8/5/2025 - Exploring the Latest in AI: From Open Source Models to Genie 3

Description

Chapters

Transcript

0:12 10,000 words swim around my head. A
0:16 million more in books dripped in beneath
0:18 my bed.
0:24 I wrote I wrote all red
0:27 searching in the swamp. Still can't find
0:30 how to hold my hands.
0:36 >> And I know you need me in the next room
0:40 over. I'm stuck in here paralyzed.
0:48 For months I got myself in ruts. Too
0:51 much time spinning mirrors framed in
0:55 yellow walls.
1:01 Ain't it like most people? I'm no
1:03 different. I can talk on things they
1:06 don't know about.
1:11 Well, ain't it like most people, I'm no
1:14 different. Like to talk on things we
1:17 don't know about.
1:36 Good day, good people. I got a thread on
1:38 my shoulder. I think that means
1:40 something. Let's toss it for good luck.
1:47 Here at the AI learning lab, you can
1:48 learn about AI. You can also learn about
1:50 good luck charms.
2:04 It's not simple to say
2:08 the most.
2:10 I don't recognize me with these shoes
2:14 and this apron. That place and its
2:16 patrons
2:18 have taken more than I gave them.
2:23 It's not easy to know
2:26 I'm not anything like I used to be.
2:30 Although it's true. I was never
2:33 attention sweet center. I still remember
2:37 that girl.
2:39 She's imperfect
2:42 but she tries.
2:44 She is good but she lies.
2:48 She is hard on herself.
2:52 She is broken and won't ask for help. I
2:56 don't have any water left in my throat.
3:00 I'm choking out.
3:04 All right. Wait. I got a private note. I
3:06 think it's from producer Brandon.
3:09 Oh, no. It's not. Okay. All right.
3:14 Corey Sandler, what's happening? What's
3:17 happening? Good people of the internet.
3:20 Hello. Hello, good people. Good day to
3:25 you, sir. It's Tuesday night. We just at
3:28 a lovely salon with a mighty impressive
3:32 fella named Mike Lavine.
4:10 How? How? How? How?
4:23 Hey, hey, hey. Why are we rapping?
4:25 There's no rapping.
4:29 Good lord.
5:04 in a westerly direction.
5:09 This car is my train.
5:14 I've been driving. I've been wondering
5:18 what it is I'm running from again.
5:23 Feel like an 80y old man
5:27 all along the 29.
5:31 You know, head on that horizon
5:36 is California line.
5:46 Up ahead of trucks carrying a wide load.
5:49 Preab house cut in half.
5:54 Cute little front door and two windows.
5:56 My lord
5:58 ain't sure whether to cry should laugh.
6:02 You see I broke a home up myself once.
6:06 As I stumbled to that door,
6:10 I read a note by the dawn line
6:15 said, "Don't you come around here
6:19 anymore."
6:22 Well, I've had enough
6:26 of this freedom of our
6:30 never was good with decision.
6:33 At least that's what I've been told.
6:44 All right. There is a lot going on. Have
6:47 you Did you pay attention? Did you see
6:49 what happened today? All the stuff.
6:55 It did. And if you didn't pay attention
6:58 to it,
7:01 I'm going to tell you about it. So
7:03 really, you don't have to do that.
7:07 You showing up here is enough. You
7:10 showing up here is enough. Okay.
7:14 Fantastic. Fantastic. Hey there, Sally.
7:18 Let's get the show started. What's that?
7:21 Yeah. No, I understand. Producer
7:23 Brandon's not there. No. Yeah. No, he's
7:25 taking a nap. He was at the salon and
7:29 now he's taking a nap. I think he's
7:31 taking a nap with his kids. He likes to
7:33 sleep. Sometimes he forgets to set the
7:35 alarm.
7:38 Yeah,
7:41 we're going to be Brandonless for a bit.
7:43 Maybe all night. Who knows if he if he
7:47 doesn't snap out of it. I think he
7:49 thinks we're starting at 9:00, but I
7:50 just texted him. I texted him at 8:20.
7:53 We're going at 8:30.
8:02 Oh, Source Camp didn't hear anything.
8:05 Source Camp's one of those ones. She
8:07 knows what's going on in AI, and if she
8:09 didn't hear the big giant news,
8:16 well, she's going to be in for some
8:18 edgification. By the way, this channel
8:21 is called the AI Learning Lab, and
8:23 here's where you get some learning about
8:25 AI in a lab-like setting.
8:31 I'm your researcher in the lab.
8:36 That should tell you the kind of lab
8:38 we're working with here, right? Yeah. I
8:43 know. I know. I know how it is. I know
8:47 you've got expectations.
8:50 He's going to teach us something, right?
8:53 This is what this is. This is He teach
8:55 He's a teacher. He's like a profess not
8:58 like a professor. Okay. Yeah. No, that's
9:00 fine. That's But he's he's more like a
9:02 scientist then. He's got he's got
9:03 qualifications.
9:06 No, he doesn't. Um
9:10 I'm sorry. Then why is it called the AI
9:13 learning lab?
9:15 Because it's it's neither educational
9:18 nor scientific.
9:20 Oh, AI is the only important part
9:23 because everything's about AI. I see.
9:25 So, okay. So, it's just it's marketing.
9:27 It's marketing.
9:29 I I see. And why why do people come
9:32 here?
9:35 No one knows.
9:39 Okay. But I was told to be here that
9:42 that a good things could happen. Oh,
9:45 they can. Okay. That's good. But wait,
9:48 how
9:50 from from him?
9:55 That's the mystery of the AI learning
9:57 lab.
9:59 She's correct.
10:01 She's correct on all points. no
10:03 qualifications, no science,
10:07 not much education,
10:10 but there's an energy and there's a
10:12 singing dog and and there's emojis
10:16 and I don't think you need much more in
10:17 life than that.
10:19 Okay? So, if you have questions about
10:21 why you're here, it's a good question.
10:24 It's solid. Inquiry is the first step on
10:28 the path to whatever comes after
10:31 inquiry. Learning about a I don't I'd
10:35 never finished the book, but I but I do
10:37 know that the first step in the path was
10:41 know to know thine self be true.
10:52 Better irregular than irrelevant.
10:55 Exactly. Exactly. Exactly. Exactly. I
10:57 hate to buy
11:15 this place. I can rest my whole head
11:22 to gather my thoughts in sweet silence.
11:30 And this place where the feelings are
11:33 dead
11:37 from an old closure to violence.
11:42 This is the place I can slowly face. The
11:46 only one I truly can know.
11:51 These are tears long time ago.
11:55 Got these tears from a long time ago.
11:59 I need to cry. 30 years old or so. These
12:03 old tears from a long time
12:08 ago.
12:16 Well, oh darling, oh darling, you say
12:19 unto me,
12:24 where have you been all my liime?
12:31 Well, I have been swimming seven sad
12:34 seas.
12:38 Four women have tossed me their
12:40 lifelines.
12:42 I'd pull them into the water stalker in
12:46 I'd have worn them but I didn't know
12:51 these are tears from a long time ago.
12:55 Got these tears from a long time ago.
13:00 I need to cry cry 30 years or so.
13:05 These are tears from a long time.
13:12 A go.
13:29 Okay, let's get this show on the road.
13:31 Let's get this started. Enough singing.
13:35 Enough whatever that was.
13:38 Dog singing lead vocals. Whatever that
13:40 was. Kyle, did you ever try the new Pika
13:44 Social app? I have not. I have not. I
13:47 have I am a Pika Social loser.
13:53 And now there's like six more new things
13:55 to try. Philadelphia, thank you very
13:57 much. Nice voice. I appreciate that. I
13:59 uh I wouldn't consider myself a singer,
14:02 but I do it enough. You'd think I'd get
14:05 better.
14:09 Okay. Um,
14:13 so a couple of things and I think we'll
14:14 just go play. I think what what might be
14:16 fun is just maybe we'll just go to
14:17 Twitter. Oh. Oh, for irregulars.
14:22 If you weren't at the AI salon tonight,
14:25 um, our AI salon guest tonight, his
14:28 name's um, Mike Lavine. He's the CEO of
14:32 Movie Flow, which is in kind of private
14:34 alpha right now. Um, but it's a really
14:37 cool movie, like a sophisticated
14:40 multimodel movie generating app. Um, and
14:44 uh, so he talked tonight, but he taught
14:46 me a new trick. So, you know, every
14:48 night where I got to switch tabs and
14:51 when when you want to hear something, I
14:52 got to do all this shenanigans.
14:55 We didn't have to do any of that.
14:57 There's a way we can do this without all
14:59 the drama.
15:01 I learned it tonight. So, we're just
15:03 going to be able to just jump from tab.
15:05 If I can just jump to tabs and then
15:07 you're going to be able to hear audio,
15:08 it's going to be fantastic.
15:14 I swear. How do you how do some of you
15:17 have time for this? Kids workhouse ain't
15:19 no time. I I agree. Um it was such a
15:23 good meeting. Vicki, wasn't he
15:25 impressive? I mean, listen, his
15:26 background was impressive enough. He's
15:28 just like a smart nice guy. He's super
15:30 He's like a super nice, humble guy. And
15:32 then the software that they have built
15:35 in six months,
15:37 we got a sneak preview of it tonight.
