7/10/2025 - Beyond the Hype: A Practical Exploration of Grok 4's Power

Description

Chapters

Transcript

0:00 Ready, champ?
0:02 [Music]
0:14 She came on like
0:22 this woman and a look in her eyes.
0:27 [Music]
0:30 She sat on the stool and he said, "What
0:33 do you want?"
0:37 She said, "Give me a love that don't
0:40 freeze up inside
0:44 [Music]
0:51 me
0:52 in my time, dear.
0:57 But to sit next to you, well, I shiver
1:00 and shake.
1:05 And if I knew love, well, I don't think
1:08 I'd be here
1:12 asking myself if I've got what it takes.
1:19 I do your icy blue heart.
1:24 Should
1:28 I stop?
1:30 Turn what's been frozen for years
1:36 into a river of tears.
1:41 [Music]
1:48 [Music]
1:55 Good day, good people. It is Thursday
1:58 night. I think the Grock announcement is
2:01 still going on. I'm pretty sure they're
2:03 still live. I That was the most awkward
2:08 painful painful
2:12 I have never in my days
2:16 seen a more painful presentation.
2:21 You know, there's videos of Steve Jobs
2:25 on the internet that you can watch,
2:28 right?
2:32 Like you could go into Grock right now
2:36 and you could say,
2:38 "Write me a presentation as if I created
2:42 something like chat GPT
2:45 and do it in a way that's exciting and
2:48 new.
2:50 And as if people gave a [ __ ]
2:57 Oh my god, that was bad. But we now have
3:01 a tool that's smarter than every PhD in
3:05 every discipline. And I know your
3:07 question
3:09 is my question. We have the same
3:12 question, people. And what's that
3:14 question? What is that question? What's
3:17 the question? Hey, archetypal architect
3:20 was shaking. The algo thinks it's a
3:23 thing, so people are going to be
3:24 confused. Um,
3:27 here's the question.
3:29 What the [ __ ] am I supposed to do with
3:31 this thing?
3:36 If we can go find out if it if it can
3:37 count the number of Rs in Strawberry,
3:39 that'll be a start. It got good after
3:42 you left. Yeah, I know. I watched it for
3:44 a little bit. But I saw that they had
3:46 some some video announcements. It's
3:48 going to it's going to be able to read
3:49 and generate video soon, but it's not
3:51 there now. It looks like that was
3:53 October or something like that.
3:56 [Music]
3:57 [Applause]
4:02 I mean, one of the things that struck
4:04 me, it was funny. I had a I had a
4:06 conversation with Leah Faston who's my
4:08 co-founder for the AI salon
4:11 and you know she was just saying like I
4:13 just like these tech companies they
4:16 [ __ ] don't have any humanity and I
4:19 was like well yes they do not because
4:25 that's not who's making these models and
4:28 it was I thought it was fascinating last
4:30 night that
4:35 like not Not only do they
4:38 not only do they not have humanity,
4:43 Grock 4 was so good that they don't know
4:46 what to demo.
4:50 Like one of the demos
4:52 was a guy who's clearly got,
4:57 you know,
5:00 double digit, you know, multiple
5:02 double-digit IQ points ahead of me.
5:07 very very smart young man
5:12 and and what he demoed was a math
5:14 problem that he didn't understand
5:18 and it solved it and gave him an answer
5:20 that he didn't understand but he could
5:22 look at how it solved it and said it
5:24 must be right
5:27 like
5:29 you know really
5:34 and you every every one of these every
5:36 one of these model releases
5:41 you've got people demoing it that you
5:43 know aren't from the humanities they
5:44 don't you know back actually back to
5:46 Steve Jobs right Steve Jobs I say I
5:48 would say is the last great tech leader
5:53 that not only acknowledged the
5:55 humanities but basically said Apple is
5:58 as successful as it is because of the
6:00 melding of art and science
6:03 Right?
6:05 You know, why did Open AI just buy
6:07 Johnny Ies
6:11 right now? Open AAI not great at
6:14 humanities, right? Their their demos are
6:17 awful. They're basically just showing
6:19 you engineering [ __ ] Like, why is there
6:21 not a screenwriter in there saying,
6:23 "Here's how to do storytelling with this
6:24 thing. Here's what it can do different."
6:27 Right? Here's here's here's an artist
6:29 showing some of the art stuff it can do
6:31 or video stuff it can do. Right?
6:33 Um,
6:37 I don't think we can
6:41 think we should hope for the frontier
6:45 model companies to get better at the
6:47 humanities stuff. I just don't think
6:49 they're going to do it. And so I think
6:51 that puts it on individuals and I think
6:54 it puts it on organizations like the AI
6:57 salon
6:58 and she leads AI
7:01 to do that. Uh Microsoft investing $4
7:04 billion in AI education. Good. They
7:07 [ __ ] should.
7:11 But their interior design is off the
7:13 charts.
7:16 They've got IKEA. Well, what was funny
7:19 was the the last Grock announcement
7:22 looked like a bad version of an OpenAI
7:24 announcement. And last night, the the
7:27 one memo that Elon Musk got from Steve
7:29 Jobs was, "Everybody wear black and
7:32 we're going to turn the lights down on
7:33 the stage so you can't see it. It's just
7:35 going to be like a black hole with our
7:37 faces sticking out of it."
7:40 Um but then there was the um inability
7:44 to talk part
7:47 and inability to know who was talking
7:49 next part. It was so bad. It was so bad.
7:53 But they announced they they they
7:55 released a model that you know is is
7:58 quite amazing on the benchmarks. I
8:00 haven't played with it a ton. We'll
8:02 we'll dick around with it a bit tonight,
8:04 but I've got some access to it within
8:07 Twitter, but if I go to grock.com, I
8:09 don't have access to it. So, I don't
8:11 know why that is. So, I might only have
8:13 like a a a a run or two left before I
8:17 have to pay for it.
8:21 And I'm kind of like the guy from Grock
8:25 that
8:26 he didn't know what to put in it to demo
8:28 it. So, I don't either. Maybe I'll do
8:31 some I'll maybe I'll do one of those uh
8:34 some sort of creative creative
8:36 challenge. I'll have to maybe we'll go
8:39 over to chat GPT.
8:41 Oh, I think I did this before. I wrote a
8:43 whole article about this. I did. I know.
8:45 I know where I've got it. Um,
8:53 actually, Brandon, if if you can jump
8:57 over to LinkedIn
8:59 and I wrote an article,
9:01 I don't know, four months ago
9:05 that was like how to use reasoning
9:07 models for
9:10 um
9:12 for arts and humanities.
9:16 [Music]
9:20 on my laptop. It's okay, but Kyle on the
9:22 TV is wearing a Story Vine shirt. Yeah.
9:28 Aren't I on a Isn't that
9:31 Am I on your TV correctly? Story Vine,
9:40 redo some older things and see how how
9:42 much better the results are.
9:45 could do. I I tried uh I tried an Eminem
9:48 rap in uh with quantum mechanic
9:51 mechanics. Oh, you just need to sit up.
9:53 Yes, I was slouched over. Oh, you know
9:56 what it is also? There you go. I had my
9:59 little my little camera was a little
10:01 low.
10:03 There you are. Night. Night people.
10:05 Night.
10:10 Hair check. Yeah, it is.
10:14 It's special.
10:17 [Music]
10:35 There's been something, baby, I've been
10:38 trying to say
10:41 for an age. And it seems I don't know
10:44 how.
10:45 [Music]
10:48 Past and a future now surrounding me.
10:53 Surrender to whatever cheap thrill can
10:55 be found.
10:59 There's been little trouble
11:01 [Music]
11:02 since you came to my rescue.
11:05 [Music]
11:09 And if you like all of the rest, I would
11:12 have quit you long ago, but I couldn't
11:14 do that.
11:20 Oh, tell me now. Women in W never went
11:24 too well.
11:27 Make a man crazy, make him cold as hell.
11:33 I a woman that you wish me well.
11:37 But in spite of your trying, still going
11:40 to have to find my own way through.
11:50 Um
11:53 [Music]
11:57 All right.
12:04 Okay. Who got access to the comet
12:06 browser? I have not gotten access to the
12:07 comment browser. I did get access to
12:10 some agentic browser
12:12 two weeks ago, but I couldn't it it was
12:15 not clear what to do with it.
12:22 So, if anyone has any comet invite codes
12:24 and you want to uh kick one over here,
12:27 that would be great. Ah, cool. Thank you
12:28 very much there, Brandon. Brandon.
12:31 Brandon, you're a good man. You're a
12:32 good good man there, dear. Dear Brandon,
12:36 that's what I'm saying there. Yeah. That
12:39 Brandon, he he's all right. He's an all
12:41 right man. He He's pretty good. He
12:43 pretty good. Yeah. All right. I like
12:46 Ters. Tater's tasty.
12:49 Never answer these questions again. Chat
12:51 GPT is free. Vibe coding. Which one is
12:55 it? How I used the reasoning engine.
12:57 That's the one. Shang dang dang dang
13:00 dang dang. Look at that cool image.
13:05 That's me. That's me.
13:09 That's the reasoning engine. You see?
13:12 It's so much bigger than me. I don't get
13:14 it. How do I use it? Well, Grock's like
13:17 10 times that big.
13:34 I use the opera browser. It has embedded
13:36 AI, but it doesn't but doesn't do much
13:39 with it.
13:43 I did. I did. Neither. Oh, you got
13:45 access. You se slug of doom got access
13:47 to comet? Will you turd monkey, do you
13:50 have any extra access codes? Come on,
13:54 give them up. Someone DM me an access
13:57 code, you pin heads.
14:00 I'll go demo it.
14:03 Perplexity. 200 bucks a month. Yeah, I
14:05 know. Exactly. I saw that 200 bucks a
14:07 month for Perplexity. I'm like, you're
14:08 joking, right? This is comedy. It's
14:10 comedy, right?
14:19 Burn.
14:20 [Music]
14:55 You've got a flash goal. I got a flash
14:57 goal on TikTok. People start flashing.
15:00 [Laughter]
15:06 Get Get your privates out. We've got a
15:09 flash goal.
15:15 Come on, people. Let's get banned for
15:17 life on Tik Tok. No, don't do it. If
15:23 I understand I understand what you're
15:24 saying. Um
15:27 All right. It looks like my LinkedIn has
15:29 calmed down a little bit, which is nice.
15:31 All right.
15:34 All right. How I use the reasoning
15:36 engine. All right. So, when did I write
15:37 write this? I wrote this January 31st.
15:40 Are you like me and can't quite figure
15:42 out the difference between your trusty
15:44 chat GPT and a reasoning engine like 01
15:48 DeepSeek R1, Gemini Flash Thinking, or
15:51 the shiny new 03 Mini?
15:56 Start prompting on Korea.
16:01 [Music]
16:03 What is a comment browser and why would
16:04 I use one? Fantastic question, Vicky
16:07 Baptiste for the win. Let's we'll come
16:09 back to the reasoning engine thing here
16:10 in a second.
16:12 Fantastic question here Vicki. Les
16:16 that's a really good question. Fantastic
16:18 question. So there's this new browser
16:21 coming out called Perplexity. You see or
16:24 no sorry from Perplexity. Good lord.
16:28 Geez. You know you try to do a show and
16:31 boy that's it's rough because you got to
16:34 read the teleprompter if you actually
16:36 want to say the words, right? You know,
16:39 you can't just do it off the top of your
16:40 head like some people I know. Anyway,
16:44 new browser from Perplexity called
16:47 Comet. And it is an AI browser.
