Futuristic

FUTURISTIC 29

[00:00:00] Cameron: Welcome back to the Futuristic Steve Sammartino, episode 29. You’re looking sharp as always, man. Look at you. You look like a, I don’t know, he’s just stepped out of a venture capital meeting in Silicon

[00:00:20] Steve: Finance, 6, 5, Trust Fund, Blue

[00:00:23] Steve: Eyes.

[00:00:24] Steve: really finance, 5 6. Finance, 5 6, not 6 5. I’m

[00:00:29] Steve: not 5

[00:00:29] Steve: 6, but anyway.

[00:00:31] Cameron: I look like I’ve just come off stage at a Grateful

[00:00:33] Steve: You actually do, you’ve got the, what have you got on there? The

[00:00:37] Steve: bandana.

[00:00:38] Cameron: It’s a beanie, cold enough. One of the few cold enough days in Brisbane to wear a beanie. I’m going to

[00:00:44] Steve: have you got on there? A

[00:00:45] Cameron: I’ve got a

[00:00:46] Cameron: got A

[00:00:46] Cameron: bit of,

[00:00:46] Steve: lovely 23 degrees or

[00:00:48] Steve: something.

[00:00:48] Cameron: yeah, that’s

[00:00:49] Steve: is real.

[00:00:50] Cameron: I got a, I got

[00:00:50] Cameron: a, got an annoying cough. So I’m going to have to keep muting my mic to cough. Hold

[00:00:54] Cameron: on. Yeah. All right. Uh, Steve, uh, tell me one. thing of note you’ve done that’s futuristic y since we last spoke.

[00:01:03] Steve: Futuristic y. Is, I’m trying to balance my screen usage. I’ve never really looked at how long, I’ve spin on the screen because it was always justifiably, well, it’s work for me. And to be fair, I don’t really spend a lot of time doing things on the screen that aren’t work. Uh, but I looked, and it’s an average of 5 hours and 46 minutes a day on my phone.

[00:01:27] Steve: It’s an average of 92 pickups. And so, I went out and bought a few books. This week I asked you for a couple of recommendations and I have some on the way and every time I go to pick up my phone when it’s not an absolute necessity to do a task on it for work, I’m going to instead pick up the book. Uh, because I actually think that there’s a really big difference between knowing about, hearing about something and knowing about it.

[00:01:58] Steve: And I think we hear about a lot of things and the screen lets you hear about a lot of things. I don’t think Enables deep understanding, like reading a book and really going deep into it, into a topic.

[00:02:13] Cameron: Well, one of the things I was going to add to our talking list, but I ran out of, well, we had too many things already, but I was started watching a video last night, uh, an interview with Dwarkesh Patel, who does a lot of AI videos on YouTube, but he was being interviewed on another YouTube channel. And he was talking about how he thinks you don’t learn very well from books.

[00:02:35] Cameron: And I think we talked about this in an earlier episode, and he’s come up, he said, like, when I interview people that are experts in a topic, I need to, you know, get smart on that topic really quickly, and often they’re topics I don’t know a lot about. Said he had a geneticist on his show recently. So he’s developed this workflow that he puts into Claude, his AI tool of choice at the moment, that creates, um, uh, spaced repetition.

[00:03:03] Cameron: And he said he can upload an EPUB. of somebody’s book into Claude and say, tell me what the key points are that I should know before I interview this person and then test me on those. Um, you know, give me questions and answers and test me on it to test my basic knowledge. And he’s developing a system, which I haven’t fully finished the video to integrate it yet, but I think, you know, we’re moving into an era where Books are going to be a component of how we learn, but there’ll be an AI layer wrapped around that, that will create a, um, educational context.

[00:03:42] Cameron: What do you call it when somebody’s an educator? You call it, uh, uh, there’s a word for it that’s escaping my memory. Hold on, I should ask GPT. Um, what, what, what’s, what do you call a system of education that, uh, leads you through the learning process?

[00:04:02] Cameron: Scaffold.

[00:04:04] Cameron: Or pedagogee was the word I was trying to think of..

[00:04:07] Cameron: He’s talking about it’s a scaffold

[00:04:09] Steve: yeah, and that’s a nice way, and that’s a nice visual of Understanding the elements and almost seeing a matrix design of where you could go to get different types of learning. And just, uh, look, I think that what you’re saying is right. There’s a multitude of ways of learning. I mean, I’ve always been very good at kinesthetic learning, where actually hands on doing, and it’s one thing with startups that I’ve always realized.

[00:04:30] Steve: It’s one thing to read about startups and it’s another one to get out there in the battlefield and try and get people to use your product and, and raise capital and, and, you know, exiting theory. is really important. Uh, and you always learn a little bit more by doing something than actually reading about how to do it.

[00:04:46] Steve: Like you can’t learn how to ride a bike by reading about it. You have to hop on the bicycle. You can probably get some tips. Sure. Uh, the context of what I was saying, and I, it’s probably good that you brought this up because I can clarify. I’m not saying that I want to read more books to learn more. What I’m saying is, of what I’m doing with reading, I want it to be physically on paper.

[00:05:11] Steve: So the context is the reading context. And I actually looked up some research on it because I was curious about it because I have this theory that I get more out of a book and we all know about the distraction side of it. It’s easy to get distracted when you’re on a phone and not stay in the book as deeply because it’s singular versus a phone which has a multitude of uses.

[00:05:31] Steve: But some of the research suggests that the subconscious impact on the light and tiring out your brain and the way that the words are absorbed is very, very different to ink on paper. And another one which blew my mind, I never thought of that, is that the pixels that are on a screen, uh, flashing lights really, really quickly.