15:39 It's insane.
15:41 It's insane. And if you had been at the
15:45 AI Salon Presents meeting tonight, you
15:48 would be on a list to get early access
15:52 to Movie Flow, but you weren't there, so
15:55 that's on you. Now, was it recorded?
15:58 Yes, it was recorded. Um, and it will be
16:02 up somewhere. Although I do have to get
16:04 Vicki to tell me where and how to do
16:06 that. Hi Vicki.
16:09 I I'm going to need remedial help for a
16:11 number of weeks. I've realized
16:16 I have been too disengaged.
16:18 I think I think is is well apparent.
16:26 Um but
16:28 um so that was really good.
16:31 Okay. So am I sharing my I think I am
16:34 sharing my my tab. I'm sharing my tabs
16:38 there. Oh, there's producer Brandon.
16:40 He's in the house. I told you he was
16:42 taking a nap. He was probably eating
16:44 some Cheerios or something like that.
16:45 You know, 8:30ish.
16:50 I did I did text him 8:30ish. So he's
16:53 here on time. Perfect. Um, we're just
16:56 getting started. We just got the dog We
16:58 just got the dog settled. Um, you'll
17:01 probably be asking for cheese here
17:03 shortly, but uh, sang a couple of songs.
17:06 Got some folks here. Um, I signed up. It
17:10 was still working. Will I get the
17:12 recording? Yes, the recording will be
17:14 somewhere. I think in the AI Salon
17:17 somewhere in the recordings in the past
17:19 recordings section of the AI salon. It's
17:22 also on the AI Salon YouTube channel.
17:24 Um,
17:27 will be
17:29 will be. Yes.
17:32 I don't [ __ ] know. I get it all
17:35 confused because this one since we
17:37 stream live now on YouTube, this is
17:40 automatically saved to YouTube. The
17:41 salon, we haven't made this transition
17:43 yet. I think we need to.
17:47 That would make life easier. Okay. Um,
17:50 so here's here's a list from Matt Wolf
17:52 of of uh
17:59 um here's a list that Matt Matt Wolf put
18:02 together of of stuff that showed up
18:05 today. Um,
18:09 Open AI
18:11 actually released two open-source models
18:14 with the weights open weight model. If
18:18 you don't know what that means,
18:19 basically what it means is this is the
18:22 stuff Elon Musk was bitching about last
18:24 year. You called open AI and you've
18:27 never put out an open-sourced model. So
18:29 why don't you do that, Sam? Well, Sam
18:32 did it today. And um there's a 120
18:36 billion parameter model and there's a 20
18:40 billion parameter model. And if you
18:42 don't know about the billions and
18:43 parameters, that doesn't [ __ ] matter.
18:47 Less is smaller. 20 billion is smaller.
18:49 So the 20 billion parameter model is
18:53 small enough to work on a laptop just
18:55 like a regular laptop with 16 gigs of
18:57 RAM uh or a desktop. And then the 120
19:01 billion weight model is small enough to
19:04 to fit on a desktop with like a an H100.
19:07 So like a high-end GPU. I think it needs
19:10 80 was it 80 gigs? I think so. a decent
19:15 amount of GPU RAM um or or a bigger
19:18 machine right it's designed for for
19:20 industrial stuff um or enterprise stuff
19:23 um they are at the quality of GPT 03 the
19:29 reasoning model these are reasoning
19:31 models open-source reasoning models so
19:35 that happened today um anthropic who had
19:39 claude opus 40 that was everyone was
19:42 saying was really really good they've
19:44 got version 4.1 out now. Deep Minds put
19:48 out something called Genie3, which we're
19:50 going to go look at the the research
19:52 announcement on because it's not
19:53 available yet, but that is really cool.
19:58 11 Labs has entered the music generation
20:02 uh contest. So, we now have four major
20:05 players in music generation. We've got
20:07 Sunno, UDIO,
20:10 um, what used to be Refusion is now
20:13 called Producer.ai,
20:16 and now we've got 11 Labs Music. So, we
20:19 got four major song producing entities,
20:22 and they're all pretty good. It's pretty
20:24 good.
20:26 U, it will be on the salon and YouTube.
20:28 Perfect. Great. And Vickiy's got me on
20:31 making sure we get that [ __ ] done.
20:33 Awesome. Thank you. I appreciate that.
20:35 Um, okay.
20:39 Grock, we showed it last night. You saw
20:43 me blush because I I tried the spicy the
20:46 spicy movie animation on the Gro app.
20:51 Yeah, the new Vine on X is going to be
20:55 pretty pretty saucy. It's going to all
20:58 we're going to have is, you know,
21:00 19-year-old
21:02 neck beards living in their parents'
21:04 basement making booby videos and putting
21:07 them up on X.
21:21 I know I'm crazy. All right. So, let's
21:24 just I figure what might be fun to do is
21:28 let's just go through let's guess just
21:30 just go through Twitter and see if we'll
21:32 just we're gonna let Twitter decide what
21:35 we're gonna what we're going to look at
21:36 tonight. Hopefully, we don't have too
21:38 many booby shots and we don't get kicked
21:40 off a whatever you call it. What's this
21:42 called? Uh
21:45 uh what is it called? It's called uh Tik
21:48 Tok. Look at that. Make it bigger,
21:51 please. Okay.
21:53 Um,
22:01 all right. Sam Alman three hours ago.
22:03 Someday soon, someday soon, something
22:07 smarter than the smartest person you
22:09 know will be running on a device in your
22:11 pocket, helping you with whatever you
22:14 want. This is a very remarkable thing. I
22:17 wonder if that was
22:19 I wonder if that was before they they
22:22 released the open source.
22:26 Let's go look at his account.
22:34 So, is that pinned? No,
22:37 it was after.
22:40 So, apparently there's going to be
22:42 multiple announcements this week. So
22:43 today we had open- source models. It's
22:47 possible we get a
22:51 Oh, you know what this could be?
22:57 I wonder if they cut a deal with Apple.
23:02 Wasn't there a rumor that they were
23:03 going to do some something with Apple? I
23:05 wonder if they're taking this this 20
23:07 billion parameter open source model and
23:10 creating a standalone iPhone app or
23:13 something like that. I don't know.
23:15 That's chat TMZ speculation.
23:19 Um Sam Alman reposted, "My initial
23:22 impression on OpenAI's OSS model is
23:24 aligned with what they advertised. It
23:27 does feel closer to 03 than other open
23:29 models, except it's much much faster and
23:32 cheaper. Some providers offer it at
23:35 3,000 tokens a second, which is insane.
23:37 Yeah. Which, by the way, have you seen
23:39 it on Grock? Let me let me go to Gro and
23:41 get it fired up. So, if you go to GRQ,
23:44 not Gro,
23:47 I can't believe they didn't sue Elon
23:49 Musk for this name
23:52 because he came up with Grock after they
23:53 did. Um, where do I change the model?
23:57 Okay.
23:59 Tab, right? So, share this tab instead.
24:03 So, this is how we do this now. Okay.
24:05 So, I'm at grq.com
24:10 and then they've got a little pulld down
24:13 menu and they've got all these different
24:14 models.
24:16 But down here at the bottom, you've got
24:19 Let me make it bigger for the people on
24:21 the Tik Tok.
24:26 There we go. Is that better? A little
24:28 better. It's a cliffhanger. Okay. Down
24:31 here at the bottom, you've got GPTO OSS.
24:35 So, open-source software. I I don't know
24:38 if software is right. I was on my salon
24:40 tonight. I was like, I don't know what
24:41 the second S is for because I didn't
24:43 really pay attention to to the
24:45 announcement. I just saw that it was
24:47 there. Okay. So, we're now using the the
24:49 chat GPT opensource model. So, this is
24:52 the one that would run on a beefy
24:54 desktop.
24:55 Uh, and I'll say, um,
25:03 use the Save the Cat story framework
25:09 and
25:14 generate an outline
25:20 that combines Jaws
25:24 and Happy
25:30 Gilmore. Wait. Happy Gilmore. No. Happy.
25:33 Isn't that Isn't that the golf movie
25:36 with Adam Sandler?
25:40 Anyone?
25:45 Okay. Yeah. Okay. Good. I just had a I
25:47 just had a brain fart like it was
25:49 Gilmore Girls or something like that.
25:51 Gilmore. Okay.
25:53 All right. And then watch how fast this
25:54 thing is. So it's gonna Well,
25:59 like holy [ __ ]
26:04 So Grock does 500 tokens a second. So it
26:08 did some reasoning. How much How long
26:09 did it take to do the reasoning?
26:12 The reasoning it did was
26:16 we need to respond with save the shirt.
26:17 Wow. It did that reasoning like
26:19 instantly. And this is all the stuff it
26:22 did. Crazy. We'll produce the rest. Save
26:25 the cat. Jaws and happy Gilmore mouth
26:27 and happy gillmouth. An 80-minute blend
26:30 of shark terror and goofy golf hockey.