16:51 Oh, I don't use Perplexity. Yeah. Well,
16:53 so you don't need you don't need to pay
16:56 $200 to to Perplexity. $200 a month to
17:01 be able to have Comet, but if you want
17:03 to get off the wait list, that'll get
17:05 you off it. So, I'm still on the wait
17:07 list. Um, so Perplexity is kind of like
17:12 a a smashing together of a search engine
17:15 and a large language model. It's really
17:18 good. It's a very powerful site. Well,
17:20 they just launched a browser, which
17:21 apparently OpenAI is also launching a
17:24 browser. What is the difference between
17:27 a
17:29 an AI
17:31 browser and a regular browser is that
17:35 all of the large language model search
17:39 stuff is built into the browser itself.
17:42 And so what that means is as you log
17:45 into your websites,
17:47 you're already logged into your
17:48 websites. So, you can just say, "Hey,
17:51 run over to LinkedIn and check this,
17:52 that, or the other." Or, or go to my
17:54 Gmail and find the latest five emails
17:57 and, you know, tell me if you think
18:00 anything's important or tell, you know,
18:01 find the latest emails for the past hour
18:03 and tell me if you think there's
18:04 anything important there. So
18:08 it it'll it allows you to, you know,
18:12 basically
18:14 do what you do sort of in chat GPT, but
18:17 where you have to like copy in URLs and
18:19 do this and do that. Um I have Wait,
18:23 it's Agentic, too. Okay, so Oh Lord,
18:26 browser wars. Yes, new browser wars are
18:28 coming. Absolutely. Maybe Netscape will
18:30 make a comeback. Yes, exactly. Yes,
18:33 that's the question I was going to ask.
18:35 Which was the question you were going to
18:36 ask? What the hell is Comet?
18:42 So, it's better or it's just Musk's
18:45 word? Grock 4. Um,
18:50 I don't know.
18:52 It I It's not just Musk's word. I mean,
18:55 if you look at the um
18:58 what's it called? the uh not humanity's
19:02 last thing but the the arc agi post.
19:06 Hang on. Let's do arc arc agi
19:12 gro 4.
19:16 So this is from
19:22 we got something from arc ai yesterday.
19:28 Oh, but now everyone everyone just
19:30 [ __ ] copies everything, so it's hard
19:32 to find
19:34 the actual post. RKGI.
19:39 H
19:52 ARC- AGI
19:57 2.
20:00 Oh, I'm not showing that. All right.
20:04 Uh,
20:07 well, whatever. Um, I can't find the
20:10 original post, but here
20:14 you can't really see it all that well,
20:16 but like all of the cluster
20:19 all of the cluster of performance down
20:22 here.
20:24 So, there was this test called the the
20:26 Arc AGI test and all of the models blew
20:29 through it and and 03 I think it was 03
20:32 mini that blew through it first and then
20:35 03 high like really blew through it. So,
20:38 this is the Arc AGI 2 test.
20:42 And so all of this crap down here is all
20:44 the other browsers. The one at the very
20:47 tippy top here is Claude Opus 4, their
20:52 thinking model. and and right right
20:54 there just below it is 03 high. So these
20:57 were the two state-of-the-art models.
20:59 Here's Gro 4. So Grock 4 is up at 16%.
21:05 These things are all at single digits,
21:07 eight, eight and a half. So it's almost
21:10 twice as
21:12 twice as good at this benchmark. Now I
21:16 forget what's all in the ARC AGI
21:18 benchmark. It doesn't necessarily me
21:20 measure AGI. Um but there's also a thing
21:23 called humanity's last exam
21:30 humanities
21:36 last exam
21:39 Grock 4
21:46 and they got it up to 50% and and like
21:49 there's there's No, no one should be
21:52 able to get to 50%. Like every question
21:54 on this test are ridiculously hard,
22:00 right? Like like even PhD students
22:02 couldn't answer these questions.
22:04 Access to Grock 4 is already in comet as
22:07 well. Very cool. The web is getting so
22:09 cluttered. We need AI to to reread
22:12 everything.
22:14 Well, that's because so so when they
22:17 talk about Comet being agentic, that
22:20 means you're going to be able to just
22:21 tell it, you know, go off to Twitter and
22:25 find anyone who commented on my stuff
22:27 and categorize them into positive and
22:30 negative, you know, AI sentiment if
22:32 they're AI related. and then go to go to
22:35 LinkedIn and do this for me and go here
22:37 to do that and take this spreadsheet
22:39 that's in Google Sheets and take all
22:41 those emails and and go it send them off
22:44 in Gmail. Um,
22:50 we're we're very quickly entering
22:55 an era
22:57 where individual websites get a lot less
23:00 relevant. Like I would argue that we're
23:02 probably already in that area. But when
23:04 you have full full access to all of the
23:08 sites that you log into
23:10 and you've got kind of native access to
23:13 large language models being able to
23:14 interact with them and you can control
23:16 that. Um you don't need to go to those
23:20 websites,
23:22 right?
23:24 So
23:26 do do clicks matter? And then the other
23:29 thing is going to be there's gonna be
23:30 all sorts of weird [ __ ] like is
23:34 is the New York Times
23:37 gonna throttle access to a browser like
23:40 Comet because it can just go in and
23:43 summarize all those articles and come
23:45 out and write article synopsies that I
23:47 can then share with the world. I didn't
23:50 have to go read those articles. I didn't
23:53 look at the advertising.
23:55 It just went and scraped all that [ __ ]
23:57 out of New York Times for me. So, I
24:00 think we're gonna see we're gonna have
24:01 browser wars. We're gonna have media
24:04 content deals.
24:06 Websites and SEO is going to get really
24:08 weird. Tik Tok Pavin, any comments on
24:11 new YouTube policies for AI generated
24:13 content for monetization? Yes. Um,
24:18 my initial reaction when when
24:21 I saw I think it was a Tik Tok video I
24:23 saw today that said YouTube, you know,
24:25 has has banned AI AI content.
24:30 My immediate response was, well, what
24:32 about the people that actually put time
24:33 into videos like Kelly Bosch or Joy
24:35 Party or, you know, things like that.
24:38 Um, and it and it the more I learned
24:41 about it, all that stuff is exempted
24:44 from this policy. I don't know how
24:46 they're going to determine that, but
24:48 basically what they're saying is if
24:50 you're putting up 10,000 videos a day of
24:53 like people talking about some product
24:55 or whatever it is, um, you're not going
24:57 to be able to monetize that anymore. And
24:59 it's coming fast. It's coming like July
25:01 15th. So that's next Monday or Tuesday.
25:05 So, next Monday or Tuesday, all of these
25:07 farms that are that that have written
25:09 scripts that generate avatars that make,
25:13 you know, YouTube shorts and things like
25:14 that, and they just vomit them out there
25:16 by the thousands, all of that stuff's
25:18 going to get demonetized.
25:20 What you're probably going to have to
25:21 do, so I wrote this article on LinkedIn,
25:25 I don't know what was it, three, four
25:27 days ago, five, six days ago, called
25:30 Chain of Craft. Remember remember when I
25:33 wrote that that little innocuous article
25:35 called Chain of Craft that basically
25:37 said, "Hey, if you've
25:41 had considerable human contribution to
25:45 generate your output, you should start
25:47 documenting that and you should start
25:49 talking about that."
25:51 This is why,
25:54 right? because
25:58 we're whether you want something
26:00 copywritten or whether you want to be
26:02 able to monetize your videos on YouTube,
26:05 at some point YouTube might say, "You've
26:07 got to prove that this is not just
26:09 slop."
26:11 So,
26:14 um, and I think there's a there's
26:16 someone commented on that article
26:18 amongst the AI haters was someone put in
26:22 a snarky comment. They're like, "Yeah,
26:24 this isn't your idea is not that big a
26:27 deal. We came up with this six months
26:28 ago or whatever it was." Which I'm like,
26:30 "Okay, fine." Um, but they have a whole
26:32 business around it. I think they call it
26:34 Chain of Creation, and they're creating
26:36 a chain of creation documenting
26:39 platform. Great. Perfect. Awesome. Um,
26:44 yeah, demonetize chain of crap. Exactly.
26:48 Um, note to self, use projects more.
26:50 Yes, use projects more.
26:54 Um, if you don't know what projects are,
26:56 projects in chat GBT are really
26:58 powerful. Um, and I can I can we can go
27:01 play with those a little bit.
27:04 Um, all right. Any other questions?
27:06 Okay. So, we talked about comet. What
27:08 comet? What comet comet is?
27:12 Um, let's go back to LinkedIn here and
27:14 look at this little article I wrote
27:15 about reasoning engines. Okay.
27:22 prompt. Analyze this screenplay for
27:24 logic. So, wait a minute. If I go to
27:29 If I go to Grock,
27:32 I can upload stuff, right? Okay. So, if
27:36 I do,
27:38 did I do virtue?
27:42 Wait. V do I do R. T
27:49 dot U
27:55 E
27:57 name contains virtue.
28:03 Okay.
28:07 So, I'm gonna So, I've uploaded a a
28:09 screenplay
28:13 that I wrote with my own hands on a
28:16 typewriter
28:18 because you can never do anything good
28:20 if it's AI. So, this is only done with a
28:25 typewriter and a writing partner who has
28:27 a real flesh brain.
28:31 His name's Brian Dyster and he's super
28:33 talented.
28:36 He did slam poetry on the Chappelle
28:38 show. He's that good.
28:47 Okay. Analyze this screenplay for
28:48 logical continuity errors.
28:52 Identify any character actions or plot
28:54 events that contradict earlier
28:56 established facts. Please summarize
28:58 inconsistencies
29:00 and suggest possible fixes. Well, I'm
29:02 going to I'm going to change this. We're
29:04 going to
29:06 intermediate analyze this branching
29:08 narrative. Evaluate this story outline.
29:10 All right. So, let's go back to Grock.
29:15 So, we're going to pop that that prompt
29:17 in there. Analyze this screenplay for
29:19 logical continuity errors. Let's say
29:21 analyze this screenplay
29:24 um
29:28 and identify
29:33 um
29:37 references
29:40 that
29:42 feel out of date
29:46 or scientifically
29:50 um
29:52 illogical
29:55 and suggest
29:57 changes
30:00 that
30:02 bring it into the future, into the
30:05 present.
30:12 Be sure to include the latest
30:18 breakthroughs
30:20 in AI
30:23 um uh what's it called? Quantum
30:27 computing
30:29 and uh
30:34 bio
30:36 computer
30:37 interaces.
30:40 Um
30:43 also
30:46 analyze
30:48 logical continuity errors, identify any
30:51 characters or actions and provide a
30:53 summary of inconsistencies.
30:55 Um
30:59 finally,
31:01 um
31:03 give an analysis
31:07 of
31:08 what
31:11 the screenplay
31:13 got right or gets right
31:19 from both a
31:24 sci-fi
31:26 and storytelling
31:29 perspective
31:33 and
31:34 um
31:44 give a report
31:46 on why modern
31:50 audiences
31:53 would like
31:56 this story. All right. So, this story is
32:00 called Virtue AN says. Okay. Okay.
32:05 Analyze this screenplay for
32:09 and identify references
32:12 wait for
32:15 references that feel outofdate or
32:17 scientifically illogical and suggest
32:20 changes that bring it into the present.
32:22 Be sure to include the latest
32:23 breakthroughs in AI, quantum computing,
32:25 and biomputer interfaces. Also, analyze
32:28 logical continuity errors. Identify any
32:31 character actions or plot events that
32:33 contradict earlier established facts.
