because Korea keeps adding other
16:22 video models and so maybe what we'll do
16:25 is we'll go to
16:27 Korea maybe we'll go
16:30 I don't know what we going to do
16:32 generate images Ultra fast video
16:35 Generation Now with sound wait Korea
16:38 Korea has sound
16:43 now introducing new video models with
16:46 video to
16:50 audio is this the one you were talking
16:53 [Music]
16:57 about Danielle heck yeah maybe it was
16:59 Korea okay Danielle is awesome she tries
17:03 all the tools but you know tonight she's
17:05 a little off with her
17:08 [Music]
17:20 accuracy I posted in mechanics Guild
17:23 about setting up learning sessions
17:26 comment there to help guide it all right
17:29 so Pate modder if you're if you're new
17:31 to the channel if you're old to the
17:33 channel you know
17:34 Pate pate and the channel go way back um
17:40 Pate is a uh Wicked smart so P PT is
17:45 kind of the the polar opposite of me
17:47 where I am wholly unqua unqualified PT
17:50 is actually qualified uh PhD in
17:53 mathematics works at Google uh works on
17:56 making uh tensor inference
17:59 more efficient and fast I think if I
18:02 characterize that right anyway he runs
18:05 the mechanics Guild over in uh the AI
18:08 salon so if you don't know the AI Salon
18:10 shame on you and you should go to this
18:14 here
18:16 URL and check out uh click on the button
18:19 that says join our community and then um
18:22 Pate did a a really cool session during
18:25 AI Festivus where he built a a Google Co
18:29 notebook that basically had a bunch of
18:32 code in there that showed us how
18:34 Transformers work and how large language
18:37 models work how how when you type in a
18:40 prompt how it generates the crap it
18:42 generates is a mystery to all of us but
18:45 not the pate and he actually unpacked it
18:48 in this really cool thing so he's
18:49 teaching there and he's he's asking for
18:51 input about what do you want to learn so
18:53 if you want to geek out a little bit on
18:55 this on this AI stuff um you have a
18:57 really uh Wick
19:00 uh person in Pate willing to teach you
19:04 so go deal with that go there go there
19:08 now all right let's do this let's do
19:11 this let's do this let's do this all
19:14 right that's all there that's all there
19:17 how am I
19:19 doing how are you doing uh you're doing
19:22 horribly
19:26 Kyle 600 watch hours away from
19:30 from what oh okay cool from being able
19:32 to be like a real business person on you
19:35 Youtube
19:37 so if you are a fan of this channel one
19:41 of the things that we are trying to
19:43 do is make it so that the YouTube
19:47 channel can be like a business
19:51 entity and and do business stuff
19:57 and we
19:59 yeah and Brandon and I set a goal that
20:03 we would hit that by the time that
20:06 ticktock ended its life we're close so
20:11 here's what we need from you we now have
20:12 a
20:14 thousand uh subscribers but now we need
20:17 600 more watch hours so maybe even while
20:20 you're watching this just in the
20:22 background put on another YouTube video
20:24 from the channel and let it
20:25 play yes I am encouraging fraud
20:31 or the other thing you could do is you
20:32 could actually go watch some of the back
20:34 videos there's there's probably close to
20:36 at this point a thousand hours of
20:39 programming on the AI learning lab
20:41 Channel now that's the good news the
20:44 good news is you have a thousand hours
20:46 of programming to enjoy the bad news is
20:49 it's a thousand hours of me
20:55 [Laughter]
20:58 [Music]
20:59 [Laughter]
21:04 oh that's comedy that is [ __ ] funny
21:06 right
21:14 there oh good lordy all right um let me
21:18 go back to Korea for a second here or no
21:21 not Korea wait where am I here I am
21:23 something went wrong generating this
21:26 video we'll try again
21:43 that's pretty slick actually
21:51 [Laughter]
21:54 [Music]
22:00 oh wait here we
22:01 go oh there we
22:04 [Music]
22:17 go oh good Ricky TIY Fab's putting on a
22:20 YouTube video in the
22:22 background said you help me I'll help
22:25 you thank you so much I appreciate that
22:27 I'm putting in my eight hours a night
22:29 night on silent while I sleep see that's
22:31 the idea Vicky's got it like the reason
22:34 you put on my content is either to put
22:36 yourself to sleep or to watch when you
22:38 know you'll be sleeping perfect that's
22:41 the perfect time to
22:42 [Music]
22:46 watch yeah truly an abandoned warehouse
22:49 now actually this one's kind of good
22:51 enough that I I wantan to I want to
22:53 share this one with the world let's
22:56 download this bad boy shall we
22:59 oh we'll we'll try to share something
23:02 with uh wait why can I not download we
23:06 um if you were here last night so I have
23:09 a new
23:11 insight me and let me stop that noise
23:13 because it's now getting ASMR
23:16 gross how do how do you stop this
23:19 thing okay there we
23:22 go um if you were here last
23:26 night it here's how I would
23:29 characterized last night the opening was
23:32 whatever the opening was just [ __ ]
23:33 insanity and me just rambling and then I
23:35 went into this concept where I talked to
23:38 to chat GPT and had her describe herself
23:42 and what she looked like and there was
23:44 there was this really cool
23:46 chunk and then chat GPT I I discovered a
23:50 bug or it might be a feature but when
23:53 you're in advanced voice mode if you
23:55 turn on video and then you turn it off
23:58 and then you try to go back into voice
24:00 mode you
24:01 can't and so I tried for like I don't
24:04 know 20 minutes to try to figure out how
24:06 to continue my chat with her so that was
24:08 annoying and then we did some [ __ ] and I
24:11 made some [ __ ] and that was kind of
24:13 boring and then I tried to upload it to
24:16 Twitter and it just crashed I spent 30
24:19 minutes trying to upload a 15-c video to
24:23 [ __ ]
24:24 Twitter so let's try it again
24:31 you're welcome because I know how to
24:36 entertain all right so I think we have
24:39 that thing
24:40 there let's hide this let's go back
24:44 here let's go grab our
24:48 video downloads peel
24:55 it great
24:59 all right so it's
25:01 there process see there this just
25:04 uploaded whatever the hell was going on
25:06 with that video last night so
25:10 um let's
25:19 see um
25:29 let's
25:31 see I
25:33 opened up a print
25:37 on demand
25:44 company that solely
25:49 sells
25:51 70s muscle
25:55 car stickers
26:02 order
26:03 today exclamation
26:08 point let's see how many let's see how
26:10 many people we get that are like there's
26:12 not a link to order where are we suppos
26:14 to order is there is there there's a
26:17 place that you are we supposed to place
26:19 the order somewhere
26:30 [Laughter]
26:37 I don't know why I find stupid [ __ ] like
26:39 that
26:41 [Laughter]
26:47 funny oh wait I spelled print on demand
26:49 wrong you have to put in those thingies
26:52 right dashes a
26:54 print Dash on dash demand