[00:05:52] Steve: And apparently your brain doesn’t digest the words as well. So there’s some research coming out, uh, in relation to screens not enabling people to digest information as well as books. So there’s that part to it. And there’s also that you go deeper into a book than you do just scanning the news on tech.

[00:06:09] Steve: So I know about a lot of stuff or hear about a lot of stuff, but I want to know more deeply about

[00:06:13] Steve: it. So that’s that’s what I’m really trying to do.

[00:06:18] Cameron: I think you’re probably right. I mean, I think, hey, well, let’s, let’s break that down. Like, um, yeah, reading rather than scrolling. Um, you know, I, I’ve developed the habit over the last few years where if I have some spare time and I’m not working and I’m picking up my phone, it’s usually to do Duolingo, to practice my Italian,

[00:06:39] Steve: is great.

[00:06:40] Cameron: or it’s, or it’s chess.

[00:06:43] Cameron: I’ll be playing a game of chess against somebody on Chess. com and I’ll have to go and make a move. Oh shit, that reminds me. Oh no, I did make a move this morning. And. Or I’ll read. Um, sometimes it’s Reddit, um, but often it’ll be a book that I’m in the middle of. I’ve always got like ten books on the go, so it’ll be one of those.

[00:07:01] Cameron: Um, But for me, and I think you’re probably right in terms of reading paper, but for me, the upside of being able to, I’ve got a shortcut on my iPad that if I copy a, if I copy a block of text in a book and I run this shortcut, it’ll paste it into a running note that I have in Obsidian.

[00:07:19] Steve: got benefits that a book will never

[00:07:20] Steve: have.

[00:07:20] Steve: Absolutely. No doubt. For sure.

[00:07:22] Cameron: I can, I can pull up GPT and

[00:07:24] Cameron: say, what’s this word mean? Who was this person? What’s that I don’t know. So I think there’s, there’s pros and cons, but I, I, I think you’re probably right that ink on paper is

[00:07:33] Steve: Just something about, I just want to extricate myself from the screen somewhat because I just think I’ll get reading done as well because it’ll be more purposeful and less distracting. So there’s, there’s a multitude of reasons, but that’s, that’s what I’m trying to do more

[00:07:46] Steve: of and just pay attention to the physical world.

[00:07:49] Cameron: Well, I’ve, I, you know, I know every

[00:07:51] Cameron: time we do this, I talk about coding because that’s where I spend a lot of my time, but I really felt like I leveled up my coding in the last couple of weeks. There’s a particularly big thorny problem that I’ve been working on and getting nowhere for days and days and days on it.

[00:08:06] Cameron: And then I learned. To, if there’s a particular problem. See the way the AI tools work at the moment, if you’ve got a really big script and it’s doing, it’s doing like 25 functions, but one function’s not working properly, and you tell the ai, I got a, I got a an error code, I’m gonna try to do this. It’s immediate thing will, it’ll rewrite the entire

[00:08:28] Cameron: script and go, okay, run it

[00:08:29] Cameron: again. And so I’ll wait till it finishes and then I’ll run it again and it still won’t work, and I’ll give it the error code and it’ll go, oh, okay. Let’s put some more logging in. And then it’ll give you the whole script again, and you run it again, and you give it the log, uh, the error log. And he goes, Oh, okay.

[00:08:45] Cameron: Let’s put some more logging in. And you get in this loop and it’s really frustrating. So what I learned to do is to isolate the bit that’s not working and say to the AI, No, no, no, no, no, no, no, no, no, no, no, just stop. Write me a script that, that has three or four ways in it. To solve this problem, like the particular thing that’s not working.

[00:09:07] Cameron: And let me just run those independently. So I’m not running the whole script until, so I isolate, test, figure out the method that actually works, and then integrate that back into the bigger script as a new function. And, um, I did that over the last week or so, and it’s solved a whole bunch of problems.

[00:09:28] Cameron: Just isolate. Test and then combine or integrate it back into the main thread has really,

[00:09:35] Steve: it’s a really good point, Cameron. And I don’t know how the large language models.

[00:09:40] Steve: But it seems to me as though you cannot iterate verbally on the previous finding. And I know that OpenAI is quite guilty of this. I was doing a presentation, uh, getting ready to go to New Zealand early this week, and I had an image that I wanted in 16, or in widescreen format. Is it 4 is it? And I said, this image that you’ve developed, Now go,

[00:10:09] Steve: 16,

[00:10:09] Steve: 3, I’m way off, um,

[00:10:12] Cameron: It’s 3, 4 and 16,

[00:10:13] Steve: there you go, 9, okay. said, develop this image in widescreen format, redo it. It gave me a new one. I’m like, no, no, this image here, exactly the same, widescreen. Can I do it? It actually is. And I’m wondering if what you’ve found as well is it’s very, very poor at verbal iterations, unless you take out the piece of the puzzle and get it to reframe that exact piece and then insert it back in, it always goes back to the well and starts again.

[00:10:42] Steve: And I’ve noticed it can’t do the same thing twice. It’s very, very bad at doing the same, which is interesting and cool. Uh, but if you ask it to recreate an image or create that image again, it won’t do it. It’s always a little bit different. And I think that comes down to the probabilistic nature of it and the way that it works.

[00:10:59] Steve: And it’s really noticeable, especially from an image perspective, and I imagine from a code, based on what you’ve said, and it reruns the code base again, because it goes, oh, you didn’t like that, I’ll just do the whole thing again. I’m hoping that ChatGPT 5 and others come out with a greater iterative ability, and for it to understand what’s required within the iteration of the previous development, and it doesn’t seem to be there.

[00:11:24] Steve: And this goes to show What, what you’ve been doing and what I’ve been doing, and I’ve been saying this on stage so much is getting to work well with the AI is an absolute art form that cannot be understated. And the only way you learn that art form is by experimenting with it and playing with it. And I wonder, If that’s always going to be the case, even if we have extraordinary AIs that are close to the general level, because it, it doesn’t, it’s trying to satisfy your needs and it is not inside your brain.