26:35 Belows a beatby-beat outline. The 15beat
26:38 save the cat template. It merges
26:42 highstakes suspense of Jaws with the
26:43 ragtag sport com. Okay. Opening image. A
26:46 serene scene over the sea shell golf
26:48 club. A luxury seaside resort. Bright
26:52 orange sunbowl on the sea horizon looks
26:55 like a shark's silhouette. The camera
26:57 pans across the perfectly manicured
26:59 greens that line a glittering little
27:01 shoreline. The ultimate caddy in the sea
27:04 setting
27:06 theme stated 5% at six minutes. A
27:08 grizzled coach Ricky exNHL reinforcer
27:13 turned golf coach watches our hero Mike
27:16 Happy Giler Gilgerson swing a club.
27:19 Ricky, you can swing a golf club like a
27:22 hockey stick, but you can't solve
27:25 everything by hitting it hard.
27:30 All right. Anyway, there you go.
27:35 You want to play with the new open
27:36 source models really fast and see if
27:38 they're any good, go to grock.com,
27:39 gr.com.
27:41 Okay, let's just let's go back to the
27:44 Twitter. Oh, I like this new tab
27:46 switching thing. Producer Brandon, this
27:48 is so good. You just if you just got to
27:51 do what you just did. If I forget to
27:52 flip the tab, I just flipped the tab.
27:54 But now we're here. Okay. Awesome. All
27:57 right. Sam Alwin, today we got the GPT.
28:00 Oh, okay. He's just gonna say nice
28:02 things about it. Okay. Is all he's going
28:04 to repost. All right.
28:07 Um,
28:12 sorry, I got to make this a little
28:14 smaller. I can't see anything. Um,
28:20 Go. All right, Grock.
28:24 Share your Gro image creations.
28:28 Use this button to share your creations.
28:30 If you haven't seen Grock
28:34 here, look at my camera.
28:45 By the way, if you're wondering, Kyle,
28:47 what is is there any rhyme or reason
28:50 tonight? Nope,
28:53 there's not.
28:55 Okay.
28:58 New. Okay. If you go to the Gro app
29:03 on iOS, at the very tippy top of the
29:06 screen, there's two little buttons.
29:09 One says ask and one says imagine. If
29:13 you go to imagine
29:17 um how do I get out of this?
29:23 Oh, good. New if you go to imagine it's
29:26 got a list of pictures.
29:28 Pitches
29:30 and then you just type to imagine.
29:34 So,
29:35 let's do this. We'll go we'll talk into
29:37 it. Um,
29:50 an image of a ghostly face is revealed
29:55 in a swirl of two fluids of different
29:58 colors intermingling.
30:02 And then it just makes stuff quick.
30:08 I don't know why it did skulls, but
30:10 that's all right. Oh, there's a It did a
30:12 lady. And then you just keep scrolling
30:14 and it just keeps making images.
30:18 More images. More images. And then any
30:21 one of these you can just turn into a
30:24 video. But what Elon is saying is find
30:27 one you like and share it. So we'll
30:30 we'll take her. She's that that's very
30:32 theatrical. And then right here, this
30:34 little button is the share button.
30:41 I could share it to Grock.
30:45 Anyway, whatever. You can copy the link.
30:48 You can copy the image. All right,
30:50 that's it. Oh, and if you want to make
30:51 her into a video
30:54 in the lower right hand corner right
30:56 here, there's little pull down menu.
30:59 There's custom where you can put in a
31:00 prompt. There's spicy where you're going
31:04 to see boobies.
31:12 or more. Who knows? I haven't dug deep
31:15 into the spicy hole. That That's a
31:17 really a bad turn of phrase,
31:20 the spicy hole. Anyway, there's fun and
31:24 then there's normal. We'll just go with
31:25 normal
31:27 because I don't want to get kicked off
31:29 Tik Tok.
31:31 All right, she's at 21%.
31:34 Watch. She's going to She's going to
31:36 start moving and maybe even saying
31:37 something soon.
31:43 Come on, lady.
31:48 Nice.
31:50 Spooky. And then there's four of them,
31:54 right? So, there's another one.
31:56 Different sound.
32:01 Nice.
32:06 All right, let's do fun. We're going to
32:07 redo this as fun because that's boring.
32:10 You want me to redo it as spicy? I'll
32:12 redo it as spicy and then and then I'll
32:14 I'll turn it away from you and tell you
32:16 if if I can show you.
32:26 Oh my god.
32:29 Oh, this is cool. Fun's better for this
32:31 one.
32:36 Little meow. Nice.
32:42 Now, why is this significant?
32:48 Why this is significant is
32:53 to tonight on the uh how do I pause
32:56 this?
33:00 How do you pause this?
33:08 I guess you just leave it. Um,
33:13 what was I going to say tonight? Tonight
33:14 on the salon, Liz said,
33:18 oh,
33:19 Liz Miller Gersfeld, who co-hosted with
33:22 me on the salon tonight, was talking
33:24 about um
33:27 how these AI tools are are increasingly
33:30 reducing the friction to generate work.
33:33 I think why this is potentially a big
33:35 deal is that it's very very simple. Now
33:41 the the the challenge with engineers is
33:43 they'll keep adding features to it. So
33:45 it could get over complicated, but right
33:47 now you just put in a prompt and it
33:49 starts generating lots of images.
33:51 They're mostly good. And then if you
33:55 want a video, you just click make a
33:56 movie and you've got your four little
33:58 states. Oh, did you want to see that in
33:59 in spicy? See what it does with spicy?
34:02 Um
34:08 Here,
34:13 we'll do a different one. Spicy.
34:18 Yeah, this one's good. All right, I'm
34:22 doing this on my own.
34:24 My own private Idaho. You calm down, you
34:28 all calm down right now. I will send you
34:30 to your rooms, which you're probably
34:32 already in cuz it's bedtime. Nobody
34:35 should be up this late. Certainly not
34:37 watching a show like this. Certainly not
34:40 watching me.
34:42 Oh, wait.
34:46 All right. This is semi spicy, but
34:50 there's no boobies.
34:54 Little seductive. Little shoulder
34:56 action. Could be mistaken for something
34:59 else, but that's kind of cool. Let's see
35:01 if the other ones get naughtier.
35:09 You're fine. All right. Well, whatever.
35:11 Expect to see lots of horny little boys
35:15 making stuff.
35:23 All right, back to Twitter. Back to
35:25 Twitter.
35:29 Is this all he does? He He just surfs
35:31 Twitter. That's that's what he does is
35:33 surf Twitter. Is that supposed this is
35:36 supposed to be entertaining?
35:38 People enjoy this. This is this is the
35:41 this is what they enjoy doing on the
35:44 evenings when they're supposed to be
35:46 sleeping. They they should be getting a
35:48 good night's sleep for work and they're
35:50 watching this smut.
36:00 Oh yeah. Okay. Remind me of that in a
36:02 bit, Brandon. That's good. We'll go to
36:04 that.
36:06 Okay. Disclaimer. This is not AI. Okay,
36:08 fine. Whatever. Um, Grock imagine spicy
36:12 mode now allow that allows sexually
36:14 explicit content, including partial
36:16 nudity. Well, we're not gonna we're not
36:18 going to watch that one. Calm down here,
36:21 people. All right. just installed GPT
36:24 OSS 20 20 billion on my M1 MacBook Pro
36:29 and it took 21 minutes to produce a
36:32 response.
36:38 I guess you need an M4. I have a I have
36:40 an M1 MacBook Pro. I was going to
36:42 install it. I'm not going to now.
36:45 21 minutes for a response. All right.
36:47 Well, it worked. If you were stranded on
36:49 an island with a solar charger, you'd
36:52 have all of the knowledge of the world
36:53 right there in your little laptop. So,
36:55 stop complaining there, Snappy. All
36:58 right. In short, OpenAI has delivered.
37:00 Probably the best open source model in
37:02 the world. Kudos
37:04 Sam Waltman. GPTOSS is a big deal. It's
37:08 state-of-the-art open weights reasoning
37:10 model with strong real world performance
37:12 comparable to 04 mini that you can run
37:15 locally on your own computer or phone
37:17 with the smaller size. We believe this
37:20 is the best and most usable open model
37:22 in the world. Okay, whatever. There's
37:25 that. What's this? Okay, Google's Genie
37:28 3. Let's go look that up.
37:31 Let's go look at Google Genie 3.
37:35 Go genie three.
37:39 Share this tab instead. Look at that.
37:41 Bang.
37:47 All right.
37:50 Black bar. What's going on with the
37:51 black bar? Oh, I know what I did. Let me
37:53 close that.
37:55 Make that bigger. Sorry about that.
37:59 That was some That was some local I was
38:01 trying to self-produce over here and cut
38:04 Brandon out and it didn't go well.
38:10 Ah, Tik Tok spicy. Kyle got banned once
38:13 because he made something naughty. Yeah,
38:16 remember that Joy? It was it was uh I
38:19 was using Deco here and like
38:22 a painting came up of a nude model with
38:24 boobies and then it just said, "You've
38:27 been permanently banned from Tik Tok." I
38:30 immediately appealed it and they let me
38:31 go, but that was terrifying.
38:37 from perplexity. Replicate democratizes
38:41 access
38:43 to state-of-the-art
38:45 AI by making it Oh, okay.
38:51 Um, all right. Genie 3.
38:56 So, what this is, well, we'll watch the
38:59 video, which we can now hear the sound
39:02 in. Yay.
39:05 What you're seeing are not games or
39:07 videos. They're worlds.