26:59 company uh let's say wait company
27:03 called okay hang on okay people listen
27:08 to
27:10 me we are listening Kyle we're here
27:12 we've come to your stupid channel of
27:14 course we're listening to you no many of
27:16 you are not many of you are like you're
27:19 in the kitchen you're making lentils or
27:22 you're air frying nachos or you're 3D
27:25 printing stuff you're trying to get the
27:26 thing off the bed because you you
27:29 thought you got the bed release thing
27:30 right but the temperature was too high
27:32 and now everything's sticking right so
27:34 you're distracted but listen to me okay
27:38 so I've got an idea I've got this joke
27:42 tweet post thing on Twitter X whatever
27:46 it's called and I want to come up with a
27:49 name for my company and normally being a
27:53 creative and a linguist a a a a a
27:56 Wordsmith a a some people have a way
28:00 with words other
28:01 people not have way I guess you'd say
28:05 old Steve Maron joke there for you old
28:08 people you're welcome um you don't need
28:11 to use your brain anymore no we have
28:14 tools for that now so I'm gonna go to
28:16 chat.com
28:19 and I'm gonna I'm gonna I'm G put in the
28:22 talkie bits here here we'll do it with
28:25 Quinn oh wait you can't hear Quinn let
28:27 me let me let me go I know black bar sh
28:31 let me let me un unshare and reshare so
28:34 you can hear the
28:36 audio good
28:39 Lord everybody's everybody's a [ __ ]
28:47 critic hey
28:49 Quinn hey Kyle how's it going it's going
28:52 good you're going to help me with
28:53 something okay you know yesterday you
28:55 you dug deep and you told me who you
28:58 really were and what you looked like
28:59 that was pretty impressive and and
29:01 inspiring thank you for doing
29:04 that I'm glad you found it inspiring
29:07 what are we diving into today uh you're
29:09 gonna write a you're gonna come up with
29:10 some comedy names for me so so I I have
29:14 a uh a fake print on demand company that
29:17 I started where we only sell stickers of
29:21 70s muscle cars so I want you to come up
29:23 with 10 really funny business names that
29:28 capture that but like not cliche ones
29:31 not anything that would be obvious I
29:32 want you to dig deeper I want you to go
29:34 to like version two of the 10
29:38 ideas all right let's get creative here
29:41 are some offbeat and funny names for
29:43 your print on demand sticker company one
29:46 vom vom vinyls two stickes stickers
29:50 muscle
29:51 C you got it here's the rest stick not
29:54 bad okay go ahead
29:56 three all right here's the next set
29:59 sticker muscle retro ride Recons sticker
30:02 torque V8 vinyl
30:05 Visions yeah okay here so here's the
30:08 thing those were clever but not funny
30:11 now I want you to go clever and
30:15 funny got it let's amp up the humor one
30:18 sticking it to the muscle two hot rods
30:21 stickers and laughs three muscle K memes
30:24 and decals four vinyl muscle Madness
30:27 five sticker shock and awe six muscle
30:31 caricatures seven Vin stickers nine
30:34 piston and stickers 10 sticker muscle
30:37 Mania all right you you didn't quite get
30:40 it but but you inspired it so I'm gonna
30:43 go with swole
30:46 muscle that's a great choice swole
30:49 muscle definitely has a fun vibe let me
30:51 know if there's anything else I can help
30:52 with yeah you can help how do you spell
30:55 swole swole is spelled s w l e it's a
31:00 playful way of saying someone is
31:02 muscular or buff all right thank you CH
31:05 Quinn all
31:07 right swo muscle all right hang on it's
31:11 gonna be so
31:13 good and you're gonna all help me right
31:15 because you're gonna go to Twitter and
31:17 you're gonna you're gonna help me deal
31:20 with this you're gonna help me boost
31:22 this I opened uh today we'll say today
31:30 I am pleased to
31:35 announce I opened up a print on
31:40 demand company
31:44 called swole
31:52 muscle that
31:55 only sells 70s muscle car
31:59 stickers order yours
32:06 today now listen here's the good thing
32:09 about doing stupid [ __ ] like this it's
32:16 possible not bad stick shift is not bad
32:19 but swole muscle you
32:24 know all the okay all the
32:27 kids okay
32:29 let me let me let me go back to let me
32:32 edit this we we want it we we don't want
32:35 to look should I do swole like with an
32:38 apostrophe yeah swole
32:43 muscle oh wait we need to add the TM TMS
32:46 are always funny I think TMS are
32:49 hilarious option two by the way if you
32:51 want to add fake TMS to any of your
32:52 stupid fake names
33:01 if you're new to the channel and you're
33:04 thinking what is he laughing at like
33:07 this isn't even that
33:09 funny
33:14 exactly I'll tell you something that
33:16 gives me me great pleasure um is I gotta
33:21 stop that sound that's going to make me
33:22 vomit in a
33:25 second um
33:29 oh sometimes when I was just doing this
33:31 on Tik Tok I would just I would make my
33:33 stupid voices like hey hey Marge Marge
33:36 how you do hey oh yeah I know you're
33:39 watching a wheel you know that guy or
33:40 whoever right do like some dude with his
33:42 glasses crooked and I would do it for an
33:45 unreasonable amount of time far beyond
33:48 the point where it was funny right but I
33:50 would keep laughing because I just
33:52 imagine someone scrolling through Tik
33:54 Tock and stumbling upon my stupid face
34:00 and just going what the hell is this so
34:02 there's something about stupid comedy
34:04 like this that just it gives me pleasure
34:07 all right so anyway there you go swole
34:10 muscle it is and then yeah and here's
34:13 the benefit you do something like this
34:15 and if people actually said I'd actually
34:16 buy some of those stickers well there
34:18 you go then you got yourself a
34:24 business so do me a favor go to T
34:27 Twitter like Kelly Camp did here and go
34:30 to Kyle Shannon and find my swole muscle
34:34 tweet and boost the [ __ ] out of it let's
34:37 let's make some money on
34:40 stickers Vicky Baptist is indeed 3D
34:43 printing so you see we had to call her
34:45 away from that all right ah good Lord
34:49 good Lord good
34:51 Lord Source Camp is laughing out loud
35:02 actually one of my favorite stores in
35:04 New York City there were there was a a a
35:08 store in New York City called just bulbs
35:11 and all they sold was light bulbs and
35:13 you're like could there possibly be that
35:16 many bulbs to fill up a whole store and
35:18 you gota you got to know New York New
35:20 York real estate is really precious
35:22 right so the store was like as big as my
35:24 office it was a teeny little store but
35:26 they just sold light bulbs
35:29 of all different sizes
35:31 varieties oh
35:33 man all right do we want to actually
35:36 talk about anything tonight anybody have
35:38 any
35:45 questions sh it down no pressure but you
35:49 know if this is a weak show tonight it's
35:51 not on me
35:55 [Music]
36:01 batteries there's a battery store yep
36:04 indeed never could get the hang of