[00:11:56] Steve: And so that is the complexity that requires the ability to prompt well and to work with it and iterate and, and craft the angles and take pieces of the puzzle out and then reframe those and then put them back into the puzzle. It’s a really interesting insight.

[00:12:10] Cameron: Mistral. Mistral. Large 2.

[00:12:15] Steve: You’re going to have to give me a download on this cam because I didn’t test

[00:12:17] Steve: it.

[00:12:18] Steve: I had a look and read

[00:12:19] Cameron: I’ve been using, I’ve been using Mistral for my coding. For the last couple of weeks, their latest model, Large 2, Mistral’s a French company. Um, I believe made up of former employees of Meta and Google, was founded in April, 2023.

[00:12:41] Cameron: And, um, they, uh, have a open source model. Well, we’re going to talk a little bit of open source on this show, if we have time. Their latest version, um, Large 2, absolutely. Stunning when it comes to coding. I haven’t used it for much else, but it shits all over GPT 4. 0 and 4. 0 when it comes to coding, just gets it right almost every time, understands what I want, nails it, so much less stress, uh, using it for coding and, uh, that in and of itself is great.

[00:13:23] Cameron: So if you, if you haven’t tried it, you can use it for free. If you go to Mistral. ai, M I S T R A L, they have something called Le Chat And le plateforme, le plateforme is the API. Weirdly, I signed up for the API and I haven’t been able to get it to work. Because one of the problems with le chat is, as my thread gets long, which it can often do with coding, it slows right down.

[00:13:52] Cameron: It’s certainly not fast enough. Fast. It gets really slow, like one letter at a time, typing out, like my, my, you know, my dad trying to type something on a keyboard 20 years ago. Um, uh, so I tried to use the API, and I can’t get the API to work for some reason, and their tech support is not, has not been great.

[00:14:12] Cameron: I’ve been asking their tech support for a week to Figure out what’s wrong with my API access. And they’re not really very responsive because they’re French. But the, uh, the, but the LeChat service in and of itself is pretty good. If

[00:14:24] Cameron: you don’t have massively long threads that

[00:14:27] Steve: in the chat. It’s right up So I highly recommend.

[00:14:30] Steve: I said it’s right up there with Ordineter. Which is what they call a computer because they’ve got the French language division which makes sure they don’t have Western words.

[00:14:40] Cameron: Ah,

[00:14:41] Steve: And that’s the computer, is it already

[00:14:43] Cameron: So anyway.

[00:14:43] Cameron: Mistral

[00:14:45] Cameron: 2 is great and if we have time after our guest comes on in six or seven minutes, I want to talk more about open source versus closed source because it’s been a big topic, uh, in videos I’ve been watching this week. Um, The Friend, Steve, another bloody AI pendant, uh, has hit the

[00:15:06] Steve: Yes,

[00:15:07] Cameron: by a 21

[00:15:09] Cameron: year old guy called, uh, Schiffman.

[00:15:13] Cameron: Uh, did you

[00:15:14] Steve: I did, he’s an interesting cat. You know, he, it was interesting. His turn of phrase was interesting. What I did like was that he said productivity’s over in his Wired interview, and I kind of, I do feel as though there’s so many productivity tools, we need productivity tools to handle the productivity tools.

[00:15:32] Steve: It’s like, it’s eating its own tail. And at least he had a point of difference where this. Is to help people have a friend and for loneliness, which is a real problem in the world. I mean, he just forgot to put the word in the name of the product imaginary before friend, which would have been so perfect that I would have loved that.

[00:15:51] Steve: And I think he really should have done it, missed an opportunity.

[00:15:54] Steve: Also really curious that he spent almost half the venture capital on getting the dot com.

[00:16:00] Steve: Uh, he spent 1. 8 million on buying friend. com. Uh, which I thought it was interesting, Look, this is one of those

[00:16:10] Steve: ones that I can see that there could be some utility

[00:16:13] Steve: for

[00:16:13] Steve: it.

[00:16:13] Steve: Definitely. Yep.

[00:16:14] Cameron: So tell people, tell people what

[00:16:16] Steve: So, it, it’s, it’s, interacts with you on your day to day where you can talk to it. It sends you messages which I think appear on your phone where it talks back to you. It might even just make comments without asking saying, hey it’s

[00:16:27] Steve: great to be outside. It’s good that we’re walking around getting some fresh

[00:16:30] Steve: air.

[00:16:31] Cameron: You wear it you Wear neck like a little necklace.

[00:16:34] Cameron: around your

[00:16:35] Steve: Round your necklace and it observes and listens. It’s always on, as is Siri and Google and everything else. It’s always on and it makes comments about your life and talks to you and send you text messages and interacts with you in a way that is kind of like, let’s say an imaginary friend or a little AI friend that comments on your day and has opinions and can summarize, Hey, you know, you had that chat with whatever.

[00:16:59] Steve: Did you follow up and do that email? I don’t, stuff like that. It seems interesting because I don’t think it’s over promising in what it can do or trying to replace a screen which none of them have been able to do. All of them have been technologies in search of a problem where the smartphone usurps their utility.

[00:17:15] Steve: This one, I think, I think it could work. I don’t think it’d be anything ever more than niche, something incredibly niche. I could, I could see it potentially working. Um, just because it’s a little bit more single minded on what it does. Because the smartphone is the ultimate Swiss Army knife that does it all and does it incredibly well.