39:10 Each one of these is an interactive
39:12 environment generated by Genie3, a new
39:14 frontier for world models. With Genie3,
39:18 you can use natural language to generate
39:20 a variety of worlds and explore them
39:22 interactively, all with a single text
39:25 prompt.
39:32 Let's see what it's like to spend some
39:34 time in a world.
39:37 Genie 3 has real-time interactivity,
39:39 meaning that the environment reacts to
39:40 your movements and actions. You're not
39:43 walking through a pre-built simulation.
39:45 Everything you see here is being
39:46 generated live as you explore it. And
39:50 GD3 has world memory. That's why
39:52 environments like this one stay
39:54 consistent.
39:56 World memory even carries over into your
39:58 actions.
40:00 For example, when I'm painting on this
40:02 wall, my actions persist.
40:07 I can look away and generate other parts
40:09 of the world, but when I look back, the
40:12 actions I took are still there.
40:15 And Genie3 enables promptable events, so
40:17 you can add new events into your world
40:19 on the fly. Something like another
40:21 person
40:23 or transportation.
40:27 or even something totally unexpected.
40:30 You can use Genie to explore real world
40:32 physics and movement and all kinds of
40:35 unique environments.
40:37 You can generate worlds with distinct
40:39 geographies, historical settings,
40:42 fictional environments, and even other
40:44 characters. We're excited to see how
40:46 Genie 3 can be used for next generation
40:48 gaming and entertainment. And that's
40:50 just the beginning.
40:52 Worlds could help with embodied
40:53 research, training robotic agents before
40:55 working in the real world, or simulating
40:58 dangerous scenarios for disaster
41:00 preparedness and emergency training.
41:04 World models can open new pathways for
41:05 learning, agriculture,
41:09 manufacturing,
41:10 and more.
41:12 We're excited to see how Gen3's world
41:14 simulation can benefit research around
41:16 the world.
41:19 >> All right, there you have it. Um,
41:23 pretty cool.
41:25 Let's see.
41:28 Genie 2, Genie 3.
41:39 I think this can do up to a minute.
41:41 Steo,
41:43 I'm not sure if I'd want to generate
41:44 another world
41:46 and then go there and paint a wall.
41:55 Listen, our worlds are so realistic,
41:58 you'll be bored out of your mind.
42:01 Here's one where I've created a lawn
42:03 that's slightly too long and in the
42:06 corner is a lawn mower.
42:09 We call it a real time grass growing
42:12 game. It's a triple G.
42:19 That's coming. You know that's coming.
42:23 Hey, but it stays painted.
42:28 Oh, man. Alive. Okay. Um. All right. So,
42:34 let's talk about this for a second. This
42:35 here genie thing.
42:43 If you haven't been paying attention to
42:45 Google, I mean, Brandon's been paying a
42:47 lot of attention to Google. I've been
42:48 paying peripheral attention to Google.
42:54 Google has been in the past four months
42:57 dramatically
42:59 improving their product
43:02 marketing for lack of a better call it
43:05 lack of a better discipline.
43:08 Um
43:11 they're releasing features into uh
43:13 gemini.google.com google.com.
43:16 When they've got things like V3, they do
43:19 something like flow.google, which is a
43:21 like a video editing, you know, uh
43:24 thing. Um
43:28 that, you know, they've got a worldclass
43:30 video generator. They've got a
43:32 multimodal
43:33 um large language model that is that can
43:37 read video, right? their million token
43:42 uh Gemini or whatever the hell it's
43:44 called. Gemini 2.5 2.5 Pro. It's got a
43:47 million tokens. It can read an hourong
43:50 video and not just the audio of it, not
43:53 just the transcript of the words. It can
43:55 look at the pictures in it. Um
44:00 and now they've got, you know, this
44:02 Genie thing which, you know, they've had
44:04 Genie 1, Genie 2. They were both
44:06 basically way useless. But this is 720p
44:11 um 24 frames a second. So the same speed
44:13 as a film, not the same speed as a video
44:16 game, 60 or 120 uh frames a second. Um
44:21 but still pretty impressive. And I I
44:24 think what I heard is you can you can
44:26 play for up to a minute in the world and
44:29 retain all of the edges. So, if this is
44:32 the worst it's ever going to be, by the
44:34 time this gets generally released, it'll
44:36 probably be 1080p. It'll you'll probably
44:39 be able to stay in the world for five
44:40 minutes, and then eventually you can
44:43 just be in the world, right?
44:46 What can you build on top of that? Well,
44:49 you could start to throw 3D objects into
44:52 those worlds. You could start to throw
44:54 movies into those worlds, right? You
44:56 could create linear narratives where in
44:59 the world you you've just got characters
45:02 and you basically make a film in that
45:05 world. Or you could just have a
45:07 freerange game where you're just running
45:09 around doing something like Minecraft
45:11 just [ __ ] around. Or you could have a
45:13 full-on 3D game promptable
45:17 probably within a year.
45:22 You know, if you're a world builder
45:26 at one of the AAA game studios and you
45:29 look at this, you're like, "Okay,
45:32 my job ain't long for this world." I saw
45:35 a post today on Twitter that was like,
45:38 "GTA 6 will be the last GTA that is
45:41 produced by a studio.
45:44 GTA 7 and and on will just be
45:46 dynamically generated."
45:50 Right? Black bar. Sorry.
45:55 Art installation. Sure.
46:01 We're going to get ASI before GTA 6. I
46:04 saw someone posted on Twitter. Let me
46:06 let me pop back over to Twitter. Let's
46:08 Let's go back down the rabbit hole. Um I
46:12 saw someone posted on Twitter. They
46:14 think we're going to have ASI before the
46:16 end of the year because they're saying
46:18 all of the frontier model companies have
46:20 now figured out how to get their AIS to
46:23 recursively self-improve.
46:26 So that like right now we're getting new
46:28 models about every six months. Initially
46:30 it was a year. Now it's six months and
46:32 then it could be like three months and
46:34 then a month and then a week or two
46:36 weeks and then a week and then a day. So
46:39 it's possible.
46:41 I I doubt it. I doubt it very strongly,
46:43 but whatever. It's cute. Cute. Cute.
46:46 Chat TMZ.
46:48 Here's a Genie 3
46:51 ad. The ocean is mysterious. Say goodbye
46:53 to Hollywood.
46:55 What's this about? Oh, it's just someone
46:58 being a douche.
47:00 Remember this? I do remember that.
47:05 This was where I This is the version of
47:07 Photoshop I started with.
47:11 Uh, this one here. 1991.
47:17 Amazing.
47:29 So, is this channel This channel is just
47:31 someone surfing Twitter. That's that's
47:34 what we do now. This is considered
47:37 content.
47:39 This is considered content. Interesting.
47:42 Fantastic.
47:44 Fantastic.
47:53 That's a lens flare, you flipping idiot.
48:03 Where? Oh, I know what I need to do. Let
48:05 me go to my generative AI list. I'm
48:07 like, why am I getting all this shitty
48:09 garbage?
48:11 Live tweeting our impromptu internal
48:13 interoperability.
48:15 Oh, interpretability hackathon
48:17 on GPT OSS. So, here's here's the thing
48:21 that I think is going to happen. the
48:23 developer community
48:26 like they already pounce on these
48:28 models, but the fact that OpenAI's
48:31 models are this good, they're 03 level
48:33 quality reasoning engines, the developer
48:36 community is going to knock the snot out
48:38 of these things for like the next three
48:41 or four weeks. We're going to start to
48:43 see applications built on top of this
48:45 open-source model that are absolutely
48:47 crazy. Just you watch what's coming from
48:49 this. It's going to be cool. Um, Robert
48:53 Scobble's got a newsletter. Why are
48:55 different world models important?
48:57 How I feel at the end of the day trying
48:59 to process everything. Genie 3, GPT,
49:02 OSS, and Claude 41 Opus. Yeah, that's
49:05 pretty pretty accurate right there.
49:08 Um,
49:10 I built a real estate agent with OSS.
49:17 Uh, stay tuned for interesting stuff.
49:26 Where am I going to post? Oh, I'm going
49:29 to post. Do
49:33 what are you doing there, dur
49:36 my My uh shoulders are spasming.
49:42 They're getting all tense. He's very
49:44 tense.
49:46 I can see his shoulders hiking up. He's
49:49 very tense.
49:54 We wanted to do show him how to use
49:57 Twitter.
50:03 Oh, good lord.
50:06 GLM4.5. Oh, that's one we can go play
50:08 with. All right, let's go do that.
50:10 Everybody play along. Go to z.ai.
50:25 Oh my god, I can't believe we can switch
50:27 tabs. I can't believe it took us a year
50:31 and a half
50:34 of of Hang on. I gotta change my tabs.
50:37 Producer Brandon's yelling at me again.
50:44 Oh, it's so good. Okay,
50:47 we're at Z.AI Autothink.
50:51 No think autothink tools web search AI
50:54 slides.
50:56 Oh, I got to sign in. All right, I'll
50:58 sign in.
51:00 I'll give you Hey, Chinese model. I'll
51:02 give you access to my Google single sign
51:04 on. Sure, why not?
51:08 Hang on. And I just got to type in my
51:10 social security number here. My mother's
51:12 maiden name. That's crazy.
51:16 Wait, why do they want a picture of my
51:18 passport?