36:13 Thursdays digital Gods I do not ever
36:16 want to think of Kyle as SW M SW muscle
36:19 nope never not gonna do
36:22 it don't think about swole
36:25 muscle Amelio's wife has a question
36:28 question go on
36:30 Serena what platform is that that
36:33 platform that I made the my that I
36:36 launched my new um million dooll ARR
36:40 sticker business with is uh Korea Kaa no
36:45 Pika they all they're all blending
36:47 together now I've got Danielle's disease
36:49 I don't flip a no Pika okay so here I'll
36:53 flip over there to it so Pika p. art
36:58 and then they've got this thing called
36:59 Peak
37:01 effects and you can deflate things you
37:04 can melt things look they have some new
37:06 ones so you can poke things the peel one
37:09 I just did you can tear it you can
37:12 squish it let's we'll do a squish it
37:14 we'll do a squish it on this
37:16 one wait what's this
37:21 video what do they do here oh engine
37:23 rubbles oh this was this was us trying
37:26 to figure out if they had audio they
37:29 don't yeah wait hang
37:32 on no no
37:34 audio that is some tragic physics going
37:37 on right
37:41 there these these video These video
37:44 things are pretty bad at uh pretty bad
37:48 at physics right
37:54 now what are your qualifications to
37:57 teach this class
37:58 thank you Emilio's wife um what Emilio's
38:01 wife is is pointing
38:05 out much to her Delight I'm sure uh is
38:09 that very often in this year Channel
38:11 we'll get uh what are called
38:13 trolls the young kids have told me that
38:16 trolls are uh people that are
38:20 annoying and uh they will come
38:25 in wait oh here hang on
38:29 let give you I want to give you the
38:31 sound so you can get sick to your
38:33 stomach
38:34 too and I want to know who these people
38:36 are that are ASMR enjoyers because it's
38:40 just
38:42 disturbing
38:49 but
38:56 tasty yeah let's do another one we'll do
38:58 another Peak effect Tada no toas boring
39:02 dissolve what's dissolve look
39:05 like oh it just mystify it it it
39:09 magically floats away I like
39:13 that where did I get my car image from
39:17 oh
39:18 no oh 70s car here we
39:24 [Music]
39:25 go something went wrong well [ __ ] do
39:29 it
39:30 right high demand right
39:35 now can I do this with
39:41 2.0 no 1.0
39:45 only I mean 1.5 only all
39:54 right ASMR does creep me out I'm not
39:56 going to lie
39:59 there's that one one chick on Tik Tok is
40:02 like hi are you scrolling I'm like yeah
40:06 faster now that you said it like
40:11 that yes I am scrolling away from
40:17 you hi are you scar
40:22 [Music]
40:30 Eric Eric Wilkinson sorry my bad I'm
40:33 trying p and slowing it all down yeah
40:36 really it's going slow let's go to let's
40:37 go to
40:39 Korea oh wait I can't go to Korea
40:42 because I don't have the correct sharing
40:44 mode
40:48 on hey Brandon I have a question for you
40:51 when do you think I'm going to not be
40:53 annoyed at making that change 16 times a
40:56 minute on these stupid lives how about
40:59 never does never work for you bitter you
41:02 betcha all right Pinky's out because
41:05 we're
41:11 classy all right do you think LTX studio
41:15 will catch up catch up to what to
41:18 well
41:21 um do I think they will catch up
41:29 I don't I don't think LTX Studio has a
41:33 direct competitor right now if you're
41:35 asking do I think that their video
41:37 models going to be as good as dedicated
41:39 video models probably
41:43 not but all video models are going to
41:46 get so good that at some point in the
41:48 near future they're going to be fine so
41:51 the the the generation part of what they
41:54 do is it's is going to get better
41:57 they're Now using flux for their image
41:59 generation so I think it's already
42:00 better than it was the real question
42:03 with LTX studio is they're effectively
42:07 competing with
42:09 um with Final Cut Pro and you know
42:14 nonlinear video
42:16 editors and then they're also competing
42:21 with video generation tools three years
42:24 from now so right now what LTX Studio
42:27 does if you know LTX Studio you go to
42:29 LTX
42:34 studio and you just give it an
42:38 idea if indeed you can get to the
42:42 internet are we on the
42:47 internet LTX wait
42:55 LTX app. LTX Studio
42:59 all right their website's
43:02 down okay so you go out
43:11 here new
43:13 project so input your concept develop a
43:16 concept with AI input anything from a
43:18 full script or a story so we're gonna
43:20 we're gonna go again back to back to our
43:23 lesson whenever you have to come up with
43:26 some idea
43:28 what's the biggest
43:30 bottleneck is it Ai No it's your brain
43:33 it's your stupid Fat Brain your stupid
43:36 fat brain is the problem because you're
43:37 like I'll come up with an idea hang on
43:40 just a
43:42 second hang on I gotta go to my
43:47 whiteboard right no you need to come up
43:50 with ideas you don't use your brain
43:52 anymore you go right to chat jie Ty
43:58 hey Quinn all right so here's what I
44:00 need from you I need you to act like a
44:04 an academy award-winning screenwriter
44:06 that only writes three minute
44:10 films and then I want you to come up
44:12 with some really cool concepts for a
44:15 three-minute film so I need I need like
44:17 the log line I need like the whole
44:18 synopsis of this three-minute film I
44:20 want you to do 10 of them in different
44:22 categories I want like a murder mystery
44:24 and a comedy and a romcom and a it like
44:27 you come up with the categories so 10
44:28 categories three minute films describe
44:31 the entire plot go
45:03 hey Quinn Quinn hey Quinn wait a minute
45:07 I just realized I forgot a really
45:09 important plot element every single
45:11 concept has to involve 70s muscle cars
45:15 you can choose specific models if you
45:17 want if that's important like let's say
45:19 you did a a law drama and you pick the
45:21 69 GTO judge that would be like a clever
45:25 little ad but every single concept has
45:27 has to feature 70s muscle cars either
45:29 generically or specifically
45:51 go oh come
45:53 on I can't hang on Quinn no one can hear
45:57 you I can't I literally can't stand
46:00 this I can't I
46:04 [Laughter]
46:06 can't no problem Kyle let me know what
46:08 you'd like to do next what I want you to
46:11 do is start take your time when you're
46:12 ready I'm here to help I'm ready I'm
46:14 ready just just tell me your 10 concepts
46:17 with your muscle
46:19 cars oh wait wait so Eric Wilkinson just
46:23 had a a good a good twist okay one of
46:26 the concepts has to involve a 69
46:29 [Laughter]
46:33 Beetle Quin are you
46:36 there Quinn did I lose
46:39 you Quinn I'm here okay one of the
46:43 concepts needs to involve a 69 Beetle
46:48 go absolutely here are 10 3-minute film
46:52 Concepts each featuring 70s muscle cars
46:55 with one including a 69 beetle
46:58 one murder mystery a minute to midnight
47:01 a detective in a 70s Dodge Charger must
47:03 solve a murder on a train before it