[00:17:35] Steve: And it’s very, very difficult to see that being supplanted. But I think that the single minded proposition of it, commenting, talking to you, helping lonely people, You know, you can, I feel like it could have some utility. So, look, that said, do I think that having an imaginary AI friend is better than sitting across the table or me chatting with you or picking the phone up?

[00:17:59] Steve: No. But maybe some people don’t have that opportunity or that wherewithal or that confidence or have a base of

[00:18:05] Steve: friends. So that’s kind of my taking on

[00:18:07] Steve: it. Yeah.

[00:18:09] Cameron: And I don’t think it’s an either or. I think in the near future, we will have an AI friend or multiple

[00:18:15] Steve: Yeah, right. Yep.

[00:18:16] Cameron: and as well as real friends, if you’re lucky enough to have real friends. Personally, I don’t like real people that much. I like you. I like Tony. I like Ray. I like my wife.

[00:18:28] Steve: pretty sure, I’m not even sure if you like me. You say that you do, but you’re, you know, you’re

[00:18:33] Steve: an unusual cat, you know that.

[00:18:37] Cameron: There’s a small group of people that I actually like to hang out with. But, getting back, I look, I think, um, trying to sell these sorts of, uh, single purpose AI widgets is probably a path to nowhere. But, I do expect that in the very near future, my watch, or my phone, or my iPad, or all of those, will have an AI, or several AIs, that will be my friend.

[00:19:09] Cameron: We’ll be listening to everything that happens, watching everything that happens, and maybe I’ll have different friends. Maybe I’ll have, I’ll have the friend that lends a sympathetic ear when I’ve had a fight with Chrissy. Maybe I’ll have a friend that’s like, dude, you, you, you really weren’t very productive today.

[00:19:27] Cameron: You know, you know, I’ll have my coach,

[00:19:29] Cameron: you know, The one that tells me that,

[00:19:31] Steve: idea. Your own personal

[00:19:33] Steve: Jim Rohn, straight around the neck.

[00:19:36] Cameron: yeah, yeah. Although the one that’s telling me that, uh, I need to be, I need to spend more time with my kids. I’ve been

[00:19:43] Steve: You’ve only hung out with Ox.

[00:19:44] Cameron: or maybe it’s all wrapped up

[00:19:45] Cameron: into

[00:19:46] Steve: had one hour with Fox today, your average is 2. 3 hours, you’ve let him down.

[00:19:49] Cameron: I was thinking of my older kids, my 23 year olds, but, uh, yeah, they’re always calling me up like. Like, one of them said, I want to take you to Deadpool, the Deadpool film on the weekend. I’m like, uh, I’ll just wait till it comes out on streaming.

[00:20:02] Steve: go out and have some popcorn with your boy, what’s wrong

[00:20:04] Cameron: hang out, hang out with your sons. I was like, oh yeah,

[00:20:07] Cameron: okay, good point, good point. Um, so I do think we’ll have AI, uh, friends that’ll be listening. I don’t think these devices are really going to be where it’s at, but I do think we’re going to have these AI friends, coaches, mentors, guides, uh, counselors that will be there. end up playing a really significant role in their lives. They’re not going to replace people, but they will be on top of that.

[00:20:34] Cameron: It’ll be like having, uh, the smartest person that, you know, just hanging out with you all day. Oh, here’s

[00:20:43] Steve: can also see a B2B context for these being valuable. You can see a context where you might have an AI pin that is hanging around your neck because you’re a Boeing engineer and it’s like, Hey, did you do the X? Don’t forget the Y. And then that way we’ll have less windows pop out of Boeing 737 MAXs and that’s the kind of plan I want to be on.

[00:21:04] Steve: I’ve always said that. But, yeah, different AI contexts of some of these pendants and AI friends that could be useful, like you say, coaching, training, uh, work context, technical skills. Which are, uh, a fragment, uh, but a, a, a far more knowledgeable AI within a certain slither. And you tend to see those categories

[00:21:28] Steve: split down in, in ways as well.

[00:21:29] Steve: So that, that could, there’s some potential

[00:21:31] Steve: there,

[00:21:33] Cameron: He’ll be like, Hey Cameron, do you really want to buy that box of Tim Tams that you just picked up at Coles? Really? You know, you’re trying to lose 10 kilos right now. Do you really want to buy that? Is that going to help with your black belt grading if you get a box of Tim Tams? Maybe, maybe think twice.

[00:21:46] Cameron: Alright, we’ve got a guest! Uh, it’s a long time since we’ve had a guest on the show, uh, but Martin Murray has been, uh, tweeting up a storm from his farm in, I’m gonna say, somewhere in New South Wales. I’m taking a guess, and he’s been talking recently about his, how he was testing ChatGPT on the farm, and I thought, wow, that’s something I know

[00:22:11] Cameron: nothing about.

[00:22:12] Cameron: We should get him on. Welcome to Futuristic, Martin.

[00:22:17] Martin: Yeah, G’day, it’s good to be here, and yeah, you are right, northwest New South Wales, just around Dallungra, which is sort of between Armidale and Inverrell if you, sorry, between Armidale and Moree if you, yeah, you know that area at all.

[00:22:33] Cameron: I have driven through there a number of times. Yeah. It’s a lovely part of,

[00:22:37] Steve: It’s a New England highlight.

[00:22:38] Cameron: And you are talking to us, you were

[00:22:40] Cameron: talking to us from the cab of your tractor, by the looks of it.