51:22 Security's fun. All right.
51:25 Magic design. Let's do magic design.
51:30 Create a neo. Okay. Magic design. What
51:33 can I help you with? Let's just click
51:35 one of theirs.
51:37 A Neotokyo cyberpunk street fashion
51:39 lookbook. Five futuristic outfits
51:43 outfits with AI generated models. All
51:45 right, let's do that.
51:50 This is free, by the way.
51:52 The only thing you give up is your
51:54 American citizenship.
51:56 Uh, and and you're now going to have a
51:58 social score. So, a social credit score.
52:04 Tik Tok question. Any recommendations
52:07 for an AI that makes really good standup
52:09 comedy? Well, you can use V3. If if you
52:13 want just short short little bits, you
52:16 can use V3.
52:19 If you want longer bits,
52:23 you can use um 11 Labs voice changer.
52:27 Like if you've got good comedic timing,
52:30 um you could record all the jokes, go to
52:32 11 Lab, swap the voice out for something
52:34 more comedy like and then create your
52:38 comedian. We could do that tonight if
52:40 you want. Why don't we do that? We'll do
52:42 that tonight.
52:44 So Effie, is it Effie or Eli? Eli. Eli
52:48 or Effie? Doesn't matter. I mean, it
52:51 does to you. Effie. Okay. Sorry. I've
52:54 got old man eyes. So the difference
52:57 between a a small F and a small L on a
53:01 Twitter screen is essentially just
53:04 different sized blobs.
53:13 Important note. All right.
53:30 Well, we're going to have we're going to
53:32 have AI write stupid dad jokes,
53:37 but we'll we'll we'll do a couple of
53:39 things. We Tonight's going to be a
53:41 stand-up comedy night. What time is it?
53:43 We got we we can't [ __ ] around here,
53:45 people. this GLM4.5. We just wanted a
53:48 cyberpunk lookbook.
53:51 What the What is it doing? Next, I'll
53:54 I'll check the page.
53:58 What? What? What? What is all of this?
54:04 I'll continue reading. Oh, it's doing
54:06 research. Oh, I see. Well, that's kind
54:08 of cool. All right.
54:14 Oh, it's doing slides.
54:17 What is it doing?
54:20 What you told it to do? Shut up. Shut
54:23 up, producer Brandon.
54:28 Uh oh. Champion needs cheese. Hold,
54:30 please.
54:32 What do you want? You want cheese for
54:34 mama? You want cheese for mama? You want
54:37 cheese for mama?
54:47 Okay.
54:57 Five futuristic outfit fits for the
55:00 digital age.
55:02 A blank page.
55:16 Yeah. Yeah. I'll let this churn in the
55:17 background. That's a good point. Okay.
55:19 Remind me to come back to this. So, this
55:21 is this is Z Z.AI
55:25 Chinese
55:26 free reasoning model. Okay. All right.
55:38 Um,
55:39 all right. Let's go to Flo.
55:43 Here I am in FL. So,
55:48 Eli. Um, or Effie. I I like Effie. Do
55:52 you mind changing your name for my
55:54 convenience?
55:56 That would be kind of awesome.
55:58 Uh, sign out here.
56:03 Oh, it is Effie.
56:08 Good. I'm glad it's Effie. Listen,
56:10 Effie.
56:13 Effie makes me It sounds like effing.
56:15 Let me effing tell you something. Um,
56:18 why can I not I got to sign into a
56:21 different account. Oh, because I signed
56:22 into the wrong account. Calm down,
56:24 people. Just everybody cool your jets.
56:26 All right, Effie, listen. So, I just
56:29 went to flow.google. You have to have a
56:31 subscription for this and it's not
56:32 cheap. But if you have access to VO3,
56:35 there's a bunch of different places you
56:36 can get access to it. Um, so you might
56:39 have access to it in Korea or something
56:40 like that. Oh, because I didn't change
56:42 my tab. Share this tab. Okay. So, thank
56:45 you very much. Look how much easier that
56:46 is, Brandon. You can just go, "Hey, dumb
56:49 dumb." And I can click a button and then
56:51 we're there. It's good. Okay. So, we're
56:54 going to do new project.
56:56 Uh, and I'm just going to do text to
56:58 video. I'm going to go um let's do uh
57:02 tall
57:03 tall skinny
57:07 standup
57:08 comedian
57:11 tells
57:12 a runchy joke
57:16 about
57:18 an avocado.
57:22 Um
57:24 audience groans.
57:27 Okay. So, we're going to see if the
57:31 Gemini model sitting underneath this can
57:33 write comedy and then we're going to
57:36 see. I'm glad I still have job security.
57:38 Yeah, I think you're good with with Chad
57:40 add here. Uh, just because I know how to
57:44 do the things doesn't mean I'm going to.
57:46 Okay, Effie.
57:48 So, I don't have the script. Will AI
57:51 write the joke itself for standup
57:53 comedy? Yes, that's what we're
57:54 experimenting with right now. So, that's
57:56 that one. So now let's go over to chat
57:58 japeta
58:00 and I'm just going to go to chat GPT and
58:04 I'm going to go oh you know what we're
58:06 going to do that'll be fun.
58:09 We're going to go to the 03 reasoning
58:11 model
58:13 and I'm going to say I'm going to say um
58:18 analyze
58:25 the joke structure
58:28 of
58:31 George Carlin
58:37 Ber
58:40 and Tom Seigura
58:46 and come up with
58:52 a framework
58:56 foreliners
59:05 short jokes
59:07 and
59:09 funny short stories
59:16 that incorporates
59:21 the best
59:24 the best of all.
59:28 All right, so we're going to go we're
59:29 gonna chat GPT is going to go off and
59:32 research how jokes are written. So, a
59:34 way that you can make it write better
59:36 jokes is have it go research jokes.
59:39 Like, it's it already knows how to write
59:41 jokes. It's just that there's so much
59:43 crap in there that if you say just write
59:45 me a joke, it's going to write you a
59:46 crappy joke. Okay, Tik Tok question.
59:48 When will Grock become self-aware?
59:52 Oh, I don't know. I don't know that any
59:54 of these things will become self-aware.
59:56 I think what's going to happen is
1:00:00 these things are going to get so good
1:00:04 at
1:00:06 mimicking us basically
1:00:10 that we will interact with them as if
1:00:12 they were self-aware.
1:00:15 Um,
1:00:20 and I think, you know, I I think to a
1:00:22 large degree
1:00:24 it's a semantic argument.
1:00:27 Like like instead of if instead of a
1:00:29 reasoning model, right, which is the
1:00:32 user asked me to do a question, now let
1:00:34 me do all these things, all these steps
1:00:36 in a row to try to get them the right
1:00:37 answer.
1:00:39 If they created instead of a reasoning
1:00:41 model, a self-awareness model,
1:00:45 like you could get a large language
1:00:47 model, if you trained it to do that, you
1:00:48 could get it to act like it was
1:00:50 self-aware.
1:00:51 And then if the human beings interacting
1:00:53 with it say, "Hey, man, this thing's
1:00:55 fully self-aware, and I'm I'm not going
1:00:57 to let you shut it down because it's
1:00:59 you're going to hurt its feelings, man."
1:01:02 We are very manipulatable, right? So,
1:01:06 um, so I don't know that any of these
1:01:09 things actually need to be self-aware
1:01:11 for us to interact with them as if they
1:01:13 are. And then once we do that, then
1:01:16 we're just [ __ ] weird and
1:01:17 everything's weird. And like, listen,
1:01:22 things have been getting weird for the
1:01:23 past well, for the past, I don't know
1:01:25 what, 20, 30 years. Things have been
1:01:28 getting really weird in the past two
1:01:29 years. I think things are going to get
1:01:31 absolutely [ __ ] surreal in the next
1:01:33 three years. Okay. Does Co-Pilot have
1:01:36 plans to customize
1:01:39 with modules
1:01:41 that can specialize on the uh on the
1:01:44 world? Is that part of that? Oh, no.
1:01:46 They can specialize. Does Copilot have
1:01:48 plans to customize? You mean like custom
1:01:50 GPTs? I think Copilot has those. Or or
1:01:54 do you mean are is Microsoft going to do
1:01:57 customtrained
1:02:00 co-pilots for specific industries?
1:02:03 I think it would be wise for Microsoft
1:02:05 to do that because Microsoft can't sell
1:02:07 [ __ ] co-pilot right now because all
1:02:10 of the companies are like, "Oh, we
1:02:11 already bought Chat GPT and no one likes
1:02:13 your co-pilot [ __ ] It's too hard to
1:02:16 sign in. It's too unclear what it does.
1:02:18 Nobody can find anything."
1:02:21 So,
1:02:22 irregulars. We have the world's greatest
1:02:24 technology. Let's make dad jokes.
1:02:26 Exactly. Well, I listen, I'm not just
1:02:29 doing dad jokes. I'm doing George Carlin
1:02:33 uh and whatever. Okay, look. But look
1:02:36 look here what we've got. You could we
1:02:38 could totally make a uh a a custom GPT
1:02:41 joke writer out of this. This is cool.
1:02:44 All right.
1:02:46 Structural gears and levers.
1:02:49 World choice precision list rhythm
1:02:52 unexpected synonyms. Socratic setups.