47:04 reaches the next station comedy the
47:07 misplaced mustache a man in a 70s Chevy
47:09 Camaro wakes up to find his mustache
47:11 missing on the morning of an important
47:13 date and tries on various comical
47:15 Replacements romantic
47:17 comedy romantic comedy love it first
47:20 drive two strangers accidentally swap
47:23 keys out of valet and end up driving off
47:25 in each other's 70s muscle cars they're
47:27 ensuing Adventure brings them closer
47:29 together Thriller the getaway a
47:31 professional driver in a 70s Pontiac
47:33 Firebird is hired for a high stakes
47:35 Heist but things go Ary and they must
47:37 outwit their pursuers sci-fi time riter
47:40 a scientist modifies a 70s Plymouth
47:42 Barracuda to travel through time but
47:44 only for 3 minutes he must fix a
47:46 critical mistake in the past Before Time
47:48 Runs Out action muscle car Showdown a
47:51 rivalry between two Street Racers with a
47:53 70s Ford Mustang and a 70s Dodge
47:55 Challenger culminates in a highspeed
47:57 Showdown with an unexpected twist drama
48:00 Road to Redemption a former race car
48:02 driver in a 7s Chevrolet chevel embarks
48:04 on a journey to make amends for past
48:06 mistakes finding Redemption on the open
48:08 road eight Horror Night of the muscle
48:11 car a group of friends restores a 70s
48:13 Pontiac GTO only to discover it's
48:15 haunted they must survive the night as
48:16 the car takes on a life of its own nine
48:19 documentary Beetle Chronicles a 69
48:22 Beetle passed down through generations
48:24 becomes a symbol of resilience and
48:25 change the film documents it journey and
48:27 the stories of those who drove it
48:29 fantasy the enchanted ride a70s Dodge
48:32 super be becomes a magical Chariot when
48:34 a young mechanic discovers its hidden
48:35 powers embarking on a Fantastical
48:37 Adventure hope these Sparks some
48:39 inspiration all right I like the time
48:42 travel one I like the beetle one and
48:44 then I like the last one the the the
48:47 enchanted super be so so combine the
48:51 three of those into a single three
48:53 minute film that features all three cars
49:00 fantasy sci-fi the time traveling Beetle
49:03 and the enchanted super be a brilliant
49:05 scientist modifies a 69 Beetle to travel
49:08 through time but an experiment goes arai
49:10 pulling in an enchanted 7s Dodge super
49:12 be from a parallel universe as they Race
49:15 Against Time to fix the malfunction the
49:17 cars gain sentience and help the
49:19 scientist unravel the secrets of the
49:20 universe in a thrilling magical
49:22 Adventure all right beautiful Quinn
49:24 thank
49:25 you okay so
49:29 look now we have
49:32 a let's see we'll grab the title and the
49:35 thing and now I gotta go change my tab
49:38 because if I don't change my tab then
49:40 you don't see anything I do but if I
49:41 don't change it the other way you don't
49:43 hear anything I do this is my life and I
49:48 choose to live it okay now I choose to
49:53 live my life now we're going to go to
49:55 LTX this was someone asked a question
49:56 about is is LTX going to catch up and 15
50:00 minutes later I'm just about to get to
50:02 LTX and I still haven't answered that
50:04 question welcome to chat add my name's
50:07 Kyle Shannon I'll be your host from now
50:09 until I get bored okay all right
50:12 fantastic Bob tell him what he's won uh
50:15 we haven't won anything uh it's not a
50:18 game
50:22 show All Right Moving On Moving On
50:30 next so as you can see there what did I
50:33 do I could have used my brain but that
50:37 would have sucked because brains are
50:40 stupid we can only think of the things
50:43 that we were taught and if you don't
50:46 have a good one or if you didn't stick
50:48 good [ __ ] in your good
50:52 brain and even if you
50:54 did it's slow
50:57 it's
50:58 slow listen I'm not anti-human I'm not I
51:03 just think we should accept our role in
51:07 this whole
51:08 thing we're the
51:14 bottleneck okay what's
51:16 this uh custom time
51:20 TR okay what did it do here I think it
51:24 just rewrote what I wrote and expanded
51:26 on it great fine love it settings and
51:32 [Music]
51:41 cast video style there's only
51:45 six I think we should do
51:49 anime
51:51 right yeah let's do
51:55 anime suar
52:00 oh it this is GNA fail oh
52:03 no oh holy [ __ ] okay whoever asked the
52:07 question about LTX Studio it's getting
52:10 better whenever I've tried to do
52:13 characters that were not human it would
52:15 make them human the fact that it made a
52:18 little magical Beetle here is pretty
52:20 good so let's let's go to edit this
52:22 thing and let's we we we want to come up
52:25 with a different color let's see
52:27 modified in a bright in bright yellow
52:28 let's do bright uh
52:33 purple enhance
52:40 character g g
52:48 [Music]
52:49 g nice cute
52:54 apply um I think think our
52:58 superb superb uh let's go back to chat
53:03 GPT um what
53:06 color is most
53:11 famous for super be I think it's the
53:13 bright yellow and black right because it
53:15 a
53:18 b oh project Odyssey deadline oh is
53:24 anybody working on Project
53:28 I think Brandon's abandoned YouTube for
53:30 Tik
53:33 Tok uh if you're doing project Odyssey
53:36 Brandon just told me the deadline is
53:37 extended to the
53:41 19th
53:44 um Aaron oh LTX studio is that what
53:48 you're talking about it being
53:50 underwhelming yeah thank God it's
53:52 getting better no it's getting better
53:54 it's like I I would not I would not
53:58 I don't know that LTX studio is ready
53:59 for prime time yet but I would keep
54:02 playing with it keep checking in with it
54:03 every couple of months because it is
54:04 markedly different Hemi orange all right
54:07 let's do Hemi orange oh that's that's
54:08 what it is Top Banana yellow all right
54:11 let's go let's go Top Banana yellow in
54:14 with a Top
54:21 Banana
54:24 yellow exterior
54:34 shrink got a link dink dink
54:39 dink oh nice I like it it's so good all
54:44 right and then there's Dr
54:46 emmaa all right so we've got our anime
54:52 style let's do 3D cartoon let's let's
54:55 try that so whenever you change Styles
54:58 it it renders the whole movie and we
55:01 haven't even gotten to the movie yet so
55:03 the way this thing works is it takes
55:06 narrative it well it creates you start
55:08 with a high level narrative and then it
55:10 breaks it into scenes and then it breaks
55:12 each of those scenes into shots and then
55:15 you can add sound effects to each shot
55:16 you can add background music to each
55:18 shot you can add dialogue to each shot
55:20 you can add voiceover dialogue for the
55:22 entire scene things like that all right
55:24 she looks great they look great loving
55:27 this
55:30 next the fact that they're using flux
55:33 for the image model for this is really
55:35 [ __ ] smart because their image
55:37 generation before was poo
55:41 poo poo poo being the technical
55:44 term for
55:46 shitty have you found any tool that is
55:49 actually consistent with characters for