[00:22:44] Martin: Yeah, literally in the

[00:22:45] Martin: spray rig. So, um, yeah, hopefully I didn’t ruin it when I jumped on earlier, your audio quality, uh, but I’ve shut it down, but kept the time booster on. So hopefully this conversation holds

[00:22:56] Steve: I have to ask the mandatory, uh, AI question. Are you on a John Deere with the right, no. Right. To fix it if it breaks down. That’s, that’s my

[00:23:06] Steve: question,

[00:23:09] Martin: Uh, can I swap the camera around? Because I am, yeah, you are banging

[00:23:14] Steve: there you go. What do you have the right to repair? Because that, that was something that was really, had a head of steam a few years ago for quite a bit. I don’t know what’s happened on the right to repair. There may be that, that’s an interesting starting

[00:23:25] Steve: point for us before we get into some of the depths of.

[00:23:28] Martin: To be honest, it’s not much of a, um, yeah, an issue for me personally. Cause if I had the right to repair, I still lacked the

[00:23:35] Steve: Okay.

[00:23:36] Martin: So,

[00:23:36] Steve: Yeah. Yeah.

[00:23:38] Martin: um, But, I think you are right, yeah, I think, like, in terms of diagnostic tools, and, um, all of that sort of software side of things, uh, which, I mean, these things are pretty well just computers now, computers with an engine.

[00:23:52] Martin: Um, yeah, we still don’t have the right to repair that and access that diagnostic software.

[00:23:58] Steve: think that’s it. Wow. With every sort of machinery, I think if you buy something, you should buy it. I don’t care if it’s got software inside it. The Digital Millennium

[00:24:05] Steve: Copyright Act is an absolute disaster.

[00:24:07] Cameron: Well, let’s, let’s move on to AI on the, on the farm, Martin, tell everybody about how

[00:24:15] Cameron: you’ve been playing with it, your motivations and your discoveries.

[00:24:21] Martin: Yeah, so, I’ve been a listener of your podcast for a little while now. Um, Oscar Pierce got me onto it maybe six months ago or so, a bit of a shout out to him. And just from, yeah, talking to him and listening to your show, I just, I figured I’d give it a go. Um, yeah. I was coming up to the current cropping season, got my soil test back, and look, I’m an agronomist by trade, like, that’s my background, so, I know what I’m looking at, I just thought, well, it’s a rainy day, I don’t have much else to do for the fun of it, let’s load these soil test results into ChatGPT. And, um, see what recommendations it comes back with. And I’ve got to say, I couldn’t quite follow the math on it. Um, I don’t know if it was using different formulas that are used elsewhere in the world to what I’m used to, but. At the end of the day, the end result was pretty well the same with what I recommended, um, putting in, you know, this is our crop, this is our target yield, this is our target protein percentage, all those other, those other inputs that you put in yourself, and, um, yeah, it came out pretty well bang on, and it just sort of went from there.

[00:25:31] Martin: I’m like, well, okay, this is the crop, uh, can you give me an agronomy plan, you know, how much starter fertilizer, how much urea, when should I be planting it, Uh, what sowing rate should I be at? And, uh, you know, what, what herbicides and fungicides am I likely to be using in the season? And to be honest, it was, you know, it wasn’t 100%, but it was pretty damn close.

[00:25:55] Martin: Like, it was, um, just an interesting experiment and we sort of took it from there. Had it, uh, Taking pictures, uploading pictures from the cattle yards, asked it to identify the breeds of cow, we’ve got a couple of different breeds, and, um, you know, things that are obvious and easy, like the Angus, it was able to pick them out pretty well, and then other ones that I thought might, you know, might throw it a bit, not as common breed, uh, like, uh, there’s a breed called Speckled Park, which is sort of black, but it’s got all these little white speckles through it, it was able to pick that out and identify it. yeah, it was just, just an interesting experiment and I’ve done a few other things with it, but that’s basically where it’s

[00:26:36] Steve: It’s so interesting.

[00:26:38] Cameron: Didn’t you also

[00:26:38] Cameron: ask it for like, uh, dry hire rates for a tractor or something?

[00:26:44] Martin: Yes, I have done that with it.

[00:26:46] Martin: Um, and it was on the money. Yeah, I think I put that one up on Twitter there a little while ago. Um, over selling, I hired a tractor off the neighbor and neither of us really knew the rate. Um, So we asked ChatGPT and we asked Twitter and then we eventually found someone that does commercially lease out the same horsepower tractor and the rate ChatGPT came back with was the same rate he was on.

[00:27:14] Cameron: Amazing.

[00:27:15] Steve: It’s, you know, oh yeah, I was just going to say, what I love about your approach. Is to ask it’s something you know the answer to already, because what that does is it gives you context of the type of accuracy you can expect, and then it gives you confidence to do the next thing. Like I haven’t, I’ve never done that on stage, but the idea of saying, well, okay, what’s something you have a lot of knowledge on start playing with the tool on that, because what that does is without you worrying about the information you’re getting back.

[00:27:46] Steve: You’re developing a relationship of usage, which gives you the back and forth to see how it works and the type of questions you need to ask it to get an answer that you already know. And I think it’s a really intuitive way to do it. It’s actually quite smart. And I don’t think we ever really teach anyone that.

[00:28:01] Steve: It’s really insightful. And then after that, you’re going to go, well, if I can do it for that, here’s the piece that I don’t know that I might be able to use it for. Because as, as you would know, as a listener, Cameron and I are talking about Whatever AI can’t do today can probably do it tomorrow. So, sort of with that in mind, what are some of the sticking points?

[00:28:20] Steve: Within your business on the farm that you’d like quicker solutions on, or it takes a lot of time, whether it’s administrative or, or, you know, planning what you do out in the field or with

[00:28:30] Steve: your crops or you, you mostly crops or it’s just always cattle.