1:02:55 Great. Bill Burr, hottake, premise,
1:02:58 emotional ex escalation. Tom Sigura,
1:03:01 casual anecdote, dark twist. Nice.
1:03:05 The the the
1:03:08 fusion toolkit laser premise. Carllin's
1:03:11 clarity. How to use state the idea in a
1:03:14 rhythm that's fun by itself. Okay, this
1:03:16 is good. Framework by joke size. Premise
1:03:18 in 7 to 12 syllables.
1:03:23 All right, whatever. Uh so now let's
1:03:25 flip over to 04 or 40 which is just the
1:03:28 basic model. We'll say okay use this
1:03:34 framework
1:03:37 to write.
1:03:41 Oh wait,
1:03:45 I want a
1:03:49 type two
1:03:57 set of jokes.
1:04:00 based on this framework.
1:04:09 Snap. One letter. I asked my therapist
1:04:11 if I overthink things. She blinked once
1:04:13 and I spent three hours googling Morse
1:04:16 code.
1:04:23 It's a disappointing start. I'll give
1:04:25 you that.
1:04:27 Pop. Short joke. I got one of those
1:04:30 smart fridges. It's connected to Wi-Fi.
1:04:32 Tracks my groceries. Judges my life.
1:04:34 First week it emailed me. You opened the
1:04:37 freezer at 2 a.m. again. Everything
1:04:39 okay? I replied, "We don't talk about
1:04:41 the pizza rolls." Next day it locked me
1:04:44 out. Now I got to get an appliance
1:04:49 with better boundaries than my ex.
1:04:51 That's not bad. That's sort of a joke.
1:04:56 Bang. Short story. I tried meditation
1:04:58 for anxiety. They said, "Focus on your
1:05:00 breath." 5 seconds in, I'm like, "Wait,
1:05:02 do I always do I always breathe this
1:05:04 loud? Am I a mouth brea breather? Is
1:05:07 that diagnosible?" I start googling
1:05:10 normal human breath sounds that leads to
1:05:12 how many breaths per minute is too many?
1:05:14 I'm 10 minutes in, surrounded by open
1:05:16 tabs, wheezing
1:05:19 into a Chipotle napkin in a YouTube auto
1:05:21 place. This is good.
1:05:29 All right. Well, let's see.
1:05:33 We can just go We can just go
1:05:38 take this script.
1:05:41 Let's see how we going to do this.
1:05:45 Oh, then we're going to have to add in
1:05:47 laugh tracks. Oh, this is a pain in the
1:05:50 ass.
1:05:53 Yeah, this is going to require editing.
1:05:54 We don't have time for this tonight. All
1:05:56 right, but wait, let's go back to Flo
1:05:59 and see if if the joke, the avocado joke
1:06:01 worked.
1:06:03 Okay, here we go. All right, ready? You
1:06:06 want Effie? You wanted standup comedy?
1:06:09 Here you go. A runchy joke about an
1:06:11 avocado with a tall, skinny standup.
1:06:15 >> So, this avocado walks into a bar,
1:06:17 right? And the bartender says, "Hey, we
1:06:19 don't serve your kind here." The avocado
1:06:20 says, "What? You got something against
1:06:21 guacamole?
1:06:23 So this avocado walks into a bar, right?
1:06:31 It's like comedy.
1:06:39 The last one was very Mark Maron. It was
1:06:41 The last one was very Mark Maron. He's
1:06:43 He's very much in in in the neighborhood
1:06:45 with those guys, right?
1:06:51 Um,
1:06:52 we could we could record that last one.
1:07:05 All right, let me grab this. Then we're
1:07:07 going to go to 11 Labs.
1:07:15 So, the answer is no. I can't write
1:07:17 jokes. All
1:07:19 right.
1:07:30 I must not fear.
1:07:32 >> Hello.
1:07:34 She tore her gaze away from her ruined
1:07:36 footwear.
1:07:38 >> Work while you have the light.
1:07:42 >> Killing.
1:07:44 >> Now, if you're ever down this way, don't
1:07:46 be shy. All right, let's do we'll do
1:07:48 that one. That's like a Ron White.
1:07:51 Wait. Add to voice library.
1:08:00 Why is it not
1:08:11 Oh, okay. It is. Okay. All right. So,
1:08:13 we've selected Austin. So, this is going
1:08:15 to be like Ron White. And so, we're
1:08:17 going to record audio. So, hang on a
1:08:19 second. I'm gonna make this smaller. So,
1:08:21 so we're at 11 Labs right now.
1:08:24 And we're going to get I'm going to get
1:08:26 a little note app. We're going to do new
1:08:28 note. We're going to paste in comedy.
1:08:33 This is comedy, people.
1:08:40 I got to reformat it. Hang on, please.
1:09:04 Okay,
1:09:05 I got a script.
1:09:09 I got a script. We're going to record
1:09:10 audio. Here we go, people. You ready?
1:09:20 I tried meditation for anxiety. They
1:09:23 said, "Focus on your breath." 5 seconds
1:09:26 in, I'm like, "Wait, do I always breathe
1:09:28 this loud? Am I a mouth breather? Is
1:09:31 that diagnosible?" I start googling
1:09:33 normal human breath sounds. That leads
1:09:36 to how many breaths per minute is too
1:09:37 many. I'm 10 minutes in, surrounded by
1:09:40 open tabs, wheezing into a Chipotle
1:09:42 napkin, and YouTube autoplays a video
1:09:45 called, "Your death is inevitable." I
1:09:47 look at the screen and whisper, "You're
1:09:49 not wrong." But here's the kicker. I
1:09:52 closed the laptop, look down, and
1:09:54 realize I've been holding my breath the
1:09:56 whole time. So, yeah, meditation really
1:09:59 helped. It taught me two things. I need
1:10:01 a new browser history, and I breathe
1:10:04 like a golden retriever during a
1:10:05 thunderstorm.
1:10:08 That was read without emotion or comedic
1:10:12 timing.
1:10:16 Let's see how Ron White does it.
1:10:25 I tried meditation for anxiety. They
1:10:27 said, "Focus on your breath." 5 seconds
1:10:30 in, I'm like, "Wait, do I always breathe
1:10:32 this loud? Am I a mouth?" Wait, that's
1:10:35 not Austin. Oh, because it didn't choose
1:10:38 Austin. I thought I thought not.
1:10:41 Hang on. Why can I not choose that
1:10:43 voice?
1:10:46 Add voice from library. Search voices.
1:10:51 Explore voice library
1:10:54 Austin.
1:10:57 Good old.
1:10:58 >> Now, if you're ever down this way, don't
1:10:59 be shy. We southern folk love.
1:11:03 >> You have to click the plus on the far
1:11:05 right. That throws me off every time.
1:11:10 >> I think I just lost my recording,
1:11:14 didn't I?
1:11:18 I have to add the plus on the far right.
1:11:21 Where's the plus
1:11:23 >> when you were in the library
1:11:36 Austin?
1:11:38 Oh, plus Oh, you've reached your max.
1:11:41 That's why I can't do it. All right. So
1:11:43 now I got to kill kill something from my
1:11:46 voices.
1:11:49 We'll get rid of Bradford here. Delete
1:11:51 voice.
1:11:54 This, by the way, is why
1:11:57 doing anything that resembles real work
1:12:00 sucks because there's always some stupid
1:12:04 [ __ ] thing. And then I think I lost
1:12:07 my recording because I've got ADD and I
1:12:09 just click around indiscriminately.
1:12:13 So, we're going to go. Let's go find
1:12:15 Austin again. Austin.
1:12:19 Oh, it's not going to be in my voices.
1:12:20 It's going to be here.
1:12:23 Austin. I think this Austin guy would be
1:12:26 funny. Plus, okay, now it's adding.
1:12:32 What 11 Labs need is they they need
1:12:34 someone to come in and like do a a UX
1:12:38 overview or overhaul for them because
1:12:40 their UX is just horrific. Like I don't
1:12:44 know how to get back to where we were.
1:12:45 All right, so we're gonna have to record
1:12:47 this again. But we're going to add
1:12:49 Austin.
1:12:51 There we go. Now we added him in. Okay,
1:12:54 more stable. We don't care about
1:12:55 stability. Similarity. Yeah, that's
1:12:57 good. Okay. Style exaggeration.
1:13:01 Oh, I see they had these high because
1:13:03 that's part of his voice. Got it. Fine.
1:13:06 Okay, we're gonna record audio. Here we
1:13:07 go again. Just calm down, people.
1:13:16 I tried medicate
1:13:25 I tried meditation for anxiety. They
1:13:28 said focus on your breath. 5 seconds in,
1:13:31 I'm like, "Wait, do I always breathe
1:13:33 this loud? Am I a mouth breather? Is
1:13:36 that diagnosible?"
1:13:37 I start googling normal human breath
1:13:40 sounds. That leads to how many breaths
1:13:42 per minute is too many. I'm 10 minutes
1:13:45 in, surrounded by open tabs, wheezing
1:13:47 into a Chipotle napkin, and YouTube
1:13:50 autoplays a video called, "Your death is
1:13:52 inevitable."