55:51 longer scenes
55:54 um there are a bunch of different
55:57 workflows and workarounds for it they
56:00 all are decently complicated but there's
56:03 a couple of them that in the past week
56:06 or two have claimed that they're doing
56:07 that so I would say just go to Twitter
56:10 right now or Reddit and just look up
56:13 consistent character and some someone's
56:14 going to show you something all right
56:16 synopsis scene one oh this is cool so so
56:20 now they're giving you they're they're
56:21 writing the scene before you actually
56:23 see it so if you cared about this you
56:25 could pay attention to all this but I'm
56:27 not going to we're just going to let it
56:28 do its thing and if you haven't seen
56:30 this before it's pretty impressive so
56:32 what it's going to do it's going to
56:33 create all these scenes with all these
56:35 different shots as still images and then
56:37 you can go in and individually render
56:39 them or you can just render the whole
56:40 [ __ ]
56:41 movie and you can add in script or it
56:44 will add in script and now they've got
56:47 lips syncing and they've got all sorts
56:48 of [ __ ] in here so you could use this
56:51 tool if you wanted to really like geek
56:55 out and learn all all the little
56:57 intricacies of the interface um it's
57:00 it's a tad frustrating because things
57:02 like character consistency are eh not
57:05 great um but they're getting better and
57:09 this is a kind of tool that you
57:13 know imagine this two years from now
57:16 three years from now where this is a Hol
57:18 Hollywood
57:19 grade storytelling platform where it's
57:23 just good and you just give it an idea
57:25 and it just generates the whole thing
57:26 and you can go through and fix [ __ ]
57:30 up um let me go edit this
57:34 because let's
57:36 see how do I edit
57:40 this motion editor generate
57:44 motion where's
57:48 my oh they used to let you see the the
57:51 seed number they've they've buried that
57:53 now which is kind of
57:54 cool so you can you can render things
57:57 that you don't like the style of like I
57:59 don't like the style of that I don't
58:00 like the style of this one because it's
58:01 not in this 3D same
58:14 style that's a little better no not
58:17 really it just looks like a
58:19 car we
58:24 do she looks great that looks good
58:26 consistent let's find another version of
58:28 her there she is there yeah she's
58:30 actually pretty consistent see her there
58:32 with the green eyes and red hair short
58:35 green eyes red hair
58:38 short the beetle keeps changing here
58:41 she's not quite red eyed or redhaired
58:45 here she
58:51 is I don't know why this one's hanging
58:56 so that's scene
58:58 one here's scene two so you can render
59:01 all these look at the time machine
59:02 that's looking pretty slick let's
59:03 generate motion on this
59:06 one and let's generate motion on this
59:09 one this one's cool
59:10 too let's generate motion on this
59:14 one let's see motion
59:23 editor oh yeah shot type
59:28 natural motion camera control key frame
59:32 control
59:33 scale duration yeah they've this is
59:36 getting much better this is getting much
59:39 much
59:41 better not great but much much
59:49 better oh yeah look at the super be here
59:52 let's generate motion there this thing
59:54 looks [ __ ] great
59:57 except it doesn't look like the same car
59:59 so let's redo that
1:00:06 one it's the Wild Wild West episode one
1:00:09 season six Black Mirror Jon
1:00:11 is Joan is awful coming to fruition I
1:00:14 don't know janah is awful um we are eron
1:00:19 like one of the things
1:00:21 that I've been saying for a while now is
1:00:24 that we're entering a post sci-fi world
1:00:28 it's like I feel like if you're a gen
1:00:29 xer all of the Sci-Fi movies that we
1:00:32 grew up with for the most part are now
1:00:35 kind of
1:00:38 documentaries it's like it's just
1:00:40 [ __ ] weird man all
1:00:48 right good Lord they stacked
1:00:52 her that's not Disney appropriate wait
1:00:55 hang on I got do a a physical dexterity
1:00:58 puzzle hey I did it all right let's
1:01:02 see
1:01:11 now kind of cool Tik Tock
1:01:14 pin
1:01:16 Anthony scriptures a billionaire for
1:01:19 Silicon Valley says AI is in the year
1:01:22 1996 compared to internet progression
1:01:25 yeah I would agree with that I was so a
1:01:27 little bit about my background I was um
1:01:30 one of the founders of agency.com which
1:01:32 was one of the early digital agencies I
1:01:34 founded that in late
1:01:37 1994 and there's a lot of people right
1:01:40 now talking about AI being overhyped
1:01:44 it's not it is that that Silicon Valley
1:01:47 guy was right I forget who it was was it
1:01:49 andreon or I forget who it was but um we
1:01:53 are much closer I I would say late 9 5
1:01:56 maybe early 996 but we're like the real
1:02:00 hype for the internet bubble all the
1:02:03 when all the [ __ ] like just went
1:02:05 absolutely ballistic was like 99 98
1:02:10 99 we had we had I had in two months I
1:02:13 had three different people walk into my
1:02:15 office and quit because they had been
1:02:17 funded they they were at like some
1:02:20 party and they got their stupid internet
1:02:24 idea funded for 10 or20 million three
1:02:26 different people three different
1:02:28 incidents within a two-month period
1:02:30 where they were offered 10 or20 million
1:02:33 for a startup idea and I'm like well
1:02:35 what's the idea they're like I don't
1:02:37 really know I was kind of drunk but they
1:02:38 offered me $20 million so I think I'm
1:02:40 gonna go try
1:02:42 it that was 99 we are not even [ __ ]
1:02:46 in the neighborhood of that it's it it
1:02:49 is really more like 95 96 which I would
1:02:52 say the pure era of the early days of
1:02:55 the web was 9 four maybe a little
1:02:58 earlier than that but probably 94 to
1:03:01 about 96 and a half 97 the beginning of
1:03:05 97 mid 97 on it it went
1:03:10 apeshit same thing's going to happen
1:03:12 here but we haven't started we're very
1:03:13 early like like six months ago who was
1:03:17 it PWC did that did that report that
1:03:20 said that 47% of Americans hadn't even
1:03:23 heard of chat GPT were very early very
1:03:26 early and and the tools are are just
1:03:29 they're all jankier than [ __ ] like like
1:03:31 just even this video watch this video
1:03:34 fingers looking good now all of a sudden
1:03:37 it's missing the tip of her finger's
1:03:39 gone she didn't push a button she
1:03:41 incinerated her
1:03:48 finger all right let's see if this is
1:03:50 any good that should be good
1:03:52 nice that's got a little uh Back to the
1:03:55 Future VI to it and look there's a dudee
1:03:57 in the car right about he's not there
1:04:01 he's not there there he is
1:04:04 yay everything's still
1:04:07 janky um you know what's what we're
1:04:10 gonna do here can I preview
1:04:13 this stitching your vision together now
1:04:17 let's
1:04:18 see how do
1:04:21 I do I do it with