32:35 Provide a summary of inconsistencies and
32:38 suggest possible fixes. Finally, give an
32:40 analysis of what the screenplay gets
32:42 right from both a sci-fi and
32:43 storytelling perspective. And give a
32:45 report on why modern audiences would
32:47 like this story. So, what this story is
32:49 about.
32:52 Um, I need to ask chat more to come up
32:54 come up with more dad jokes. That's
32:56 pretty good. Hey, A. Murphy, what's
32:57 happening? Source Camp's in the house.
33:00 She was late, but she's here. Um, okay.
33:05 So, this is a screenplay about a cop
33:08 who's framed for his partner's murder
33:12 and he's taken off the case because he's
33:14 a suspect,
33:16 but he but he's a rogue cop. He's not a
33:19 rogue cop, but Source Camp was on time.
33:21 I don't buy it. I don't buy it. I think
33:23 producer Brandon uh has been um paid off
33:27 by Source Camp. I He's been defending
33:30 her a lot lately. I listen, it's total
33:34 speculation. There might not be any
33:36 money exchanging hands, but I'm just
33:38 saying,
33:40 as evidenced by the behavior,
33:42 something's up,
33:44 right?
33:47 She's pay she's paying me in custom
33:49 GPTs.
33:51 At least someone's paying you.
33:54 Okay. Um, so he's he's framed for his
33:58 partner's murder. And the more he
34:00 investigates it, the more it seems like
34:03 he did it, but he knows what he was
34:06 doing that night. He was banging some
34:09 skanky chick from a bar. Okay.
34:13 Or was he?
34:16 So what what the the plot of the movie
34:18 is, it's like Manurion Candidate where
34:21 they've created a a uh a series of five
34:25 biological molecules that they inject
34:27 into the base of your brain and they
34:30 assemble at the base of your brain and
34:32 become a transduc a transmitter uh a
34:35 transceiver. It can receive signal and
34:37 it can send signal of your brain
34:39 activity and so they can broadcast in
34:42 someone else's experience. So your brain
34:46 experiences what they did physically and
34:48 then they can drive your body around
34:51 with someone else in a in a tank going
34:53 off to do whatever deed they wanted to
34:55 do. So that's that's the screenplay. Um
34:58 the the the thing at the base of the
35:01 brain, virtue is the name of the
35:03 screenplay and virtue is an acronym that
35:06 stands for for v virally integrated
35:09 reality transceiving unified enzyme
35:13 which I think sounds pretty good and I
35:15 think it might be plausible but like
35:18 maybe not. So let's find out. Let's have
35:20 Grock tell us how [ __ ] shitty our
35:23 sci-fi is.
35:28 All right, here we go. Grock is doing
35:31 its thing. It's thinking. It's thinking.
35:33 Now it's not doing anything.
35:35 The G is animating. Right there's the
35:38 little G animating.
35:42 [Music]
35:48 I wonder if I don't have access to Gro 4
35:50 anymore because it doesn't say thinking.
35:54 It's just doing its thing. Can I click
35:56 on this? No.
35:59 Um, T-Rex and a typewriter sounds like
36:02 V3 prompt worth trying.
36:06 The think and deep research button has
36:08 disappeared in my Gro 4. Is that normal?
36:10 Um,
36:12 probably if if they're just
36:15 automatically doing it anyway. If that's
36:17 part of Grock 4. Talk about your
36:19 two-finger typing. I listen I am a solid
36:24 I'm a solid five five finger typer.
36:29 Okay. A PDF. Okay.
36:31 The screenplay appears to be a 1990s era
36:34 draft evidenced by cultural and tech
36:37 references like 1980 1980s gun buyback
36:41 programs, 13-inch TVs, answering
36:44 machines, and a focus on nuclear waste
36:46 disposal as a plot device.
36:52 This is great. This was written in 1994.
36:56 U and it was supposed to be set kind of
36:58 in the present. It draws from cyber
37:01 cyberpunk tropes, VR immersion tanks
37:04 popularized in films like The Lawn Mower
37:06 Man and Virtuosity in 1995. Several
37:09 elements feel dated or illogical in a
37:11 2025 context. Outofdate technological re
37:15 uh references the VR immersion tank. The
37:18 lab describes a man in a wet suit with
37:20 circuits. Fine.
37:24 Fine. Um, computer screens and molecule
37:27 docking. The opening montage shows
37:29 molecules docking on a computer screen
37:32 in black space with glowing repulsions.
37:35 Feels like 1980s or 90s CGI. Eg. Early
37:39 molecular modeling software.
37:42 Modern simul simulations use AIdriven
37:45 tools like Alphafold, Google Deep Mind,
37:48 and protein folding predictions rendered
37:50 in high-res 3D real-time interactions.
37:52 It's exactly I I knew that. I could have
37:55 said that.
37:58 Nuclear reactor in a secret facility.
38:00 CAT scans for migraines. Tony gets a CAT
38:03 scan, likely a CAT scan, now called a CT
38:06 scan for headaches. Dated. Modern
38:08 diagnostics use MRI or fMRI for brain
38:12 issues, especially migraines, which can
38:15 involve neuroinflammation
38:17 visible via advanced imaging, answering
38:20 machines and landlines, gun buyback
38:22 program, media and news. Okay.
38:25 Scientifically illogical elements. This
38:28 is pretty good. This is pretty good. Um,
38:32 mind transfer and cloning. Tony is
38:34 revealed as a copy of Dennis via mind
38:37 uploading or cloning with migraines as a
38:40 side effect. Illogical because current
38:42 neuroscientist shows consciousness isn't
38:45 really transferred like data.
38:48 Yeah, but you got to have some [ __ ]
38:51 suspension of disbelief here. Grock,
38:53 give me a [ __ ] throw me a bone. It's
38:56 tied to a biological process. No
38:58 evidence for full mind uploads. F, shut
39:02 up.
39:04 But it did mention Neurolink, so it's
39:06 got Neuralink in there. The heartburn
39:08 backdoor code triggering a meltdown is
39:11 cartoonish.
39:13 Oh no. Real systems use multiffactor
39:17 authentication and AI anomaly detection.
39:20 I knew that. Nuclear waste melting
39:24 people. VR induced real world effects.
39:28 Air shaft escape during meltdown. I
39:30 always hated that scene. And migraine is
39:32 a plot device. All right. Suggested
39:35 changes to bring it into the present.
39:37 So, let's see if this does any good.
39:39 Let's see. Let's see if it can if it can
39:41 help my screenplay.
39:45 Oh my god, that's funny. Okay.
39:54 Uh, do we like Grock 4? I don't know. I
39:58 don't know yet. I mean, this seems
40:01 decent.
40:04 To modernize, integrate 2025
40:07 breakthroughs, multimodal LLMs like
40:09 GBT4, quantum computing, eg error
40:12 corrected cubits from IBM, Google,
40:15 enabling unbreakable encryption, and
40:17 BCIS, brain computer interfaces, neural
40:20 links, brain implants for thought to
40:22 action. Aim for plausibility while
40:24 keeping a thriller pace. Update VR to
40:27 BCId driven metaverse. Fine.
40:31 Replace immersion tanks. Oh, that's
40:33 good. With Neurolink style implants.
40:35 Yeah, that's good. Where it's more like
40:37 playing a video game. That's fantastic.
40:39 Incorporate AI breakthroughs. Virtue
40:41 tech uses AI for predictive policing or
40:44 deep fake propaganda against Karns.
40:47 Powers could employ LLMs to generate
40:49 disinformation campaigns. That's good.
40:52 the heartburn code becomes an AI
40:55 backdoor, eg a prompt injection
40:57 vulnerability in a gro-l like model
41:00 controlling the facility. So I got to
41:04 say that
41:05 Grock in this case
41:08 is actually doing really good at what
41:11 it's replacing the [ __ ] with.
41:15 Really good
41:17 quantum computing integration. The
41:20 facil's reactor becomes a quantum data
41:22 center. Eg using Google's sycamore
41:24 processor 2019 scaled up. Quantum bits
41:28 cubits simulate infinite realities for
41:31 mind uploads but risk entang
41:34 entanglement errors causing migraines.
41:37 Quantum decoherence affecting neural
41:39 patterns.
41:41 Instead of a meltdown, have a quantum
41:43 cascade failure erasing data instead of
41:46 radiation. Love it.
41:49 That's [ __ ] good because you could
41:52 see a character going uh like like uh uh
41:56 we're having entanglement errors, sir.
41:57 Well, fix the errors. It's not that
41:59 simple. Quantum decoherence is affecting
42:01 the neural patterns.
42:05 We've got a quantum cascade failure,
42:07 sir. Fix it. Fix it now.
42:10 All right. The facility's reactor
42:12 becomes a quantum data. Okay, that we
42:14 did. Okay. Biomputer interfaces. Tony's
42:16 migraines stem from a faulty BCI implant
42:20 hidden from him, allowing remote access.
42:24 Update to include optogenetics,
42:27 lightbased neuron control, advanced
42:29 2020s, or non-invasive ultrasound BCIs.
42:34 Assassins use BCI for precision
42:36 targeting like thought controlled
42:37 drones. Oh yeah, that's pretty good,
42:40 too. Yeah. See, like Yeah. So, so Virtue
42:44 Tech, I think is the name of the
42:45 company. They could be a defense
42:48 contractor and they're doing like mind
42:50 control drones that go off and
42:51 assassinate people. So, this is their
42:53 advanced program using quantum neural
42:57 BCI quantum neural injection. All right,
43:01 good cultural references. Replace
43:03 landlines with smartphones and AI
43:05 assistance.
43:07 Gun buyback involves smart guns. News
43:10 via Tik Tok streams and AI factchecking.
43:14 Yeah, it's good. All right. Self-driving
43:17 cars. Good. Scientific fixes. Ground
43:20 mind transfer in connection mapping.
43:24 Eg. Via open worm project or human brain
43:28 project from 2023. Waste disposal equals
43:31 quantum encrypted data wipes. Escape via
43:34 hacking. Okay, good. Logical continuity
43:37 errors. Gun continuity. Tony's knowledge
43:41 character relationship. Anna gives Tony
43:43 a photo related to Lenny's murder, but
43:46 is later revealed as connected to
43:48 virtue, her father. Inconsistency. She
43:51 knows Lenny, but doesn't explain how.
43:53 All right. Assassination attempt.
43:55 Meltdown timeline. All right.
43:59 What the screenplay gets right. Sci-fi
44:02 and storytelling perspectives.
44:05 Preient themes. Explores mind control.
44:07 secret government tech and identity
44:09 duplication relevant to 2025 concerns
44:12 like AI ethics, open AI safety debates,
44:16 BCI privacy, neural link trials 2024,
44:19 and quantum risks, eg hacking quantum
44:22 keys, the VR reality bleed anticipates
44:26 metaverse issues, eg the addictions in
44:30 Ready Player One. Plausible tech core
44:33 molecule docking as a metaphor for mind
44:37 integration aligns with real
44:39 biioinformatics.
44:40 Nuclear secrecy echoes blackbudget
44:43 projects blah blah blah. It gets right
44:45 that in advanced things storytelling,
44:47 pacing and structure. Fast montage
44:49 scenes, character arcs, thematic death
44:52 depth.
44:54 All right. Um
44:57 there you go.
45:00 So let's see. Let's let's ask it to do
45:02 one more thing. Um we'll say um
45:07 pick a scene
45:11 you think needs
45:14 the most updating
45:18 and
45:21 rewrite it
45:24 to
45:26 maintain
45:28 compelling dialogue.