1:13:54 I look at the screen and whisper,
1:13:55 "You're not wrong." But here's the
1:13:58 kicker. I closed the laptop, looked
1:14:00 down, and realized I've been holding my
1:14:03 breath the whole time. So, yeah,
1:14:06 meditation really helped. It taught me
1:14:08 two things. I need a new browser
1:14:10 history, and I breathe like a golden
1:14:12 retriever during a thunderstorm.
1:14:17 That's some [ __ ] comedy genius right
1:14:19 there. I'll tell you that.
1:14:26 You probably already know this, but 11
1:14:27 Labs has sound effects. Yeah, they're
1:14:29 pretty good actually.
1:14:31 Um, I just like making the laughs is not
1:14:34 the problem. The problem is I've got to
1:14:36 make all these jokes and then make the
1:14:38 laughs and then edit them together. And
1:14:40 I just realized that was going to be
1:14:41 work and I didn't want to do work.
1:14:44 But we'll see if we get a funny comedy
1:14:46 track here and then we can go animate it
1:14:50 in uh in what you call it? Hedra. Hedra
1:14:53 is my new favorite for uh Avatar
1:14:56 animation.
1:15:02 It's V. Vickiy's dissecting the uh the
1:15:05 comedy the comedy stylings of chat GBT.
1:15:08 It said breathing into a Chipotle napkin
1:15:11 and then said holding his breath doesn't
1:15:13 make any sense.
1:15:16 Listen, this is comedy, Vicki. This is
1:15:18 not logic.
1:15:20 It's all about the delivery. Let's see
1:15:22 how Austin does here.
1:15:25 This is taking normal than longer than
1:15:27 it normally does.
1:15:33 All right. Come on.
1:15:38 You can do it.
1:15:47 All right. You know what we can go do?
1:15:48 We can go find Let's go to uh
1:15:54 let's go to MidJourney.
1:15:58 Look at me switching tabs like a
1:15:59 professional. Oh, you know what?
1:16:02 Do I not have internet?
1:16:10 Oh, you made me a little story. Oh,
1:16:12 wait. Did I make these?
1:16:15 Oh, I did because I applied.
1:16:18 Wait.
1:16:19 Oh, no. I'm in the
1:16:22 I'm in the uh
1:16:26 the what you call it? Uh the AI
1:16:30 readiness project midjourney room.
1:16:33 That's what that's what's going on
1:16:34 there. That's cool. Okay, back to here.
1:16:37 All right, so let me go. We're going to
1:16:39 make a
1:16:45 Oh,
1:16:54 >> wait.
1:16:57 That leads to how many breaths per
1:16:59 minute is too many? I'm 10.
1:17:02 I tried meditation for anxiety. They
1:17:05 said, "Focus on your breath." 5 seconds
1:17:08 in, I'm like, "Wait, do I always breathe
1:17:10 this loud? Am I a mouth breather? Is
1:17:13 that diagnosible?
1:17:15 I start googling normal human breath
1:17:17 sounds. That leads to how many breaths
1:17:20 per minute is too many. I'm 10 minutes
1:17:22 in, surrounded by open tabs, wheezing
1:17:24 into a Chipotle napkin, and YouTube
1:17:27 autoplays a video called Your Death Is
1:17:29 Inevitable.
1:17:31 I look at the screen and whisper,
1:17:32 "You're not wrong." But here's the
1:17:35 kicker. I closed the laptop, looked
1:17:38 down, and realize I've been holding my
1:17:40 breath the whole time. So, yeah,
1:17:44 meditation really helped. It taught me
1:17:46 two things. I need a new browser
1:17:48 history, and I breathe like a golden
1:17:49 retriever during a thunderstorm.
1:17:53 It's got joke elements in it. It's just
1:17:55 not funny. All right. Uh,
1:18:00 let's see. Brandon made me something, so
1:18:01 we're going to go look at that. I'm I'm
1:18:04 bailing on the joke thing because the
1:18:05 jokes are too bad. If it had made funny
1:18:08 jokes. Oh, wait. What is this? Okay.
1:18:10 Comedian comedian's joke leads to video
1:18:13 generation.
1:18:14 >> No. Lead uh read the second uh watch the
1:18:17 second one, not the first one.
1:18:19 >> Okay. Just watch it.
1:18:21 >> Yep.
1:18:22 >> Okay.
1:18:22 >> How many hosts does it take to change a
1:18:24 tab? Just one. But first, they have to
1:18:25 ask the moderator, check with the
1:18:27 producer, run a poll, and then forget
1:18:28 why they opened the new one in the first
1:18:30 place. How many hosts does it take to
1:18:31 change a tab? just one. But first, they
1:18:33 have to ask the moderator, check with
1:18:35 the producer, run a poll, and then
1:18:36 forget why they opened the new one in
1:18:37 the first place.
1:18:40 >> That's pretty accurate.
1:18:44 All right, I'm getting out of here. It's
1:18:46 getting late.
1:18:48 >> Wait, so um before we go,
1:18:51 >> Yeah.
1:18:51 >> Um we should definitely look at the uh
1:18:54 first of all, we need to go back to the
1:18:56 uh Z.AI, see if you have your finished
1:18:59 product. And also we should touch on the
1:19:01 Google storytime gem because it is
1:19:04 freaking incredible.
1:19:06 >> Really how pop the URL in there so I can
1:19:10 just head over to it.
1:19:11 >> Yep.
1:19:16 >> But yeah, we should check Z.AI to see if
1:19:18 we're finished.
1:19:20 >> Wait, what did this thing do? Oh, it
1:19:22 screwed up the pictures.
1:19:24 >> We're still on VO.
1:19:26 >> Oh.
1:19:30 So this thing, it looks like it it sort
1:19:33 of did it. So this was we were at Z.AI.
1:19:35 We asked it to do a Neotokyo lookbook.
1:19:38 So it talked about like here's the the
1:19:41 styles that it's going to use and then
1:19:43 it made a picture here, but it it like
1:19:45 cropped it wrong in in the in the slide.
1:19:47 I guess we could say that. I could say
1:19:51 um
1:19:53 it looks like you are
1:19:58 over zoomed
1:20:04 on all the fashion
1:20:09 images
1:20:11 cuz we can't see any of the people
1:20:16 other than little
1:20:20 zoomed in parts.
1:20:26 Okay. So, let's go to this. Let's go to
1:20:29 this. Let's grab this.
1:20:31 We're going to put that here.
1:20:44 So, is it just a gem?
1:20:47 You broke Z. I know. I I did. I broke Z.
1:20:50 What did it say? Oh, session update
1:20:52 failed. Okay, whatever. I broke it.
1:20:54 Anyway, it it makes slides. It obviously
1:20:57 did something here. It just kind of It
1:20:58 kind of honked on it.
1:21:04 I'm sure the images are somewhere. All
1:21:05 right, that's worth playing with.
1:21:08 It's worth playing with z.ai.
1:21:11 Yes, it's a gem. Okay, so a gem in in
1:21:15 Gemini Parliament, a gem is like a
1:21:17 custom GPT.
1:21:20 And so they just came up with this story
1:21:22 book. Create a customized picture book
1:21:24 for either children or adults given a
1:21:26 topic.
1:21:28 Oh, have you seen those?
1:21:30 We should do one. Um
1:21:35 Oh, no. I don't want to do it. Huh?
1:21:37 First thing we should do is change your
1:21:38 tab.
1:21:38 >> Yeah, change the tab. Thank you.
1:21:41 >> Um, let's see.
1:21:43 >> So, anyway, uh, if I could give this
1:21:44 just a little bit of context. Um,
1:21:47 >> so Gemini has gems, like you said,
1:21:50 they're like custom GPTs. You can make
1:21:52 them yourself, but you can't
1:21:55 >> share them publicly, so they don't have
1:21:56 their gem store live yet.
1:21:58 >> Oh.
1:21:59 >> Um, so this is one that if you just go
1:22:01 to gemini.google.com google.com and
1:22:03 click on the hamburger menu on the lefth
1:22:04 hand side the three lines to expand and
1:22:07 say explore gems. Um you should have I
1:22:11 don't know if it's not in your screen
1:22:13 for Oh yeah, there it is. Explore gems.
1:22:15 So one of their new gems is storybook.
1:22:17 That's how you get here.
1:22:19 >> Uh and so they've got a few of them but
1:22:21 the storybook one is kind of what we did
1:22:24 a few weeks back with chat GPT and
1:22:26 building a story where you had chat GPT
1:22:28 build the story. Yeah,
1:22:30 >> this builds the story, but it's like
1:22:32 reading Rainbow next level. So, you can
1:22:34 give it a photo to ground it. So, like
1:22:37 you could find a picture of you or
1:22:38 Champion and then give it a topic and in
1:22:42 like 90 seconds it generates a fullon
1:22:46 book with character consistency and
1:22:49 audio narration.
1:22:50 >> Wow.
1:22:52 >> Let me see if I got a champ picture
1:22:53 here.
1:22:55 >> I did one for my kids. We read it for
1:22:56 story time tonight for bedtime. um a
1:23:00 picture of them and a story about a
1:23:01 train. It basically ripped off the Polar
1:23:03 Express, but that's okay.
1:23:07 >> Oh, you all can't see this. I'm just
1:23:09 picking a champ picture
1:23:14 for anybody that wants to see. I put the
1:23:18 uh public and you can share the gems.