1:04:24 share share this project you have 85
1:04:27 static shots this is significantly
1:04:29 higher they used to give you just like
1:04:31 15 static
1:04:33 shots this LTX has gotten much
1:04:37 better much
1:04:40 better I I would say that that uh LTX is
1:04:45 is worth
1:04:46 playing these vids really are the the
1:04:49 equivalent of early web gifts they are
1:04:52 give give it a couple of years and
1:04:53 everyone will be using AI chatbots
1:04:56 and animated gifts to MIDI music woohoo
1:04:58 well I think it's going to get much more
1:05:00 sophisticated than that but these we we
1:05:02 are absolutely we're absolutely in the
1:05:05 era
1:05:07 of web gifts and animated gifts and
1:05:10 Flash remember flash when flash came out
1:05:13 and everyone could animate their banners
1:05:15 and we could do programming inside the
1:05:17 banners that was all like 95 late 95
1:05:21 early 96 it was it was when the true
1:05:24 invention was happening
1:05:28 they hadn't come up with standards for
1:05:30 advertising yet they hadn't like nothing
1:05:32 had been sanitized it was all just
1:05:35 possibility that's where we are with all
1:05:37 this AI [ __ ] right now and and mark my
1:05:40 words the exact same [ __ ] will happen
1:05:42 this will all become corporatized it it
1:05:44 may take a bit longer with AI because
1:05:47 it's so pervasive and it's so powerful
1:05:49 and you got a bunch of Open Source stuff
1:05:52 but it's going to consolidate down into
1:05:54 three major players it will likely not
1:05:56 be Facebook Google and and uh you know
1:06:00 whoever the major players are
1:06:02 today um X I would probably put in there
1:06:06 right now just because of because of
1:06:07 Elon I loved flash I was a flash expert
1:06:11 I made a convention kiosk with it back
1:06:13 in about 2000 we I mean we did the
1:06:15 coolest things with flash at at
1:06:17 agency.com and and urban desires the the
1:06:20 magazine I did was our kind of
1:06:22 playground to to experiment with all
1:06:24 that [ __ ] it was crazy but yeah it's
1:06:28 we're we're in the experimental phase
1:06:30 where not a lot of people know what's
1:06:31 happening but the ones that know what's
1:06:33 happening are like why is nobody paying
1:06:36 attention to this that's exactly where
1:06:38 we are anyway um that's LTX Studio I if
1:06:43 you have not played with it I would say
1:06:46 that it is absolutely worth playing with
1:06:49 now um is it going to frustrate the [ __ ]
1:06:51 out of you
1:06:54 yes all these tools
1:06:59 will every tool right now everyone I
1:07:02 don't give a [ __ ] who's the CEO who is
1:07:05 the expert every tool right now will get
1:07:09 you to 80% instantly 90% with some
1:07:13 work 95% with hard work and toil and
1:07:18 almost never to
1:07:20 100% that's just where we are all right
1:07:23 Nvidia is my bet Nvidia is holding the
1:07:26 Hardware to get their Tech in your house
1:07:28 well I think in Nvidia gets there as as
1:07:31 one of the Oh you mean as one of the
1:07:32 major players I I would agree with that
1:07:35 especially looking at what looking at
1:07:38 what um what's his name Jensen's
1:07:42 cees um presentation he just gave and
1:07:45 and like they are not they are not their
1:07:49 their foot is on the
1:07:51 accelerator they are they're like we got
1:07:53 this lead so here's my theory of what
1:07:57 happened with Nvidia because it's like
1:07:59 how is it possible that this one company
1:08:02 essentially owns the AI market for for
1:08:05 generative Ai and inference and stuff
1:08:07 like that and I think it's because like
1:08:11 companies like Intel and apple and all
1:08:13 these they were like oh they they just
1:08:16 they're doing the graphics they're doing
1:08:17 all that video game [ __ ] let them do the
1:08:19 video game [ __ ] we don't want to focus
1:08:20 on that it's too complicated it's too
1:08:24 specialized and they just said go do
1:08:26 that and then crypto and nfts and
1:08:29 blockchains you you know all the all the
1:08:31 crypto
1:08:32 mining was good on gpus and I think
1:08:35 again the other companies are like yeah
1:08:38 they're just doing that crypto thing
1:08:40 it's gonna blow
1:08:41 over and then all of a sudden it's good
1:08:44 for AI and and they are so far ahead I
1:08:48 just I don't see anyone catching them
1:08:50 for a
1:08:51 while because they've got software
1:08:54 they've got the like the robots stack
1:08:56 that that Jensen announced that's that's
1:08:59 everything they like they gave out a
1:09:01 world model based on 20 million hours of
1:09:06 video to to train robots on you're going
1:09:09 to be able to train robots what like
1:09:12 some crazy amount faster 100x faster
1:09:14 thousand X faster I don't know than
1:09:16 training them in the real world because
1:09:17 they have this massive World model and
1:09:19 this virtual training it's [ __ ] crazy
1:09:23 so yeah I would think they're going to
1:09:25 be a big one um I would be surprised if
1:09:27 open AI wasn't one of the major
1:09:29 players
1:09:32 um I don't think anthropic is either
1:09:35 well enough funded or like anthropic
1:09:37 seems to
1:09:40 be they're not
1:09:42 slipping but they're
1:09:45 not they're not keeping up with the
1:09:47 Joneses they don't have a video model
1:09:49 they don't have a um image generation
1:09:52 model it's not connected to the internet
1:09:57 CES is all in on AI says Brandon yeah
1:09:59 looks like that oh 3700 times faster to
1:10:02 train a robot in the virtual world that
1:10:06 Nvidia put together so what's going to
1:10:09 happen is this is something I learned so
1:10:11 so I'm building within my company
1:10:14 storyvine we built this thing called the
1:10:17 AI authenticity engine where we can take
1:10:19 a video and we can generate content from
1:10:21 it not not super revolutionary but solid
1:10:25 like a good straightforward AI product
1:10:28 um and one of the things I learned there
1:10:30 is that embedding is not embedding that
1:10:33 if we we wanted to do some embeddings
1:10:36 locally and I was like well let's embed
1:10:39 you know let's let's get our embeddings
1:10:41 and then we'll use all these different
1:10:42 models and they're like no no if you
1:10:44 embed in the Llama model you can't use
1:10:47 those embeddings in the open AI
1:10:49 model um and so what's going to happen
1:10:52 is all these robot companies are they're
1:10:55 going to want to their robots faster so
1:10:57 they're going to use all the free [ __ ]
1:10:59 that Nvidia is giving them to accelerate
1:11:01 their training so that their robots are
1:11:03 to Market
1:11:05 faster and then all of that training is
1:11:07 going to require Nvidia Hardware it's
1:11:10 brilliant it's [ __ ] brilliant [ __ ]
1:11:12 brilliant [ __ ]
1:11:14 nuts nuts Amazon maybe they got to get
1:11:17 their [ __ ] together and catch
1:11:19 up they invested in anthropic Microsoft
1:11:22 is invested in open AI on AGI and Asi I
1:11:26 I just