45:34 Um
45:38 compelling dialogue. um
45:42 your
45:44 scientific
45:46 and
45:48 cultural updates
45:52 and
45:54 um
45:57 other improvements
46:03 implied
46:07 in other scenes. means you will also
46:12 update.
46:14 Okay, let's see if this thing can
46:16 [ __ ] write.
46:18 James Dutder, this is blowing my mind.
46:20 Yeah, this is this is pretty good so
46:21 far. Like, normally when I do stuff like
46:25 this, it's okay,
46:28 right?
46:30 Um, it'll be curious to see if this
46:32 thing can write. Selected scenes for
46:34 update. The assass the assassin's
46:36 punishment. Pages 14 to 16. The scene
46:40 needs the most updating because it
46:42 relies on outdated and scientifically
46:44 implausible implausible nuclear waste
46:48 disposal. Instant melting is impossible.
46:52 So listen, I don't have a degree in
46:53 nuclear physics. I have a degree in
46:56 acting.
46:59 You dumb dumb. I didn't have AI when I
47:01 wrote this thing. I know I shouldn't be
47:04 getting defensive, but you know, it also
47:07 misses opportunities to integrate modern
47:09 breakthroughs like quantum computing,
47:10 decoherence, and biomputer interfaces.
47:14 Culturally, it feels like a '9s sci-fi
47:18 horror because it was written in the
47:20 90s,
47:23 the flyesque body horror. Updating it to
47:26 quantum AI uh facility emphasizes
47:30 current fears. Okay. maintaining powers,
47:33 chilling charisma while implying broad.
47:35 Okay, so rewritten scene.
47:41 It wrote a [ __ ] scene here.
47:44 All right. Interior. Virtue tech
47:47 building. Quantum core room at night.
47:49 Camera pans. A sleek humming chamber
47:52 dominated by a circular array of
47:54 cryogenic quantum processors. Glowing
47:57 blue glowing be glowing blue cubits
48:01 suspended in vacuum-sealed pods.
48:04 Two adjacent
48:06 cylindrical chambers made of reinforced
48:08 graphine glass flank the core each with
48:12 biometric airlocks synced to neural
48:15 implants. This is a really good
48:17 description.
48:19 And we are gonna go right now straight
48:22 to midjourney
48:24 and we're gonna see if we can make that
48:26 image. Well, now I know we can make the
48:28 image. We're gonna see if we can get it
48:29 to be good. So, I'm going to turn on my
48:32 personalization.
48:34 We're going to do stylization to 500.
48:40 Um, and we'll do dash dash style.
48:47 No, we don't want raw style. We want it
48:49 personalized. We'll do Do we want to do
48:51 anything?
48:53 No.
48:56 We'll do, you know what we'll do? We'll
48:57 we'll do We'll do one like this and
49:00 we'll do one where we increase the
49:01 weirdness. All right. So So
49:06 that's being generated now. And then
49:09 we'll do it again. We'll do style 500.
49:12 And then we'll do weirdness to 500,500.
49:18 All right.
49:20 Oh, this is so cool.
49:25 Look, people
49:27 look here.
49:32 So now this room looks like the modern
49:39 That's really cool.
49:42 The quantum the quantum quantum [ __ ]
49:47 That's really cool. It's really dang
49:50 cool.
49:56 Oh. Oh, this was with weirdness turned
49:59 high. Yeah, these are good. Like you
50:01 could cut between all these.
50:04 In fact, let's
50:09 let's see. Let me do let's turn
50:12 weirdness down to two 700 600. We'll do
50:16 weird weirdness down to 600.
50:19 We're going to do stylization to 600
50:23 and we're going to go bang.
50:26 Then we're going to go d-r
50:31 um four. So we're gonna we're gonna make
50:34 16 of these images. Okay. All right.
50:37 Let's go let's go back to Grock.
50:42 Okay, so interior virtue tech building
50:45 quantum core room night. Sleek humming
50:47 chamber dominated by a circular array of
50:49 quantum cryogenic quantum processors.
50:52 That's really cool. Okay. The latest
50:54 quantum supremacy chips from outfits
50:56 like IBM or Google. Oh, Powers. So,
51:00 Powers offscreen. So, Powers is like the
51:02 I think he's the guy from Virtue. The
51:05 latest quantum supremacy chips from
51:07 outfix like IBM or Google child's play.
51:10 Officially they're breaking codes or
51:12 simulating molecules. But this powers
51:15 coldly intense in his late 40s gestures
51:18 grandly. The assassin shifts nervously
51:21 beside him. Still in rally gear. Becker
51:24 stands to guard the door.
51:26 This setup could have funded a fleet of
51:28 hypersonic drones all off books siphoned
51:32 from IRS slush funds and crypto mining
51:34 ops. That's good. Impressive? No. No
51:37 messy fision reactors. Just pure quantum
51:40 entanglement powering simulation
51:42 simulations that rewrite reality. Public
51:45 would riot if they knew we're uploading
51:47 minds forging perfect agents in in a dig
51:51 in digital metaverses. The assassin's
51:54 eyes. The assassin eyes the cubid array
51:57 uneasy.
51:59 Yes, sir. What exactly do we do here?
52:02 Whatever we damn well please. Which is
52:04 why a medddler like Joseph Karns pushing
52:08 for AI regulations and quantum export
52:10 bans is an existential threat. Now check
52:13 this out. Um P steps into a cylindrical
52:16 chamber, sealing the neural sync airlock
52:20 with a palm scan. He beckons the
52:22 assassin into the adjacent one. Through
52:25 the intercom, P's P's voice echoes with
52:29 weird phase shifted reverb like a
52:32 glitched AI hallucination. This is
52:34 really good. This This is feeling
52:36 familiar like I wrote it but with better
52:39 references. So, this is good. Powers
52:42 continued. I love how entanglement warps
52:45 the sound in here. Don't you? Feels like
52:47 your thoughts are decohering. The
52:50 assassin assassin nods wearily. The
52:53 chamber's interface lighting up with
52:55 holographic neural maps. These pods
52:58 twice a year they purge entangled data.
53:00 Failed mind uploads. Corrupt corrupted
53:03 Sims like wiping a neural network with
53:05 cubits collapsing entire consciousness.
53:08 The assassin's eyes widen in concern.
53:11 Powers laughing. Relax. They get qu They
53:14 get a quantum scrub down. I just wanted
53:16 to hear the echo.
53:19 Powers exits the chamber effortlessly.
53:21 The assassin's door remains sealed. A
53:23 red holographic lock pulsing. Powers
53:26 speaks in a wristmounted BCI controller.
53:29 Brain computer interface controller.
53:31 Step forward. The neural sync should
53:33 disengage. Uh, the assassin approaches.
53:36 Uh, but the doors hold. Powers fains
53:39 confusion. Checking for a holographic
53:41 display on his smart watch. Huh? Would
53:44 you look at that time? Rotten luck.
53:47 Purge. Purge cycle starts now. Six
53:50 months on the dot. The assassin pounds
53:52 the glass. So basically, he's about to
53:55 wipe the assassin's brain. Um, amber
53:57 warning. Hologram flash. Uh, quantum
54:00 decoherence initiated. A low hum builds.
54:03 Cubits in the C core flickering
54:05 erratically.
54:06 Um, in a setup like this, entanglement
54:09 failure hits fast. Your implant
54:11 overloads. Neurons firing in infinite
54:14 loops. memories fracturing and shat like
54:16 a shattered LLM. You don't melt, you
54:19 just cease. Horrific. But over 20 But
54:23 but over in 20 seconds, no body to bury,
54:26 no digital ash. Um tears stream down the
54:29 assassin's face as he begs silently
54:32 through the glass. No sound escapes the
54:34 sealed pod. Powers. This is the price of
54:37 failure. Good night, mister. Funny. I
54:41 can't recall your name. Must be the
54:43 entanglement.
54:44 From overhead vents, a cascade of
54:46 ionized particles flood the chamber,
54:48 simulating a quantum eraser wave. The
54:51 assassin clutches his head, convulsing
54:53 as neural feedback loops erase his mind.
54:56 Powers presses his face to the glass,
54:59 screaming mockingly, mimi mimicking the
55:03 silent agony while the assassin's form
55:05 slumps, eyes vacant, body intact, but
55:09 consciousness obliterated.
55:13 Um,
55:14 pretty [ __ ] good.
55:18 It's pretty [ __ ] good.
55:23 Apparently, you need to get one. Um,
55:28 what did I miss? Um, I uploaded so we're
55:31 testing we're testing Grock 4
55:35 and I wrote an article back in January
55:37 on LinkedIn about how to use reasoning
55:40 models for the humanities, right? So for
55:43 for non STEM [ __ ] Um, and the first the
55:48 very first prompt was analyze a
55:50 screenplay. So, I I uploaded a
55:52 screenplay that I wrote and had it
55:54 analyze it for scientific in uh like
55:58 outofdeness basically um and update it
56:01 and then I said rewrite a scene for me
56:03 and it rewrote a scene and it it did it
56:05 pretty [ __ ] good. Uh 11 Labs Reader.
56:09 Yeah, this would be good to put into 11
56:11 Labs Reader. Be pretty good. Um
56:17 that's pretty impressive. And then we
56:18 went over to midjourney.
56:22 And so like here, so this is this is the
56:26 hall where powers would have been. All
56:28 right. So let's see.
56:31 Um, let's let's go grab
56:37 let's go grab we'll grab this whole
56:39 scene
56:46 and then I'm going to go to chat GPT
56:51 and I'm going to say um I need you to
56:57 create
56:59 Five mid journey prompts
57:06 that include
57:09 the characters
57:13 that you feel are the most
57:18 visually
57:23 interesting.
57:25 And then I'm going to paste in my
57:27 script. And then I'm going to go um
57:32 be sure
57:35 to include
57:38 detailed
57:40 character
57:43 descriptions
57:45 in each prompt. as if
57:52 the prompt is not aware
57:56 of a character
58:00 description in the prev.
58:06 Each prompt needs to stand alone and be
58:12 understood
58:15 by an image
58:21 generator.
58:24 All right.
58:26 So, let's see how chat GBT does with
58:28 giving us some prompts. So, so what
58:29 we're going to do now is we're going to
58:30 go over
58:35 Oh, you know what? Let's see. Um,
58:39 let's go grab this prompt.
58:43 Copy.
58:45 Go chat GBT.
58:48 Okay. Um,
58:51 these
58:52 are good.
58:55 Um,
58:57 and I already
59:01 generated
59:03 images based on
59:07 images of
59:09 the facility
59:13 based
59:19 on this prompt.
59:25 Update the prompts you wrote
59:34 to include
59:38 any references
59:42 here
59:45 that
59:47 might have been missed.
59:56 Okay.
59:58 You can use image to video on VO3 now.
1:00:01 Much easier for consistent characters.
1:00:03 Yep. Definitely going to do that. Um,
1:00:06 okay. Powers in the quantum course.
1:00:09 Cinematic wide shot of a sleek humming
1:00:11 chamber dominated by a circular array.
1:00:13 Glowing blood two cylindrical.
1:00:16 Okay. Powers. A sharply dressed man in
1:00:18 his late 40s with piercing eyes, salt
1:00:20 and pepper hair, sllicked back, wearing
1:00:23 bespoke blah blah blah. Okay, good. The
1:00:25 assassin's dread. The assassin. A gaunt
1:00:29 someone.
1:00:32 A gaunt woman in her 30s with a buzzcut
1:00:34 and haunted eyes. Great.
1:00:39 Becker, stocky man in his 30s. Fine.
1:00:54 power stands. Okay. All right. So, these
1:00:56 look decent. Great.