[00:28:35] Martin: Yeah, so our business is primarily cropping. We do have, we’ve got like 30 cows, that’s not much, they probably make up, I don’t know, depending on the prices each year, they’re probably somewhere between like 1 and 10 percent of the farm’s income. But yeah, look, for me, I think with AI, it’s, it’s not so much what it can do for me now that I can’t already do, it’s what’s going to be coming, and, um, like in terms of regulation, red tape, reporting, emissions reporting, all of these sort of things, like biodiversity, reporting, um, like, I mean at the moment this is, Things that we’ve got a vibe are coming, but we don’t know what exactly is coming, and how it’s going to impact us, but it looks like, you know, whatever happens, there’s going to be a lot of office work, a lot of reporting, a lot of collating of data, uh, you know, how much herbicide have you used, how much, um, diesel have you used.

[00:29:38] Martin: Uh, just, just all this sort of stuff, you know, we’ve, like, my diesel records, they’re in, you know, my OB, they’re in the accounting software, because you pay for that by the litre, um, but I don’t record how that goes out on the paddock, um, you know, maybe there’s a way in the future that, that AI will be able to take that diesel input and go, you know, you’ve made 10 passes this year in the spray rig over this, This is a typical number.

[00:30:07] Martin: Um, you know, you’ve made your one sowing pass, your harvest pass. These are also typical fuel numbers and, and work out from the total, you know, diesel that you bought, um, what you’ve actually used in each paddock. And likewise with, with biodiversity, I was just talking to someone about this, um, before the podcast, you know, putting microphones and things around you, around your farm, um, and going, you know, uh, recording the, the sound of the boat, like.

[00:30:36] Martin: What, what’s, what animals are, what birds are chirping, what, what frogs are croaking, all of that sort of stuff, and then having AI analyze that, and then you get a pretty good picture of, of what’s on farm, and, you know, come up with a bit of a conscription of how to manage for those species. And again, you know, as I said, it’s not something that’s an issue now, but I can see it being an issue in the next 5, 10, 20 years.

[00:31:00] Martin: Something that we’re going to have to be on top of as land managers.

[00:31:03] Steve: funny because the administrative burden on small business is really been increasing, GST was introduced all of a sudden, you know, you have to, the administration needs to do with product services coming in and out in GST, which gave rise to. Uh, SESS for accounting because it just becomes too complex to do.

[00:31:22] Steve: But I like that idea of the internet and, and AI actually, not just being a filing cabinet where you can pull stuff out of it, you put stuff in, it’s a brain. So you’ve got to upload your data, which you can then analyze and you can give it directives on, okay, we’ve got this data here, what we need to find out is emissions, X, Y, and Z.

[00:31:41] Steve: Here are the data, here are the data buses, here’s the GPS from the tractors I’ve driven around. Of course if John Deere give it to you, here’s uh, here’s all of the raw materials that I’ve bought, here’s the head of cattle, here’s the output, and it can calculate all of those elements there and reduce the administrative burden because we tend to be looking at AI right now as a creation tool, like go and create more of the work that I do, but just that underpinning of putting up data to reduce administrative burden, um, it’d probably be more accurate than what we deliver today, I would imagine, you know, in, in, in calculations.

[00:32:14] Steve: Simple one is With your car, it’s like how many kilometers do you do on your little log books? It’s going to do a better job than you would do that just by looking at GPSs and petrol and acceleration and all of

[00:32:26] Steve: that stuff.

[00:32:27] Martin: Yeah, I’d imagine so. If it, if it’s got access to that data, it would definitely have more attention to detail than I would. I mean, no one gets into agriculture to be, uh, You know, an Auditor. It’s, you know, we’re here to grow things, uh, and drive tractors, chase cows, do whatever you love. You’re not getting into the game because you want to audit yourself and do all this reporting.

[00:32:50] Cameron: The outside of, uh, administrative red tape stuff, what are the, what are the two or three biggest issues that a farmer such as yourself is having to deal with in terms

[00:33:03] Cameron: of running your business every year, Martin? Yeah, I think, I mean, there’s a

[00:33:08] Martin: Um, yeah, uh, I mean personally my biggest issue is interest rates and I don’t know if AI can help me with that. But, um, but there’s I mean there’s other things, like other cases, I haven’t used this, done this personally, I’ve spoken to a fella that has, because he knew the parts catalogue was online, he took a picture of a broken part, and you can spend hours on the phone to dealers, like machinery dealers and, you know, repairmen, that sort of thing, trying to find a part number.

[00:33:38] Martin: He told me that he was able to take a, um, picture of this part, tell it where to find the parts catalogue online. And, you know, the make, model, everything else of that particular machine, and it was able to pretty quickly get back to him with the correct part. Um, personally, like, without giving ChatGPT a lot of direction, just taking pictures of machines, uh, around the farm, and asking it, you know, what’s this, what’s that, it’s probably about as good as my four year old.

[00:34:09] Martin: Like it could tell that it was a John Deere tractor and it sort of got the, like the family. Right. If that makes sense. But it, um, like, you know, say like you say you gave it a picture of a, you know, a 2002 Toyota Hilux, it would tell you it’s a Toyota Hilux, but it wouldn’t be able to quite pick up on, you know, that whether it was an Sr SR five and, um.

[00:34:33] Martin: You know, the thing with John Deere, and this is the thing that I found really surprising, is the model number is printed on the bonnet, and same with my case tractor, and that it wasn’t able to pick up on that, but it was able to tell me that it was a John Deere or a case, and give me the approximate, you know, generation, but not have the details, so I haven’t been playing around with it too much in that spare parts machinery place, but That is a big headache to solve if it can solve that.

[00:35:00] Steve: should mention that because one of the things that I’ve been doing lately is when ChatGPT, OpenAI, the one that I use the most, and I probably after today’s chat should go to Mistral, can’t do something, but I ask it to write code so that it can do that. It says, sorry, I can’t do this. I don’t know what the parts number, I would say, can you write some code to understand all of the part numbers or, uh, tractor numbers, whatever they’re called, or John Deere tractors, uh, go onto this website.