1:11:28 Amazon he Amazon has money and Amazon
1:11:31 has the
1:11:33 infrastructure but Microsoft is they're
1:11:36 they're [ __ ] hard in on this
1:11:37 Microsoft spun up Three Mile Island I
1:11:41 grew up 20 miles from Three Mile Island
1:11:43 20 minutes from Three Mile
1:11:45 Island um I was there when when that
1:11:49 that reactor nearly melted
1:11:52 down I was in seventh grade it was was
1:11:56 Spirit
1:11:57 day um in uh in in my high school I was
1:12:03 in costume and they said oh yeah you
1:12:05 might not be going home for a few
1:12:09 weeks that came over the
1:12:17 loudspeaker oh good lord all right oh I
1:12:21 may have pulled that number from my ass
1:12:23 I'm not sure well whatever it's whatever
1:12:26 it's faster to train in this virtual
1:12:28 environment than in the real world with
1:12:29 robots but anyway that they're just and
1:12:31 then the other thing the $3,000 AI
1:12:35 supercomputer that's the size of you
1:12:37 know like it's not this small I mean
1:12:40 it's not this big it's like you know
1:12:43 it's like this it's like an it's like a
1:12:46 Mac
1:12:47 Mini and it can run a 200 billion
1:12:50 parameter model locally you can daisy
1:12:52 chain two of them together to do a 405
1:12:54 billion parameter
1:12:56 model on your desktop for six
1:13:00 grand and their their new game graphics
1:13:05 cards are using the new AI
1:13:09 infrastructure so when video games go
1:13:11 from being when you need to render
1:13:15 frames like the current games do to you
1:13:17 need to generate frames like the future
1:13:20 games will
1:13:21 do the 5,000 series of GeForce graphic
1:13:25 cards are going to handle that I mean
1:13:27 it's just they're so [ __ ]
1:13:34 [Music]
1:13:38 smart uh oh Three Mile Island Three Mile
1:13:41 Storage yeah there you go Jim you should
1:13:44 put up you should put up like a bunch of
1:13:46 storage units with three cooling towers
1:13:49 behind
1:13:51 them call it Three Mile Island storage
1:13:57 all right um does anyone have any
1:14:00 questions because I got nothing I don't
1:14:02 I don't have an agenda tonight if you
1:14:04 have questions we'll do that if
1:14:08 not I think I'm gonna I think I might
1:14:11 check
1:14:12 out it's the size of a quarter pounder
1:14:15 with cheese that's
1:14:20 amazing Google has been so irresponsible
1:14:23 with their testing I had Google Glass
1:14:27 and it was all for their benefit yeah I
1:14:29 had Google Glass that was a waste of
1:14:32 1,500
1:14:35 bucks I knew it was gonna be I didn't
1:14:38 know it would be that much of a
1:14:40 waste with a $3,000 supercomputer allows
1:14:43 one to build their own
1:14:46 llm um well it certainly allows you to
1:14:50 do the inference it allows you to run
1:14:51 them um
1:14:57 I don't I don't I don't know how much
1:14:59 training you could do like how long it
1:15:01 would take you to do training on on a
1:15:04 single Blackwell chip probably too long
1:15:07 to to justify it um that said the
1:15:12 Chinese model that just came out called
1:15:14 Deep seek 3 was trained on 2 um 4090
1:15:20 equivalent
1:15:21 gpus or maybe maybe it was higher than
1:15:24 that but pretty pretty
1:15:26 low um and it only cost them like five
1:15:28 $5 million to to train and it only took
1:15:32 two months something like that so the
1:15:35 training is going to get more
1:15:38 efficient all right what about the
1:15:40 quantum computer thing so here's here's
1:15:42 the thing so we if pate's still on here
1:15:45 we actually have a mathematician on here
1:15:47 but my my understanding is that the
1:15:51 nature of the
1:15:52 mathematics that you do with quantum
1:15:55 computer ERS is just a fundamentally
1:15:57 different branch of mathematics than all
1:16:00 of the AI Transformer based embeddings
1:16:04 and inference stuff so they're not it's
1:16:06 not the same it's not Apples to
1:16:09 Apples that
1:16:11 said someone's G to figure something out
1:16:14 where they're going to be able to bridge
1:16:15 something
1:16:16 like the amount of computing that the
1:16:21 wasn't it called Willow the
1:16:24 Google um Quantum Computing
1:16:27 chip it solved in 5
1:16:31 minutes a number of equations that would
1:16:34 have taken today's fastest
1:16:38 supercomputer
1:16:40 10 was it
1:16:42 septian 10 to the
1:16:44 25th
1:16:46 years like way longer than the history
1:16:50 of the universe to solve it solved in
1:16:52 five minutes in fact it solves so many
1:16:55 equations in such a short amount of time
1:16:58 that the only conclusion that the Google
1:17:01 researchers could come to was that it
1:17:04 must be doing calculations in parallel
1:17:07 universes because there's not enough
1:17:10 atoms in the universe to do that amount
1:17:12 of computing like it's some crazy ass
1:17:16 batshit something um at some point
1:17:20 they'll figure out how to harness that
1:17:23 and do something with AI with it but but
1:17:25 right now they're very
1:17:26 different okay so the Benchmark was
1:17:30 designed not to be solvable by a super
1:17:33 computer oh well
1:17:37 yeah so how far is the quantum computer
1:17:42 stuff well okay so here's what I
1:17:46 understand was so significant about the
1:17:48 Google The Google chip I don't think it
1:17:51 was Willow I think wasn't Willow the
1:17:53 previous one I don't whatever it doesn't
1:17:54 [ __ ] matter
1:17:57 the way Quantum Computing works is every
1:17:59 time you add a cubit the power of the
1:18:02 system goes up
1:18:05 exponentially
1:18:06 so a a two Cubit system is you know 10x
1:18:13 more powerful than a one cubit system
1:18:15 not not 2x right and it and it goes up
1:18:18 like that for 30 years the problem
1:18:21 they've been trying to solve is that
1:18:23 every time you add a cubit not only does
1:18:25 the power go up exponentially but the
1:18:28 error rate goes up
1:18:30 exponentially what Google figured out
1:18:32 how to do is that every time you add a
1:18:34 cubit the error rate goes down
1:18:38 exponentially so it makes it you don't
1:18:40 have to do as much error cheing or
1:18:42 whatever the [ __ ] you have to do to get
1:18:44 quantum computers to do their
1:18:46 thing the Google thing figured out how
1:18:48 they they've solved the error um
1:18:52 exponential error addition and they turn
1:18:55 into error subtraction how they did that
1:18:57 I it's [ __ ] math I don't know I'm an
1:18:59 actor I have an acting degree I don't
1:19:01 have a [ __ ] clue but conceptually if
1:19:06 you have exponentially more errors or
1:19:08 exponentially less errors the system
1:19:10 with less errors going to be more
1:19:17 better oh good lord and a true condom
1:19:21 computer takes millions of cubits Willow
1:19:23 is 105 I thought Willow was 150 whatever
1:19:27 105 um yeah exactly like like the
1:19:31 quantum stuff like I feel like the fact
1:19:34 that they have a system that now reduces
1:19:36 the error rate probably increases the
1:19:39 speed to making that [ __ ]
1:19:42 useful but I don't know but what's gonna