1:00:58 Okay. So, let's
1:01:04 grab the prompt.
1:01:06 Go to midjourney.
1:01:09 Pop that bad boy in there.
1:01:14 We'll just do the We'll do the five
1:01:15 scenes.
1:01:17 So, we got that one.
1:01:20 We're just going to copy them. They seem
1:01:21 to be decent. So, I think we'll be okay.
1:01:23 Copy.
1:01:25 And we could use omni reference. If we
1:01:27 wanted if we wanted the characters
1:01:30 really consistent, we'd use omni
1:01:32 reference or we'd use some other tool.
1:01:37 So, I'm popping back and forth between
1:01:40 chat GPT and MidJourney. Right now,
1:01:46 we're just grabbing our five most
1:01:47 interesting scenes from that little
1:01:49 scene that it rewrote.
1:02:11 All right.
1:02:13 So now
1:02:21 [Laughter]
1:02:23 that was a little it's a little
1:02:25 dramatic, but you know
1:02:31 they look too sci-fi.
1:02:54 LinkedIn comment. I had my Grock book
1:02:57 chapters turned into five comic book
1:02:59 pages.
1:03:01 Oh, that's interesting.
1:03:03 Let's go back to Grock and try that
1:03:05 because it can do images, right? Oh, now
1:03:06 I got deep research back. Oh, no,
1:03:10 that's not. So, okay. Wait. So, turn
1:03:15 this scene
1:03:18 into a series
1:03:21 of panels
1:03:24 for a
1:03:26 modern
1:03:28 graphic novel
1:03:30 with dialogue.
1:03:35 Create as many panels as you need
1:03:43 while it does that. Tik Tok question. Is
1:03:45 China outpacing the US and AI? Um,
1:03:54 that's very hard to know.
1:03:57 I think that um
1:04:04 one of the things that China's doing,
1:04:07 which is a a really interesting
1:04:10 strategic political move,
1:04:14 is that it seems clear to me that
1:04:16 they've got either access to
1:04:22 private
1:04:24 documents from like Open AI from from
1:04:26 the major model companies because
1:04:30 they're coming out with open-source
1:04:32 models that are kind of as good or
1:04:34 better than chat GPT than OpenAI's
1:04:37 state-of-the-art stuff like within weeks
1:04:40 or a month, right? So, so they seem to
1:04:43 be they seem to be developing things in
1:04:46 parallel. The interfaces are identical.
1:04:50 How the things work are very very
1:04:51 similar. They have different
1:04:52 personalities.
1:04:54 Um, but what they're doing, which I
1:04:56 think is fascinating, is they're open
1:04:58 sourcing everything.
1:05:00 Um, and by open sourcing everything,
1:05:02 they uh
1:05:06 they're making it harder for for
1:05:09 um,
1:05:11 you know, OpenAI to maintain a
1:05:13 commercial edge. That said, I don't
1:05:17 think a lot of big companies are
1:05:19 trusting
1:05:21 all those models necessarily. So I I
1:05:23 don't know. Um I think the one place
1:05:27 that China
1:05:29 was gaining some ground um was on the
1:05:33 video generation stuff, but I think with
1:05:35 MidJourney and VO3 from Google, I feel
1:05:38 like we're ahead. I think what Grock
1:05:41 just released is way ahead of what
1:05:43 anything China's done. So I don't know.
1:05:46 Uh Mr. S1, what do you guys run? An AI
1:05:51 lab or no. So, so no. So, couple of
1:05:55 things.
1:05:57 This channel, AI learning lab, think of
1:06:00 this channel as what we're talking about
1:06:02 here is not building the models, but
1:06:07 using the commercial models that were
1:06:10 really kicked off with chat GPT. So,
1:06:14 November 30th, 2022 is what is
1:06:18 essentially when this channel was born.
1:06:20 um and when the AI salon was born. So,
1:06:22 if you go to the AI salon, go to the
1:06:24 salon.ai and click on join our
1:06:26 community. That'll take you into um a
1:06:29 community. So, it's a community of AI
1:06:32 optimists and AI adventurers, people
1:06:34 that are using AI in in an applied way,
1:06:37 in a creative way. Um, you know, a salon
1:06:40 is is not the place you get your hair
1:06:42 done, but it's based on the, you know,
1:06:44 19th century salon of people getting
1:06:47 together and having intellectual banter
1:06:50 about what's happening, what's possible.
1:06:53 Um, so, so think of it as people that
1:06:56 are using AI rather than creating AI.
1:06:59 Um,
1:07:01 yeah. So, okay, let's see. So, I created
1:07:05 two images here.
1:07:07 Uh, make more. Let's see if it does
1:07:11 that. Correction from link. Sorry, I did
1:07:13 this in chat GBT. The book was called
1:07:15 Grock. Ah, it's not the tool I used to
1:07:18 do it. Chat GPT does it perfectly. Just
1:07:22 feed the whole chapter in and make a
1:07:25 one picture at a time, a panel of six
1:07:28 vertical 8 and a half by 11 pages. It'll
1:07:30 break down the dialogue and character.
1:07:32 Okay, cool. So, yeah, this is not doing
1:07:34 it well.
1:07:37 I mean, listen, not for nothing, but
1:07:39 Grock is it it at least makes images.
1:07:42 It's not putting dialogue there. Okay,
1:07:44 let's let's go back to um here's Mid
1:07:48 Journey. Oh, this these are cool.
1:07:56 These are all stylistically very very
1:07:58 different, aren't they? So, we'd have to
1:08:01 do some work here to get these things to
1:08:02 be stylistically.
1:08:05 This is one of the reasons that these
1:08:09 are all different. Well, there's a
1:08:11 couple of reasons. I have the I have
1:08:12 stylization turned high and I have
1:08:14 weirdness up. So, it's going to the
1:08:16 weirdness is going to generate um
1:08:18 variations.
1:08:20 Um
1:08:23 but I also don't have a consistent
1:08:27 style description for each of these
1:08:29 prompts.
1:08:31 So, if we were to do this right, what I
1:08:34 would likely do is I would go into chat
1:08:35 GPT and I would have chat GPT come up
1:08:38 with a prompting schema
1:08:40 like a like a JSON
1:08:43 um tagged schema that says, you know,
1:08:47 here's how we're going to d uh describe
1:08:48 the cinematography, the set design, the
1:08:51 character designs, all that sort of
1:08:53 stuff. So we come up with a whole
1:08:55 framework so that every prompt that we
1:08:58 put into midjourney has got this whole
1:09:00 package that that gives you consistency
1:09:03 of of
1:09:05 cinematography and and and you know
1:09:07 camera work and and things like that
1:09:10 across
1:09:11 all the all the prompts. Um so that's
1:09:14 not in there right now. But that said,
1:09:17 we can go in here and we can grab we can
1:09:19 just start animating some of these
1:09:21 things that we think look decent. All
1:09:23 right. So, let's even though they might
1:09:25 not be consistent, let's go grab some of
1:09:27 these. That doesn't look good. That's
1:09:29 not bad.
1:09:31 So, let's We'll just do low motion on
1:09:33 that one.
1:09:35 We'll do low motion on that one. We'll
1:09:38 do low motion on that one.
1:09:41 That looks pretty good. We'll do high
1:09:43 motion on that one. I don't know what
1:09:44 that's going to do. We'll do high motion
1:09:46 on that one. So, these would be sort of
1:09:48 cuts.
1:09:50 That's pretty cool looking. Cryogenic
1:09:53 [ __ ]
1:09:55 for quantum computers.
1:10:00 That looks pretty decent.
1:10:03 Do motion on that.
1:10:06 That's pretty good there. We'll do low
1:10:08 motion on that.
1:10:11 We'll do high motion on that. Maybe
1:10:12 we'll get some cool camera movement. And
1:10:14 we could be prompting these, but I'm
1:10:16 just being lazy.
1:10:18 All right. Which one of these does not
1:10:20 look comical? They all look pretty
1:10:21 [ __ ] comical.
1:10:25 That one's pretty good. So, we'll do
1:10:28 that one. Let's copy this image. And now
1:10:32 we're going to go to flow.
1:10:36 And we're gonna
1:10:38 Oh, wait. I think I have to save the
1:10:40 image, don't I? So, I'll download this
1:10:42 image. Then we'll go to flow.
1:10:46 We'll do new project.
1:10:48 So flow, if you don't know it, this is
1:10:50 Google's
1:10:52 uh video creation thingy.
1:10:56 And so we're going to do frames to
1:10:58 video. So we're going to go grab that
1:10:59 frame we just downloaded.
1:11:05 All right, there it is.
1:11:11 And then we'll say
1:11:14 um
1:11:16 security guard
1:11:18 with a gruff voice
1:11:22 says
1:11:27 I'd
1:11:29 suggest getting in the pod.
1:11:35 What's the worst
1:11:39 that can happen?
1:11:42 He erases
1:11:46 your brain.
1:11:48 [Laughter]
1:11:50 Which is exactly what happens, right?
1:11:52 All right. So, there we're going to make
1:11:54 that scene.
1:11:58 If you're if you've joined recently,
1:12:00 what we're doing here is
1:12:04 uh one of the nicknames for this channel
1:12:06 is Chat Add.
1:12:08 And what I wanted to do was start to
1:12:10 play around with the Gro 4 and see if
1:12:12 it's good at [ __ ] Um and I wrote an
1:12:15 article back in January about how to use
1:12:19 reasoning models with humanity's
1:12:22 challenges. like what are things you
1:12:24 could do in non-scientific
1:12:27 challenges because I'm not a scientist.
1:12:29 I'm a humanities person. I have a fine
1:12:31 arts degree in in acting and uh I've
1:12:34 written seven screenplays. So I know
1:12:37 writing, I know storytelling, I know
1:12:39 image stuff, things like that.
1:12:41 So I upload a screenplay I wrote in the
1:12:43 90s and I I had it analyze it and update
1:12:47 it from a scientific perspective and it
1:12:49 it did pretty [ __ ] good. And then I
1:12:51 actually had it rewrite a scene that was
1:12:54 the most out ofd and the most [ __ ] up.
1:12:56 And what it wrote was kind of good. And
1:12:59 now I'm taking that scene and we're
1:13:01 going into midjourney and and different
1:13:03 tools to bring the scene to life. Um,
1:13:07 okay.
1:13:08 Uh, let's go back to here. So, so that
1:13:12 we're going to make that dude talk.
1:13:15 That's actually a good image of
1:13:19 powers.
1:13:22 So, we're going to download that. Let's
1:13:24 go grab some dialogue.
1:13:31 All right. So, powers.
1:13:48 All right. In a setup like this,
1:13:51 entanglement failure hits fast. Your
1:13:53 implant overloads, neuron neurons firing
1:13:56 in infinite loops, memories fracturing,
1:13:59 and like a shattered LLM. I think
1:14:01 that'll fit in 8 seconds. So, we'll grab
1:14:03 that piece of dialogue. We're going to
1:14:05 go to this thing. We can watch that in a
1:14:08 second. Let's put
1:14:13 image in there.
1:14:17 Uh, we'll say maniacal
1:14:23 tech,
1:14:25 CEO,
1:14:27 visionary,
1:14:33 evildoer.
1:14:37 Um,
1:14:43 crackles.
1:14:46 with a sinister
1:14:52 voice
1:14:55 and says,
1:15:03 "How do you spell maniacal?" Oh, yeah.
1:15:06 Like that. You You have AI fix it for
1:15:09 you. That's what you do in a setup like
1:15:12 this. Entanglement failure hits fast.