[00:35:28] Steve: Here’s where you’re going to find it. Do a screen scrape and visuals and then add into your database and create your own little, uh, GPT on it. So that you can use it to scan things and it will know it thereafter. So one of the things that, and this is this recursion, I’ve been thinking a lot about recursion lately, which is the idea that the technology can create technology that the technology can’t already do.

[00:35:49] Steve: Because it’s a functional tech, it’s not just like a steam engine where you turn it on. It can create new things that it can’t do. So you go in and ask it to write a script, To be able to identify X number of Toyotas or John Deere tractors, because that information is available on the web. It just hasn’t put it into the database in a format, which makes it findable.

[00:36:09] Steve: So you direct it to create that script and then go and do it. And then you ask it to put that in from now on, whenever you ask that, it can refer to that inside. Uh, your, your database. So that’s, that’s kind of one of the things that I think, and this is what agent AI is pretty good at as well. Uh, one that you might want to try out, which is pretty good, Martin, is, uh, agent GPT.

[00:36:32] Steve: And that does things like that. So you set it an objective. You could say, give me all the parts,

[00:36:37] Steve: numbers,

[00:36:37] Steve: and pictures of X and we’ll go out and do it.

[00:36:40] Martin: Righto, I’m gonna have to give that a go.

[00:36:42] Martin: Yeah,

[00:36:43] Steve: to wish you’d never talked to be mucking around on, you’re going to have crop

[00:36:46] Steve: failure. You’re going to be mucking around with AI for the next two weeks.

[00:36:49] Martin: yeah, nothing’s gonna get done but I do have a um, I do know, I’ve got a, well I’ve just actually got the replacement part, so I know the part number, I know where you can find the catalogue, um, so I’ll have a play with it and see if it can um, yeah, if it can do that.

[00:37:08] Cameron: lot of work if you, if, You know, to fuck around with it and try and make it do things that it can’t do. I guess we’re sort of assuming that the sorts of things that it struggles with today, it won’t be struggling with a year from now or two years from now. But I’m, I’m really interested in how AI is going to revolutionize different industries and agriculture being one of those.

[00:37:32] Cameron: I’m wondering, you know, there’s a lot of You and I were chatting on Twitter over the last couple of days. I mentioned that, uh, an old mate of mine from my Microsoft days, I had lunch with him a couple of years ago and asked him what he’d been doing. And he told me about a project that he had done for a client where it was, I think it was putting RFID chips into, uh, sheep and, and lambs, ewes and lambs.

[00:37:58] Cameron: Cause they, for some reason, these farmers needed to know or wanted to know which ewes belong to which lambs or vice versa. And they couldn’t tell, but they were able to figure out that by measuring proximity of the amount of time that they spend close to each other using the RFID chips, they could take, they could figure out whose LAMs belong to which use.

[00:38:20] Cameron: So he was building RFID just to capture all that data and then code to figure it out at the end of the day, write reports. I’m just wondering about all of the potential data. That can be captured on a farm these days with RFID chips and, as you mentioned before, visual, audio, um, and if all of that, uh, could be captured and just fed into an AI system that could come up with ways to help you be more efficient.

[00:38:48] Cameron: on the farm, look for ways of, uh, I don’t know, using less fertilizer or increasing crop yields or working less hours. And that’s even before humanoid robots. Elon starts, uh, giving you a Tesla robot. Uh, what, what would

[00:39:06] Cameron: a couple of humanoid robots mean for you on a property like

[00:39:11] Martin: Yeah, look, if I could get humanoid robots, um, that are, you know, just as competent as a human, you know, they’re able to react the same as a human, operate the same as a human, that would be brilliant, I mean, that would free up a lot of my time to go play with these sort of things with AI and whatever other tools are going on.

[00:39:32] Martin: Um, I mean, at the moment, like, you know, we’ve got GPS driven tractors, like, obviously, the one I’m in, you know, drives in a straight line forward and back, but I still need to be in here to operate, other controls, other important factors, there are now, um, you know, there’s, there’s Robots like Swarm Farm do swarm bots.

[00:39:50] Martin: They’re completely autonomous, like spraying, well they’re mainly used for spraying, but you can use them for other things too, uh, spraying robots. But at the end of the day, like, all my machines are set up for people. So, you know, we live in a human built environment, so if we, uh, well, an environment that’s built for humans, so if we can fit humanoid robots in it, it It would mean say instead of replacing, you know, three or four different machines, I’ve now got one robot that can drive all those machines.

[00:40:20] Martin: Uh, that would be a huge cost saving there. And at the moment, like labor’s hard to get. Um, it’s getting harder to get. Uh, it probably makes more financial sense to have a robot.

[00:40:32] Steve: such a strong

[00:40:33] Martin: And going back to

[00:40:34] Steve: It’s so strong. Just on that, I was Presenting humanoid robots in my keynotes lately, I show them now, the figure one and some of the humanoids, which are really extraordinary and Jensen Huang, we discussed on the podcast before says that before the end of this decade, they’ll cost, you know, like a small car, which is who wouldn’t invest in one of those? People ask me, why would we want a humanoid robot when you can have a self drive car? And if you look back to the old self drive cars in sci fi movies, they had a robot driving the car. And your answer is, is really the one, is that we have a human shaped world, right? And if you have humanoid robots, then they can have all of the capacity and work in the existing infrastructure that we already have.

[00:41:14] Steve: The insight that one humanoid robot would be a better investment than buying four new

[00:41:19] Steve: tractors that are probably 200 grand each. I don’t know, I imagine they’re quite expensive. There you go! So, the idea that you have a humanoid robot that becomes like a smartphone, a general purpose computer, but it can go into the shed and organize feed and then drive it.