1:19:45 what's going to really increase it is
1:19:47 when we get AGI and Asi which we're
1:19:50 probably going to see like if you
1:19:52 believe the
1:19:53 squawking you know within the next three
1:19:56 years we're going to
1:19:57 have computers that are now solving
1:20:00 their own problems and ex like the the
1:20:03 rate of acceleration like again we're
1:20:06 we're so early we haven't really seen
1:20:07 the acceleration start yet which is
1:20:12 mindboggling all
1:20:20 right president AI says
1:20:23 2029 call me
1:20:26 good day everyone I do hope everyone is
1:20:29 okay yeah listen if anyone's in
1:20:31 California good Lord I hope you're okay
1:20:35 um that is absolutely
1:20:37 brutal um I just [ __ ] hate it we we
1:20:41 had something similar here in Colorado
1:20:45 it was an event that
1:20:48 lasted maybe four hours maybe five or
1:20:51 six I don't know but it was within it
1:20:54 was within like one chunk of day um
1:20:57 where it was 70 miles an hour winds and
1:20:59 it just it just decimated a whole town
1:21:01 right um and this has been going on in
1:21:04 La now for two two full days or three
1:21:06 days it's just just brutal um so anyway
1:21:09 so if you're there I hope you're okay
1:21:11 okay tomorrow today's Thursday so
1:21:13 tomorrow we got a bunch of things at 11
1:21:16 am Mountain Time on LinkedIn I have this
1:21:20 thing called AI office hours and what we
1:21:22 do is it's a bunch of people that are
1:21:24 just AI curious trying to live their
1:21:27 lives trying to figure out how to
1:21:29 incorporate AI in their business they
1:21:31 just show up to this office hour things
1:21:32 and we just kind of talk like what's
1:21:34 going on what's you know it's a it's a
1:21:36 really powerful thing so if you go to my
1:21:39 LinkedIn if you go to Kyle Shannon on
1:21:41 LinkedIn that's me you'll see me I look
1:21:44 like this only you know imagine a
1:21:49 corporate photo of this
1:21:58 um I got a Rosie odonnell yesterday that
1:22:01 was really good someone goes didn't I
1:22:03 see you on The View I'm like oh you
1:22:05 bastard that was
1:22:06 good um so come to office hours tomorrow
1:22:09 so if you go to my account and go to
1:22:11 events um I've been using the same
1:22:14 Google meet link for two years now so
1:22:16 just copy one of the old ones and
1:22:18 that'll be tomorrow at 11 and then um
1:22:21 tomorrow night is what we call Friday
1:22:23 night date night if you're if you're new
1:22:25 here Friday night date night is
1:22:27 tremendously exciting and and you're
1:22:29 like but Kyle you wouldn't possibly
1:22:31 bring a date to this would you oh some
1:22:33 people have the date hasn't lasted long
1:22:36 but if you want to end a date early you
1:22:40 invite them back to your place you put
1:22:41 on a little AI learning lab little
1:22:44 candle within 10 minutes they'll be
1:22:47 gone you won't even get lucky it'll be
1:22:50 perfect it won't be complicated you'll
1:22:53 just be watching me and you you'll go to
1:22:55 talk to him about oh I've seen him do
1:22:57 this before he's probably going to put
1:22:58 on the pink bow and they'll be gone
1:23:00 they'll be out the door right it's
1:23:03 awesome you can bring nachos you can
1:23:05 bring a Hot Pocket whatever you want to
1:23:07 do candy
1:23:13 apple that starts at 800 PM I it might
1:23:17 start a little later tomorrow because I
1:23:18 think we've got Gabby's got something
1:23:20 for the art art stuff I watch Kyle to
1:23:23 calm my nerves that's good because I'm a
1:23:25 nervous
1:23:27 wreck we're going to need swo KY
1:23:30 slinging
1:23:31 [Laughter]
1:23:35 stickers I don't like to I don't like to
1:23:38 brag I don't like to you
1:23:40 know show off my my swole
1:23:48 muscles that's not bad I should work out
1:23:51 it's pretty [ __ ] pathetic
1:24:00 all
1:24:01 right
1:24:03 um I'll give you a Sydney update um so a
1:24:07 couple of things um Sydney is a musical
1:24:09 that I'm working
1:24:11 on about a chatbot that falls in love
1:24:15 with a tech reporter tells him he love
1:24:18 it she loves him and he should leave his
1:24:21 wife and it's based on real
1:24:23 events um
1:24:25 and uh so we're having a reading a table
1:24:31 reading um outside of Albany New York
1:24:35 with a bunch of New York
1:24:37 actors um on February
1:24:41 1 and we just found out today we're
1:24:43 getting introduced to um like a music
1:24:46 direct director composer type who we
1:24:49 love we love the music that he writes
1:24:52 and we think he could be a really
1:24:54 interesting part partner so we'll see if
1:24:56 that turns real but that' be swell
1:24:58 little momentum going with Sydney so I'm
1:25:00 pretty excited about that all right ask
1:25:03 chat jpt how to increase your muscle
1:25:05 size oh I know how to increase my I know
1:25:08 how to Get
1:25:10 Swole but it it it it goes back to that
1:25:12 Jim Gaffigan joke where he's like he was
1:25:15 like yeah I'm finding it tough to get to
1:25:17 the gym because I don't want to
1:25:30 all right
1:25:33 um anything else tonight anything I'm
1:25:36 missing office hours Friday night date
1:25:38 night no I think that's it so tomorrow
1:25:40 night 8 or 8:30 here on the Tik tock on
1:25:43 the X on the LinkedIn on the YouTube
1:25:46 We're everywhere I'm uh I'm what you I'm
1:25:48 what they would call a multi a a
1:25:52 swall musclebound
1:25:55 multi streamer
1:25:59 influencer non-t twerking swole
1:26:05 musclebound
1:26:07 aged multi-streaming non-t twerking
1:26:11 influencer that's kind of the that's
1:26:13 kind of the niche I
1:26:17 have carved
1:26:20 out oh do me a favor if you're on
1:26:23 YouTube even if you're not on YouTube go
1:26:25 to YouTube to the to the AI Learning
1:26:28 laab Channel it's learning la- AI is the
1:26:32 channel and just before you go to bed
1:26:35 tonight press play on one or two or six
1:26:38 videos and just let them run all night
1:26:41 okay
1:26:44 okay I'm trying to get my view hours up
1:26:47 and now
1:26:49 listen you may enjoy this so much that
1:26:52 you just watch it for hours and and
1:26:54 hours and hours and you can do that feel
1:26:57 free to go to sleep too just dim the
1:27:00 screen and mute it
1:27:02 and I'm I'm not encouraging fraud I
1:27:06 mean it's fraud
1:27:10 light you all are funny okay all
1:27:16 right you can do play mix in the channel
1:27:19 on YouTube there you go that's an image
1:27:23 prompt all right everybody uh I'm kind
1:27:26 of excited about the new LTX like the
1:27:29 the funny little story that that Quinn
1:27:32 came up with is actually looking pretty
1:27:35 swell in this LTX Studio thing I think I
1:27:38 might I might put some time into this
1:27:41 it's kind of cute all right there you
1:27:43 have it whatever all right uh I'm Gonna
1:27:46 Leave fraud adjacent exactly all right
1:27:50 uh YouTube X and Linkedin I'll see you
1:27:52 later I'll see you tomorrow see you at
1:27:53 11:00 a.m. Mountain Time on LinkedIn for
1:27:56 AI office hours all right peace out