1:15:14 Your implant overloads neurons. Neurons
1:15:17 firing in infinite loops. Memories
1:15:19 fracturing like a shattered LLM. Bang.
1:15:22 Let's go make some acting.
1:15:25 Those of you with acting degrees, you're
1:15:27 no longer needed. We're good here. Carry
1:15:29 on. All right. Could you curate your 20
1:15:32 favorites and define an SREF then
1:15:35 character sheets to define the
1:15:37 characters? Yeah, there's there's all
1:15:39 sorts of ways to do this, Rick. I think
1:15:41 so. I've been thinking about this a lot
1:15:43 with Sydney. I think the real trick is
1:15:47 that you would basically
1:15:49 um design all of your core settings
1:15:57 and your
1:16:02 I need a shot caller. Champ. Champ.
1:16:05 Champ. Champ. Hang on. Oh, please.
1:16:35 All right. There was an evil man with an
1:16:37 iPhone.
1:16:40 do need a shot caller, but not for
1:16:41 champ. Okay. Um, what was I saying? Um,
1:16:46 oh, Rick McCauley. So, I think what you
1:16:48 do is you do you would do settings, you
1:16:51 do cinema cinem uh like cinematic style,
1:16:55 and then I think you would do character
1:16:56 sheets independent, right? So, I would
1:16:58 do I would do a um like a portrait of
1:17:01 the person, and then I would do like
1:17:03 like uh costume character studies,
1:17:06 things like that. And then I would use
1:17:08 all of those as reference images
1:17:11 in something like runway where you can
1:17:13 put a scene image, a style image, and
1:17:16 character references or you know
1:17:18 costumes and characters and just have it
1:17:20 assemble them all there. Um, but there's
1:17:24 these kind of workflows like LTX Studio
1:17:27 right now is trying to automate all
1:17:29 these workflows. I think these kind of
1:17:32 workflows are going to be incredibly
1:17:35 valuable. So, anyone that that can start
1:17:38 to crack these and and really get their
1:17:40 head around it is going to be is going
1:17:41 to be super good. They're going to be
1:17:43 super good. They're going to be in
1:17:45 demand because cuz filmmakers like
1:17:48 directors and storytellers are going to
1:17:50 want to make films and they're going to
1:17:52 be able to, but they're they're going to
1:17:54 want character consistency and things
1:17:56 like that.
1:17:58 tools will come over time, but how you
1:18:00 get to those tools is you have people
1:18:02 that understand the current process,
1:18:06 learn enough about AI to understand
1:18:08 which tools to employ when. Like all of
1:18:10 this add jumping that we're doing where
1:18:12 we're jumping between these different
1:18:13 tools,
1:18:16 those are skills that
1:18:20 not everyone's going to have, right?
1:18:22 people that understand storytelling and
1:18:24 technology and AI and workflow and cons,
1:18:29 you know, character consistency and
1:18:30 logical consistency.
1:18:34 There's the like the the the world set
1:18:37 of people that can combine all that
1:18:39 stuff gets very small very fast. So
1:18:42 those people are going to be in demand.
1:18:44 And think about this for every different
1:18:46 kind of industry, right? We're just
1:18:47 talking about storytelling right here.
1:18:49 All right, let's see if these uh things
1:18:52 are any good. These videos,
1:18:57 one scene I just one set of dialogue I
1:19:00 just wrote
1:19:02 and then one came out of the thing that
1:19:04 V that uh Grock 4 wrote
1:19:11 black bar.
1:19:13 All right, here we here we go people.
1:19:18 I'd suggest getting in the pod.
1:19:21 >> What's the worst that can happen? He
1:19:23 erases your brain.
1:19:26 >> I'd suggest getting in the pod.
1:19:29 >> What's the worst that can happen? He
1:19:31 erases your brain.
1:19:34 I'd suggest
1:19:35 >> pretty good. Pretty [ __ ] good.
1:19:37 >> Getting in the pod.
1:19:39 What's the worst that can happen? He
1:19:41 erases your brain.
1:19:43 >> All right. I like it. That works. All
1:19:46 right, let's see if we get a maniacal
1:19:48 CEO visionary evildoer.
1:19:53 >> In a setup like this, entanglement
1:19:55 failure hits fast. Your implant
1:19:57 overloads neurons firing in infinite
1:19:59 loops. Memories fracturing like a
1:20:00 shattered LLM. In a setup like this,
1:20:03 entanglement failure hits fast. Your
1:20:05 implant overloads neurons firing in
1:20:07 infinite loops. Memories fracturing like
1:20:08 a shattered LLM. In a setup like this,
1:20:11 entanglement fail. It was a lot of
1:20:13 dialogue for eight seconds, but it got
1:20:15 it in. It's not horrible.
1:20:18 It's not [ __ ] horrible, is it?
1:20:20 >> Failure hits fast. Your implant
1:20:22 overloads neuron.
1:20:23 >> Are you using a paid version of Grock?
1:20:25 No, I'm not. I'm just using a regular
1:20:27 old shitty old I don't know how long how
1:20:30 much longer I'm going to have access to
1:20:31 it
1:20:33 because if I go to grock.com, I don't
1:20:35 have access to Grock for, but I'm in X
1:20:39 and I I pay for X. I'm like a I don't
1:20:42 know $17 a month subscriber.
1:20:46 That's a YouTube sponsor ad.
1:20:49 And now let's go to Mid Journey and look
1:20:51 at all the animations that we made and
1:20:53 see if any of these feel like things
1:20:56 that we could put into a film.
1:21:01 That doesn't, but it's not bad.
1:21:06 It's actually kind of cool.
1:21:09 That's better. That looks more like
1:21:12 except something just materialized out
1:21:14 of nothing there.
1:21:24 That's pretty cool.
1:21:28 >> Yeah. No, this is You all are awful
1:21:32 people. I've probably shown six
1:21:35 different things none of you could see
1:21:37 and you blame me for it. Thank you,
1:21:39 producer Brandon.
1:21:42 All right, what did I show? What did I
1:21:44 talk about? Uh, what was I doing? I went
1:21:46 to Someone asked about Grock. So, I was
1:21:49 just showing Grock. Here, here's Grock.
1:21:52 Um,
1:21:55 this you all saw that was the stuff in
1:21:58 V3 and then I went back to Midjourney.
1:22:01 Okay, so let me go to
1:22:04 out here. So,
1:22:06 here's the animations we made in
1:22:09 Midjourney. New YouTube spotlight
1:22:11 softbot creation. What is the best tool
1:22:13 for text to image? Midjourney or
1:22:16 Leonardo? Um,
1:22:21 it depends on your goal. If your goal is
1:22:24 aesthetic beauty, midjourney is it?
1:22:29 Um, Leonardo is
1:22:34 is doing really well. Unless they've
1:22:36 changed it, they're still based on
1:22:37 stable diffusion and they do a lot with
1:22:40 custom loras and things like that. So,
1:22:44 Leonardo's pretty good at getting a high
1:22:47 aesthetic, but I would say just as a
1:22:50 consistent
1:22:53 high aesthetic, midjourney is is still
1:22:55 the goat.
1:22:57 Um, I don't know how well Leonardo does
1:23:00 with text generation, like text panels.
1:23:04 If you need that, chat GPT is better
1:23:07 than anything right now. So, I would say
1:23:09 chat GPT, Leonardo, or MidJourney are
1:23:12 all good. Um, if you just want [ __ ]
1:23:14 that's kind of pretty and and
1:23:16 aesthetically beautiful and it now has
1:23:19 animation, but I think uh Leonardo also
1:23:21 has animation now. Um, but MidJourneys
1:23:24 is is quite good. So, these
1:23:27 animations that we're looking at here,
1:23:29 none of these are really usable. They're
1:23:30 just weird. That's not really usable.
1:23:32 That could be usable. That's not bad for
1:23:35 a bunch of quantum computers. This is
1:23:37 better where they're staying in place
1:23:39 and just rotating. It's kind of cool.
1:23:43 Oops. Let me zoom in again.
1:23:48 That's pretty good.
1:23:51 Yeah, that's a good shot there.
1:23:54 So, That's kind of cool. That kind of
1:23:57 works.
1:24:00 This definitely works,
1:24:06 right? These are all establishing shots
1:24:07 of the tech.
1:24:12 You cut away to these, you know,
1:24:14 closeups of the the cryogenic
1:24:18 quantum cooling units,
1:24:24 right? You could hear P's voice over.
1:24:28 Yeah, this is state-of-the-art here. Let
1:24:31 me
1:24:33 let me
1:24:35 pop something out here. Um,
1:24:59 okay.
1:25:01 So, I'm going to take this. I'm gonna
1:25:02 throw it
1:25:05 in here.
1:25:17 Okay. So, here's what I'm going to do.
1:25:19 Oh, wait. I can't cover that up because
1:25:22 that's covering up all your abil ability
1:25:25 to see anything
1:25:28 there. That's good. All right. Let's
1:25:31 see.
1:25:33 [Music]
1:25:37 Okay. So, we go back here.
1:25:40 So, you can see as as these things
1:25:43 are on screen,
1:25:46 you could hear power is saying
1:25:49 the latest quantum supremacy chips from
1:25:51 outfits like IBM or Google. Child's
1:25:54 play. Child's play.
1:25:58 officially
1:26:00 they're breaking codes or simulating
1:26:02 molecules, but this
1:26:05 um
1:26:07 this setup could have funded a fleet of
1:26:09 hypersonic drones all offbook siphoned
1:26:12 from IRS slush funds and crypto mining
1:26:14 ops. Impressive? No. No me messy fision
1:26:18 reactors, just pure quantum entanglement
1:26:21 powering simulations that rewrite
1:26:23 reality, right? as you're cutting to,
1:26:26 you know, shots like this
1:26:29 while you have the voice over. That
1:26:31 could work. That's a [ __ ] movie.
1:26:35 It's [ __ ] movie, people.
1:26:38 See what I'm saying?
1:26:41 Oh yeah.
1:26:43 Oh yeah. Tell them what they've won,
1:26:46 Bob. Oh, you've won some quantum
1:26:48 entanglement.
1:26:50 Reality distortion field, my ass. We're
1:26:52 rewriting reality,
1:26:57 right? And then at some point, the dude
1:26:59 says, "What's the worst that could
1:27:00 happen?" They erase your brain. And then
1:27:03 they proceed to erase her brain.
1:27:08 All right, there you go. All right,
1:27:09 let's try Rick McCauy's thing with Chat
1:27:11 GBT. Okay, so so now we've we've had it
1:27:15 right that. So, I'm going to say um
1:27:18 what would be the best
1:27:22 graphic novel style
1:27:26 to bring this
1:27:31 scene to life?
1:27:38 [Music]
1:27:47 [Music]
1:27:52 Billesque Tech Noir
1:27:55 Jar will Williams surreal precision
1:28:02 um
1:28:04 bonus hybrid concept. Let's
1:28:08 go hybrid
1:28:12 and
1:28:13 generate
1:28:16 a
1:28:19 multi-
1:28:21 multi-
1:28:25 panel
1:28:30 vertical page
1:28:33 that
1:28:36 captures the early
1:28:40 part of the scene.
1:28:44 Don't skip
1:28:47 anything.
1:28:50 I want to make a full
1:28:55 graphic novel of this
1:29:00 scene.
1:29:02 So, just do the uh just do whatever
1:29:09 can fit on page one. And then I'm going
1:29:14 to flip over to 03 reasoning engine
1:29:18 because I want it to think about what it
1:29:22 can what it's going to put in the panel.
1:29:26 All right.
1:29:28 So, now we're making graphic novels.