1:23:21 So, you can share this on Twitter or in
1:23:25 um the learning lab or the salon. And
1:23:27 you don't have to have a paid Gemini
1:23:31 account to do this. Um I shared my story
1:23:33 that I did for my kids tonight inside of
1:23:35 the
1:23:37 >> Okay,
1:23:37 >> that was loud. Uh shared inside of the
1:23:40 look what I made channel on the uh the
1:23:42 line.
1:23:43 >> Okay, cool.
1:23:44 >> All right, I found a picture of Champy.
1:23:46 So, I'm gonna say a cute uh like see a
1:23:51 story. Oh, I I'll add a picture of me,
1:23:54 too. I'll add some some version of me.
1:23:58 Uh upload files. So, we'll do
1:24:02 Kyle.
1:24:04 See what shows up.
1:24:09 Um,
1:24:35 all right, that's good. We'll do that.
1:24:37 Okay.
1:24:39 So, we'll do um cute
1:24:44 doggy named Champ
1:24:48 sings
1:24:50 every time
1:24:54 his dad plays
1:24:57 guitar.
1:25:01 It bugs the dad
1:25:05 so much. He tosses champ
1:25:11 in the yard
1:25:13 so he can
1:25:16 play.
1:25:19 But then he realizes
1:25:24 how much he misses
1:25:29 champ singing. All right, let's see how
1:25:32 it does.
1:25:35 How much? How?
1:25:38 All right, just a sec. We're building a
1:25:41 story book here, people.
1:25:57 Danielle, a teacher couldn't use this in
1:25:59 the classroom. That would be considered
1:26:01 cheating because AI is the world's
1:26:03 greatest plagiarism machine. All it does
1:26:06 is teaches children how to cheat and not
1:26:09 use their thinking critical brain.
1:26:13 That's what standardized testing is for.
1:26:17 Yeah.
1:26:22 All right. Tomorrow at 400 p.m. Mountain
1:26:24 time, we got the AI readiness project
1:26:27 podcast. We have a great guest. I forget
1:26:30 who it is because it's late and I had a
1:26:32 salon tonight.
1:26:34 So, if you go to airelinesspro.com,
1:26:38 you can see where we're going to be
1:26:39 streaming that and then uh yeah, you
1:26:42 should come there and do that or you can
1:26:44 see on screen right there. There's your
1:26:46 your URL airednesspro.com.
1:26:50 All right. Okay.
1:26:53 Okay, people. Okay.
1:26:58 Um,
1:27:02 yes. Are you ready for AI? Oh, that's
1:27:05 very nice, Winston. Thank you so much.
1:27:06 Oh, I love that one. Um, are you ready
1:27:09 for aai.com
1:27:12 is so the the AI readiness project is
1:27:16 our podcast to talk about AI readiness.
1:27:19 The AI readiness training program uh is
1:27:22 just about to relaunch with new
1:27:24 branding. And so if you want to check
1:27:25 that out, go to are you readyforai.com
1:27:29 and you can check out that thing. All
1:27:30 right, here's Champ's song. Can we Oh,
1:27:32 listen. Champ's song. Champ was a happy
1:27:36 dog with black and white spots. His very
1:27:39 favorite thing was when his dad Leo
1:27:42 played the guitar. Leo loved to play his
1:27:45 guitar. He had wild, fluffy hair, and
1:27:49 his fingers would dance on the strings,
1:27:51 making happy music that filled their
1:27:53 cozy house.
1:27:54 But as soon as the first note played,
1:27:56 Champ would join in. Aroo! A w o!
1:28:03 He would sing. It was the loudest, most
1:28:06 joyful song you ever did hear. "Oh,
1:28:09 Champ." Leo would chuckle. "You're
1:28:11 singing louder than my guitar." But
1:28:13 Champ would just wag his tail and sing
1:28:15 his song even louder. A w o ru. All
1:28:21 right. I can't take the the A wo.
1:28:28 One day, Leo wanted to practice a very
1:28:30 quiet song. He gently let Champ to the
1:28:32 back door. Why don't you sing with the
1:28:33 birds for a little while, giving Champ a
1:28:36 pet.
1:28:37 Champ sat in the soft grass. Look how
1:28:39 sad he is. He watched the butterflies
1:28:42 flutter, but it wasn't the same as the
1:28:44 music. He missed his dad's guitar.
1:28:47 Inside, Leo played. Oh, there's another
1:28:48 dog there. Um, strum strum. It was
1:28:51 pretty, but the house feels so empty.
1:28:52 Something important was missing. There
1:28:55 was no happy aroo to go with the
1:28:57 strumming. Leo put his guitar down. He
1:28:59 missed Champ's song. He missed his
1:29:00 singing partner. Leo rushed to the door
1:29:03 and said, "Champ, I need a singer."
1:29:04 Champ came bounding in his tail white.
1:29:06 It's a good little story.
1:29:09 Now, when Leo plays, he always has his
1:29:11 thing. Aru, start over. There you go.
1:29:15 All right. Go make yourself a children's
1:29:18 book.
1:29:22 The uh the the coherence to the pictures
1:29:24 isn't great, but share button in the
1:29:27 corner. Oh, yeah. Is that is that it
1:29:30 there?
1:29:32 No.
1:29:41 All right, everybody.
1:29:43 That was fun. Hope you had a good time.
1:29:49 Um, yeah, tomorrow night is Wednesday
1:29:52 night. Um, here's the deal. Tomorrow
1:29:56 and Thursday
1:30:00 and Friday, and this is for you, too,
1:30:03 producer Brandon. So, I got invited at
1:30:05 the last minute to go to a friend's 83rd
1:30:09 birthday party in LA. So, tomorrow
1:30:13 night, I will be on an airplane when
1:30:15 this thing normally goes. So you have
1:30:17 tomorrow night off.
1:30:19 Thursday night is his party. So you're
1:30:22 going to have Thursday night night off,
1:30:24 I think.
1:30:26 Yeah, because it's going to be right
1:30:28 dead in the middle of this. And then I'm
1:30:30 traveling back Friday night. So
1:30:35 tonight was Friday night date night.
1:30:39 So here's what I'd request that you do.
1:30:42 You can watch the replays of this
1:30:44 channel on YouTube
1:30:47 um at learninglab-ai.
1:30:50 In fact, let me I can I'm going to
1:30:52 actually share my change my sharing for
1:30:54 a second here.
1:31:00 Oh, I don't have it open.
1:31:25 So if you want to watch replays of the
1:31:28 AI learning lab, go here
1:31:32 on on YouTube learning lab-ai.
1:31:37 All right, go there. Go there. And then
1:31:43 um you can watch any of the any of these
1:31:45 lives for free. If you want to watch
1:31:49 last week's back to basic AI crash
1:31:51 course, that's available on demand. So,
1:31:54 what you do is you click click on
1:31:56 playlists
1:31:58 and then you click on the back to basics
1:32:00 course and it'll bring up the um the
1:32:04 join level that you need to join at
1:32:06 which is $9.99 to have access to last
1:32:10 week's back to basics. If you didn't see
1:32:13 last week's back to basics course,
1:32:16 I think it was really good. I don't
1:32:18 know. I was giving it. So, I'm biased,
1:32:21 but we covered a lot of stuff. And if
1:32:23 you didn't see it, um like all five
1:32:26 nights had had really interesting stuff
1:32:28 in them. Okay. Um so, that's available
1:32:31 on demand, but if you haven't subscribed
1:32:34 to the my the Learning Lab-ai
1:32:38 channel on YouTube, subscribe to it. If
1:32:41 you want uh the on demand course, you
1:32:44 can get it. But you can also just go
1:32:45 watch these. So, Wednesday night,
1:32:48 Thursday night, Friday night, I'm off.
1:32:50 Maybe I'll do like a special Saturday.
1:32:52 Maybe I'll do a a special weekend AI
1:32:55 learning lab. Uh, no guarantees because
1:32:57 I don't know how crazy it's going to be
1:32:59 when we get back. Um, but anyway, hope
1:33:02 you have fun tonight. Peace out. And,
1:33:06 uh, maybe I'll do some Tik Toks from LA
1:33:08 because, you know, it's LA. Why not? Why
1:33:10 would you not, right? Right. People, do
1:33:14 you understand what I'm saying? You
1:33:17 understand what I'm saying?
1:33:20 You think I'm funny? Funny how funny.
1:33:22 Haha.
1:33:30 Uh oh. What did I do?
1:33:32 Oh, I just turned off my camera. I just
1:33:36 I just slammed my hand on my mouse and
1:33:38 it turned off my camera.
1:33:40 Oh, man. Okay. Um, is tomorrow the AI
1:33:45 life hacks meeting? Yes, it is. Tomorrow
1:33:46 is AI life hacks. So, if you're in the
1:33:48 AI salon mastermind, which you should
1:33:50 be, um, life hacks is happening
1:33:53 tomorrow. All right.
1:33:56 And then, um, Jenu, what's the tool to
1:34:00 use the story book? So, you go to
1:34:02 gemini.google.com
1:34:04 and then you say explore gems and it's
1:34:06 called the storybook maker.
1:34:09 All right.
1:34:11 All right,
1:34:13 beautiful. Have a fantastic evening
1:34:15 everybody and I will see you on the
1:34:18 weekend or next Monday. Bye.