1:29:32 So, we've got
1:29:34 Gro 4, which is smarter than any PhD
1:29:37 student in any discipline, which I
1:29:40 assume includes humanities,
1:29:43 and it analyzed our screenplay, and it
1:29:45 did it pretty [ __ ] good. And then it
1:29:46 rewrote a scene for us, and it did
1:29:48 pretty [ __ ] good.
1:29:51 Um,
1:29:54 and now we're using Midjourney to make
1:29:56 some images. And then we used V3 to make
1:29:59 it act. And then we used MidJourney to
1:30:01 animate the images that we made. And now
1:30:04 we're seeing if we can get this thing to
1:30:06 create.
1:30:09 Yeah, the reasoning engine's
1:30:12 defining the panels.
1:30:17 Um,
1:30:19 I want you to now make this image of
1:30:25 page one.
1:30:28 Uh, uh, let's see if it does it.
1:30:37 Here we go.
1:30:39 We are off. Would this make games too?
1:30:42 Yes. I mean, sure. Um, in fact, we can
1:30:45 do that next. We can do that next.
1:30:53 Exits.
1:30:56 Oh, exists.ai
1:30:59 Vibe game platform launches tomorrow.
1:31:01 That's cool.
1:31:03 That's cool. Well, so wait, so we'll go
1:31:05 to let's go back to Grock now. Let's say
1:31:09 um
1:31:11 I want you
1:31:16 to
1:31:18 come up with a
1:31:21 webbased
1:31:22 video game
1:31:26 that
1:31:29 is inspired by the screenplay.
1:31:35 um describe it in detail.
1:31:41 It should have
1:31:45 great graphics,
1:31:49 trigger
1:31:50 twitching
1:31:52 gameplay
1:31:55 and make someone
1:32:00 familiar with the movie happy to play
1:32:04 it.
1:32:09 All right.
1:32:11 Open beta on exist.ai open beta on their
1:32:14 Discord. Holy [ __ ] Let's go play. Is
1:32:17 does the open beta launch tomorrow?
1:32:19 >> Yeah. So, the open beta launches
1:32:21 tomorrow and if you go to exist.ai
1:32:24 today, they've got a really cool proof
1:32:27 of concept of what it's going to be and
1:32:29 a link to join their Discord so you can
1:32:32 gain access to the beta tomorrow. All
1:32:33 right, great. I want What did Oh, this
1:32:39 thing's still working. Virtue entangled
1:32:41 minds game overview.
1:32:44 Okay, so now we're having Grock for
1:32:46 grock for read our screenplay,
1:32:50 analyzed it for sci-fi and logical
1:32:53 inconsistencies,
1:32:55 made recommendations, rewrote the most
1:32:58 problematic scene in the screenplay,
1:33:01 and we've now asked it to design a game.
1:33:03 Okay. And then we'll go check out
1:33:05 exist.ai in a second. Anyone have any
1:33:08 questions, thoughts?
1:33:10 Why aren't you using Grock's website?
1:33:11 Because when I go to Grock's website, it
1:33:14 doesn't give me access to Grock 4. I
1:33:16 have to upgrade. But if and logged
1:33:18 logged in with my, you know, my account,
1:33:21 but if I go to Twitter
1:33:23 X and I go to Grock, I do have access to
1:33:26 Grock 4. So, I don't know what the
1:33:28 glitch is, but there's a glitch. All
1:33:31 right. This wrote an incredibly detailed
1:33:33 game description here, which is what I
1:33:35 asked it to do.
1:33:37 But let's go look at that. Uh, what are
1:33:39 we waiting for to render? Wasn't
1:33:41 something rendering?
1:33:43 Oh, the the graphic novel thing in chat
1:33:46 GBT decoherence part one. Wow, look at
1:33:49 this.
1:33:51 Um, let me download this.
1:34:05 That's kind of a cool style.
1:34:07 Decoherence. Virtue tech quantum core
1:34:10 room midnight. There's powers.
1:34:15 IBM and Google's quantum supremacy
1:34:20 toys. Child's play.
1:34:22 The assassin. What do we do here? What
1:34:25 exactly do we do here? Simulations that
1:34:28 rewrite reality. H. Airlock engaged.
1:34:32 I'll show you. Purge cycle. So it didn't
1:34:36 it didn't really we would have to yeah
1:34:39 this was this is not great chatbt did
1:34:42 okay here but we would have to have it
1:34:45 um
1:34:48 write each of these panels first and I
1:34:50 think the way you do these I don't know
1:34:52 I don't know there's probably ways I
1:34:54 could prompt that better but that's not
1:34:56 great anyway interesting okay back to
1:34:59 Grock
1:35:01 so game design virtue and tangled minds
1:35:04 is webbased action adventure video game
1:35:07 directly inspired by the screenplay
1:35:09 Virtue reimagined as an interactive
1:35:12 thriller that captures the essence of
1:35:14 the story's highstakes conspiracy,
1:35:16 mind-bending sci-fi elements, and pulse
1:35:19 pounding action. Okay. Built using WebGL
1:35:24 for stunning browser compatible
1:35:25 graphics,
1:35:28 powered by frameworks like 3JS or
1:35:30 Babyon.js
1:35:32 for 3D rendering. The game runs
1:35:34 smoothly. The core hook.
1:35:37 Players step into the shoes of detective
1:35:40 Tony Mullen.
1:35:43 The game is free play with optional
1:35:45 microtransaction for cosmetic skins. All
1:35:48 right, I'm going to say um this is too
1:35:52 ambitious.
1:35:54 Let's
1:35:56 make this a very impressive.
1:36:00 What do they call that? casual
1:36:03 gaming experience.
1:36:11 I love your thinking,
1:36:16 but let's bring
1:36:20 it down
1:36:23 to a game mechanic
1:36:29 that will
1:36:32 please gamers and
1:36:36 captivate.
1:36:37 Um, non-gamers.
1:36:41 All right. Boom.
1:36:43 Have chat GBT give you the prompt. Wait.
1:36:47 Bouncing back and forth between
1:36:48 ideoggram the reference image from chat
1:36:51 GBT along with his prompt
1:36:54 to stack the quality. That's pretty
1:36:55 good. Have Chat GBT give you the
1:36:57 prompted prompt and feed it to ID. Okay.
1:37:01 Get Claude to create the game. Yeah, I
1:37:03 was going to run over to Claude to
1:37:04 create the game or or lovable um
1:37:09 exists AI office hours.
1:37:14 Oh, office hours tomorrow. Got it. Yeah,
1:37:16 got it.
1:37:17 >> Just a note for you for later.
1:37:18 >> Yep. Got it. Okay, cool. Uh, Virtue
1:37:21 Quantum Echoes is a web-based casual
1:37:23 puzzle adventure game inspired by the
1:37:25 screenplay
1:37:26 designed for bite size play sessions.
1:37:34 Play time per session 5 to 15 minutes.
1:37:36 Story arc.
1:37:38 Graphic style. Graphics are very
1:37:41 impressive for a web game. Vibrant
1:37:42 cinematic 2D art 3D effects.
1:37:48 All right. It seems to have described
1:37:49 this quite well. Okay. So, we're just
1:37:52 going to go capture all of this.
1:37:58 Time check 9:40. Okay, fine. We're going
1:38:00 to do it quick. We're going to do it
1:38:01 quick. We're going to go to
1:38:04 here. We're going to go to Claude. We're
1:38:05 turn this into Claude. So, we're going
1:38:07 to go to two different
1:38:10 two different vibe coding platforms. So,
1:38:12 Claude is just your regular Claude. Deep
1:38:15 answers and more sources.
1:38:20 Oh,
1:38:23 try integrations and research today.
1:38:26 Connect your everyday tools. Add
1:38:28 pre-built integrations
1:38:31 or create your own.
1:38:33 Hours of research in minutes. Okay, we
1:38:37 just want a [ __ ] game.
1:38:39 Please
1:38:40 make this game
1:38:49 YouTube comment. Lovable is great for
1:38:51 this. Tell Grock
1:38:54 that's where you're that's where you're
1:38:57 going for correct formatting. Oh, okay.
1:38:59 Great.
1:39:02 All right. We'll see how Claude does
1:39:04 with that. We'll go back to Grock and
1:39:05 we'll say um
1:39:08 this is great.
1:39:11 I'm
1:39:14 going to vibe
1:39:19 code this on lovable.dev.
1:39:24 Please re
1:39:28 rewrite
1:39:31 this spec
1:39:36 so that lovable
1:39:39 creates a remarkable
1:39:48 and impressive
1:39:50 game. exclamation point.
1:39:54 Boom. All right. So, that's going to go
1:39:57 off and think. Let's see what's Claude's
1:39:58 doing here. It's writing code, baby.
1:40:01 We're writing code. We'll do it live.
1:40:04 We'll do it live.
1:40:17 Um, yeah. Uh, let's see. Justin,
1:40:20 awesome. If you are new to this
1:40:23 generative AI stuff, um
1:40:27 what I'm doing right now is sort of
1:40:29 particularly chaotic and and chat add-
1:40:33 like. Um,
1:40:37 but what's happening is this.
1:40:40 A test that I did in Gro 4 was good
1:40:43 enough
1:40:45 that it
1:40:48 shook my brain up and said, "Oh,
1:40:52 now that you have this updated
1:40:53 screenplay, let's run over to
1:40:56 Midjourney." Oh, look. Begin. So, here's
1:40:58 our game.
1:41:01 Begin investigation.
1:41:04 What am I supposed to do? Quantum
1:41:06 resonance.
1:41:08 Quantum shard zero.
1:41:11 What am I supposed to do?
1:41:22 How do you play?
1:41:30 Fun. Match three puzzle system with
1:41:33 intuitive tapclick controls. Well, it's
1:41:36 intuitive, but I don't know how to play.
1:41:43 >> My guess would be that you click three
1:41:46 of the same colors.
1:41:48 >> Yeah, but
1:41:48 >> if it would let you click more than one.
1:41:50 >> It's not letting me click more than one.
1:41:54 Quantum shards.
1:41:58 >> All right.
1:41:58 >> Quantum jank. Quantum Jang. All right,
1:42:01 that's shitty. Okay, so here's the
1:42:03 lovable.
1:42:05 Here's the lovable description.
1:42:12 Fully tailored for lovables AI. All
1:42:14 right,
1:42:16 we're going to go to Lovable now. So,
1:42:18 you lose, Claude. You suck. Um,
1:42:22 lovable.dev.
1:42:26 I beat chat GPT into submission. And you
1:42:29 all be proud of me. I finally got got
1:42:31 something finished. Wait, finally got a
1:42:33 finished what? I finished
1:42:36 God damn that thing artist CV. Oh, good.
1:42:40 Yeah, sometimes you got to beat these
1:42:42 things up. Don't One of the most
1:42:45 important things if you're new to AI, if
1:42:47 you're new here and you're new to AI,
1:42:51 um
1:42:53 do not just take what these systems give
1:42:56 you. Don't just if they give you [ __ ]
1:42:59 tell them they gave you [ __ ] Don't
1:43:01 accept it.
1:43:07 I you know, I almost feel like the the
1:43:10 teams is now pro. I don't care about
1:43:12 that. I just want this game. Go make me
1:43:15 a game. Come on, lovable.
1:43:17 Make it not suck.
1:43:24 Did you try to click two of the same
1:43:25 color next to each other?
1:43:28 No. But that's gone now.
1:43:33 All right. And we're going to get out of
1:43:35 here soon. I'm just going to We'll see
1:43:36 what we'll see what Lovable does here.
1:43:40 See if Lovable gives something playable
1:43:42 on the first shot. I also I haven't read
1:43:44 I didn't really read what Grock wrote,
1:43:46 so it's possible it wrote a game that's
1:43:48 not possible.
1:43:53 narrative forever. I'm glad to hear you
1:43:54 say that, Kyle, because sometimes I'm
1:43:56 not patient with chat GBT. Yeah, it's
1:43:59 it's super important.
1:44:02 Think of think of every single one of
1:44:05 these le uh uh um generative AI tools.
1:44:10 every single one from large language
1:44:13 models to image generation tools to
1:44:16 movie tools, all of them to fancy
1:44:19 sophisticated tools like LTX Studios,
1:44:21 all of them
1:44:26 lose the expectation that the tools are
1:44:30 the genius. No, you're the genius and
1:44:34 you strap on generative AI tools like a
1:44:38 backpack, like a jetpack,
1:44:41 and and they're amplifying your ideas.
1:44:44 So rather than thinking, I've got to get
1:44:46 this thing to to generate something
1:44:48 genius. No, no, you're the genius. The
1:44:52 AI is there to amplify your ideas. So,
1:44:55 if you kind of shift that mental mindset
1:44:58 from it's going to do something amazing
1:45:01 to you're going to do something amazing
1:45:04 with AI, it helps a lot because then you
1:45:07 can just say to the tool, nope, you got
1:45:08 it wrong. That's not what I want. That's
1:45:10 not what I want. That's not what I want.
1:45:12 And then you'll get better and better
1:45:14 and better
1:45:16 at feeding your prompt
1:45:19 your point of view. And the better you
1:45:21 get at feeding it your point of view,
1:45:23 the better it's going to get at
1:45:24 reflecting that back to you. And that's
1:45:27 when these things get magical. It's a
1:45:30 it's a it's a big shift. You're right.
1:45:33 Because I feel like sometimes I have to
1:45:35 hold its hand and walk it through stuff
1:45:36 that makes complete sense in my head. It
1:45:38 just doesn't seem to grasp it. Right.
1:45:40 Because it doesn't it doesn't grasp it
1:45:43 right.
1:45:45 it it can only grasp it when you give it
1:45:48 enough uh context
1:45:52 or point of view,
1:45:54 right, for it to understand what the
1:45:57 [ __ ] you're talking about. And that's
1:45:59 the art of AI. The art of AI is you
1:46:03 being clear enough about
1:46:07 what you're trying to accomplish
1:46:10 to ask for that and then iterate back
1:46:13 and forth until you get it. And when you
1:46:15 do it right, what happens is the AI
1:46:18 gives you back something that maybe
1:46:20 isn't quite perfect, but maybe there's
1:46:22 some new idea there, and you're like,
1:46:24 "Oh, part of that's really good. Let me
1:46:26 take that new idea and incorporate that
1:46:28 into my vision." and and then you start
1:46:31 building and expanding your ideas. And
1:46:33 that's where it gets really [ __ ]
1:46:35 magical. Okay. Quantum echoes ripple
1:46:38 through your consciousness.
1:46:40 Connect quantum orbs to unlock memory
1:46:43 fragments.
1:46:46 Okay, we're getting there. I like it.
1:46:56 So, I broke it somehow.
1:46:59 How do I restart? But this kind of
1:47:01 worked.
1:47:04 I've got 400 points. I've got two
1:47:06 shards.
1:47:14 Skip cutscene. Reset level. Quantum
1:47:17 boost. Reset level.
1:47:20 All right. So, this one sort of worked.
1:47:22 Doesn't quite work. We're not going to
1:47:23 publish it. It's not great. All right.
1:47:26 That's it. Um,
1:47:30 let me
1:47:32 ask not what AI can do for you. Ask what
1:47:34 you can do with AI. John Kennedy. Yeah,
1:47:37 exactly. No, my thing is this. I I think
1:47:40 we ask the wrong questions. I think
1:47:42 we're asking a lot right now. How can I
1:47:45 get the most out of AI?
1:47:48 And I think if you shift that question
1:47:50 to say, how can I get the most out of
1:47:52 myself with AI,
1:47:55 it completely transforms things. So I'm
1:47:58 working on a book right now called feed
1:48:00 your prompt. And the whole idea is that
1:48:02 you feed your prompt you
1:48:05 you use using it to amplify you rather
1:48:08 than just feed it stuff. So it's the
1:48:10 genius. It's not the genius. You're the
1:48:12 genius. So you figure out what you want.
1:48:16 You lead with that and then everything
1:48:18 you interact with is in the context of
1:48:20 what you want. It'll transform
1:48:22 everything. All right. Oh, pop over to
1:48:25 exists.ai. Okay, I will do that.
1:48:29 One more thing. Exists.ai.
1:48:33 Let's see if this looks any good. Do
1:48:35 they do they have a website? Yeah, the
1:48:37 website's okay. Sign up for the beta.
1:48:42 Go to Discord.
1:48:44 join exists.
1:48:50 Welcome to exist. Let's customize.
1:48:53 >> So, you can finish doing this later, but
1:48:56 if you go back to the website and scroll
1:48:59 down a little bit, um there's
1:49:03 a more to the website than just the
1:49:05 signup page.
1:49:06 >> Okay, cool. But I want to get signed up.
1:49:09 Am I signed up?
1:49:12 Where do I sign up? Do I need to?
1:49:15 >> I think just being on the server as far
1:49:17 as I know.
1:49:18 >> Welcome.
1:49:19 There you go. Hey Kyle, Shannon,
1:49:21 welcome.
1:49:22 >> Go to general.
1:49:25 Excited to be here
1:49:29 on AI Learning Lab Live.
1:49:35 I'm gonna go tick tock.comarninglab
1:49:45 right now.
1:49:49 Okay. Um, going back to the website,
1:49:54 scroll down. Oh, that was cool.
1:50:00 Okay, so they use something like
1:50:03 Unreal Engine or probably not even
1:50:06 probably Unity
1:50:09 that. So maybe they threw an AI front
1:50:11 end on top of Unity. That's probably
1:50:13 what they did. Cool.
1:50:16 Because that doesn't look like Unreal
1:50:17 Engine 5.
1:50:20 Looks more like Unity
1:50:23 game play.
1:50:29 Yep. All right. Everything online. Sign
1:50:33 up for the beta.
1:50:36 Games meet cinema
1:50:39 content showcase.
1:50:47 Huh?
1:50:49 Are there movies in here?
1:50:54 Just these ones. Does the show like a
1:50:56 little bit of gameplay, right?
1:51:07 So, I wonder if you're going to be able
1:51:08 to
1:51:11 prompt the games into existence. Well,
1:51:13 we'll find out tomorrow. All right.
1:51:18 You can only click on three or more more
1:51:20 adjacent. Yeah, that that I got. Okay,
1:51:22 cool.
1:51:23 All right, G for real. I think it it has
1:51:27 a little angst now. I like it. Okay.
1:51:29 Yep. Right now I'm dealing
1:51:32 I can change have good internet, not
1:51:35 both. Can you make a game exactly like
1:51:38 another only different pictures and not
1:51:41 copyrightiting?
1:51:43 Um,
1:51:45 yeah. I mean, you can
1:51:48 Let me go look at some lovable things
1:51:50 here. Is this lovable? Oh, that's
1:51:54 um
1:51:54 >> that's my that's my version of Missile
1:51:57 Command.
1:51:57 >> Yeah, exactly.
1:51:58 >> With zombies and kids in sand.
1:52:01 >> Exactly. That's pretty funny.
1:52:03 >> Let me go back to my lovable
1:52:06 dashboard.
1:52:13 Come on. Where's my dashboard?
1:52:16 Lovable.dev. Go.
1:52:20 Okay. Uh Kyle's lovable
1:52:25 come here
1:52:30 workspace.
1:52:41 Why can I not find my Where's all my
1:52:43 games? Oh, here they are.
1:52:46 There's feed your prompts.
1:53:02 Is this a good one?
1:53:06 Yeah. So, here's an Asteroids game.
1:53:09 I asked
1:53:13 Press space to start.
1:53:18 Yeah, this one. So, let me let me share
1:53:24 differently.
1:53:29 Okay. Um,
1:53:35 so this
1:53:37 I just this was a single prompt.
1:53:41 This was a what do they call it? A one
1:53:43 shot. a one shot
1:53:49 and it just [ __ ] made asteroids this
1:53:52 good. This is Lovable that did this.
1:53:54 Lovabledab. So yeah, you can
1:53:58 you can say, you know, make Flappy Birds
1:54:01 but with turtles. Make Flappy Turtles,
1:54:04 Jumpy Turtles.
1:54:08 I mean, the game play here is
1:54:11 staggeringly good for,
1:54:16 if you've been watching this channel for
1:54:18 a while, you've seen me make some really
1:54:19 bad clones of Frogger.
1:54:22 And uh this is a pretty [ __ ] good
1:54:25 clone of Asteroids. If you have ever
1:54:27 played the Atari 2600 Asteroids, you
1:54:30 know the definition of disappointment in
1:54:33 game cons.
1:54:36 Um, this is more like the arcade than
1:54:38 the Atari 2600. Anyway, all right, I'm
1:54:42 getting out of here.
1:54:45 [Music]
1:54:52 Okay, beautiful. Tomorrow, today is
1:54:56 Thursday. So, tomorrow at 11:00 a.m.
1:55:00 Mountain time on LinkedIn, I have this
1:55:04 thing called office hours.
1:55:06 And how you get there is you go to my
1:55:10 profile, Kyle Shannon, and you know, you
1:55:13 know what we've got to do? I Rick
1:55:15 McCauley did this, who's in the in the
1:55:17 live tonight. Um,
1:55:21 I've got to put Oh, yeah. Go to my
1:55:24 LinkedIn and and just ignore all the
1:55:26 hate comments from from my chain of
1:55:29 craft uh article. No, go to my LinkedIn
1:55:33 and go to events and if you see AI
1:55:36 office hours, just copy that URL, throw
1:55:38 it in your in your calendar for tomorrow
1:55:40 at 11:00 a.m. Um it's a really good
1:55:43 group of people that shows up. There's
1:55:45 lots and lots to talk about. Um,
1:55:50 [Music]
1:55:51 so there's that. Tomorrow night, it's
1:55:54 going to be Friday night date night, but
1:55:56 it's going to be Friday night late night
1:55:58 date night because my wife has an art
1:56:01 opening tomorrow night. So, I'm going to
1:56:02 be at the art opening. Um, which means
1:56:04 I'll be back here late. So, it might be
1:56:06 900 p.m. Mountain time. It might be
1:56:08 9:30. Um, it'll probably just be a short
1:56:11 one, but we will have Friday night date
1:56:13 night. Okay? So, come back here tomorrow
1:56:15 night. And then uh and then let's see
1:56:18 that's it. Come to office hours
1:56:20 tomorrow. Um join the AI salon, the
1:56:24 salon.ai.
1:56:26 Um
1:56:29 that's it. I got to get my [ __ ] together
1:56:31 on marketing all the different things I
1:56:33 have going on. So I think we're going to
1:56:35 do that with QR codes. So we'll be able
1:56:37 to just pop up graphics that have QR
1:56:39 codes on them. You can zap them. All
1:56:41 right. Yay. Congratulations, Gabby.
1:56:44 Yeah, the show is really beautiful. It's
1:56:45 really, really good. All right. Um, I
1:56:49 think that's it. Yeah, that's it. All
1:56:52 right. Peace out and I will see you on
1:56:54 tomorrow. Bye.
1:56:57 [Music]