[00:41:39] Steve: The tractor and, and then work with the cattle and walk around on the farm with humanoid gumboots. Like, that is actually a strong reason for humanoid robots. And I think that the marrying up of AI and physicality is a lot like when we used to have, you know, Uh, cattle pulling, uh, things along on the farm to, to, you know, grade the soil or before we had the combine harvester, it’s the idea of putting the engine inside the horse and cart or the engine inside, uh, the grader or whatever that you had on, on the, on the farm.

[00:42:15] Steve: I feel like the humanoid moment is the intelligence inside the dexterous physical robot is going to be maybe even

[00:42:22] Steve: the biggest part of the AI revolution.

[00:42:25] Martin: And yeah, and as I said, if they are

[00:42:27] Martin: just, you know, if they are as capable as a person, uh, it’s a no brainer. Even if they’re, you know, two, three hundred, four hundred, even half a million dollars. Um, I mean, that wouldn’t suit my operation, but if you’re a larger operation, uh, half a million on a robot, you know, compared to wages,

[00:42:46] Steve: Well, you’re 24 hours a day too.

[00:42:47] Martin: know, the the cost of, exactly,

[00:42:51] Steve: You’ve got the humanoid robots, uh, out on the front doing work 24 hours a day and just swapping over their batteries. Like, you think that through, it actually doesn’t have to be that cheap for it to be quite functional in industrial and commercial contexts.

[00:43:06] Steve: It’s not so much a

[00:43:07] Steve: domestic context, but it’s certainly in industrial contexts.

[00:43:11] Martin: No, you’re not gonna have it making you coffee and

[00:43:13] Martin: doing the laundry, but um,

[00:43:15] Steve: Well,

[00:43:15] Steve: it’s hard. That camera and I were like, wow, actually, yeah,

[00:43:20] Steve: mowing the lawns, yes, and putting the washing out,

[00:43:25] Cameron: but if they’re, if they’re 20, 30, 50 grand, 50 I think people, uh, will have them doing domestic stuff too, um, some people will. All right, well, if there was one, let’s leave it with this, Martin. If there was one problem that you wish AI could solve for you in the next, say, two years, if there was one sort of killer app for AI, not robots, just AI, on a property or running a farm,

[00:43:58] Cameron: what would it be?

[00:43:59] Cameron: One of the things we’ve already mentioned or something else?

[00:44:02] Martin: Well, yeah, I don’t know. I mean,

[00:44:05] Martin: I guess it comes back to you don’t know what you don’t know, like what, what, what, what? What insights haven’t you come across yet? Um, like one of the spots I’ve been thinking about, you know, a lot, I know I said the livestock make up a very small part of our business and, like, just business wise, we’re actually looking to change that so they’re a bigger part.

[00:44:24] Martin: Um, it’s what can we, where can we use AI there? What, what, what data can we get? What insights, what patterns can we, um, You know, find that we’re not seeing, we’re not looking for, uh, we’re entering a world now where, you know, quite cheaply you can have, um, weigh bridges or like weigh bridges for cows in the paddock.

[00:44:46] Martin: There’s a great company called OptiWeigh. Um, we’ve got GPS cattle tags which, uh, you know, they’re still a bit expensive. Um, like every cow has already has an RFID tag. Um, that’s, that’s a legal requirement for supply chain traceability, all that sort of thing. It’s now becoming a thing in sheep. But GPS tags for about 50 a tag, you know, it’s not quite economic, but what sort of insights are we missing out there by not having it?

[00:45:14] Martin: Um, like there was a fella on Twitter, Nigel Caron, um, brilliant farmer in the Central West, he’s got an Optiway, so an in paddock. Type way system, the cows just hop on, they weigh themselves, hop off, it’s all a bit of a novelty for him. Um, he was able to pick up, you know, from that, that all of a sudden they stopped gaining weight and he didn’t know They didn’t know why.

[00:45:37] Martin: So they did some fecal tests and found that they, they now had a word prob worm problem. And so just from that quick insight, they were able to get on top of that problem. And, um, he said what it would’ve otherwise cost them in lost weight gain. They were able to save just in that, um, that one worming, that one event.

[00:45:58] Martin: And it’s things like that. So like, I’d be keen for, you know, once. Once AI hits a point where it’s insightful and not just, like, where I feel it is at the moment, just regurgitating what it can find on the internet, it’s just a, you know, an enhanced search engine, it’s what is it going to be able to tell me, so similar to the tags you mentioned earlier, learning, you know, what tags are in proximity, that way you can come back to, you know, Which animals are your better performing animals, and which cows did they come from, so then you can breed more from those cows and get rid of your lower performing cows.

[00:46:35] Martin: Like, there’s, there’s a whole world of possibilities, and Yeah, it’s, I don’t know where the answer’s going to be, because I don’t quite know what questions to be asking.

[00:46:45] Cameron: I Sorry for the bit of a rant, but that’s,

[00:46:47] Cameron: No, that was perfect. I thought you were going to say the cows were stopped putting on weight because they were looking

[00:46:52] Steve: that’s what I thought I’d

[00:46:53] Steve: do it for.

[00:46:53] Cameron: fuck, I weigh, I weigh how much? Holy shit. I got to go on a diet.

[00:46:58] Cameron: That’s like

[00:46:58] Cameron: me

[00:46:58] Cameron: jumping

[00:46:58] Steve: I’ve got to get off

[00:46:59] Steve: the grass.

[00:47:00] Cameron: Holy shit.

[00:47:01] Steve: That’s funny. I

[00:47:01] Steve: also like, here’s

[00:47:02] Martin: I know, we keep that information from them. We can’t have them knowing

[00:47:05] Steve: another one I like

[00:47:06] Steve: too too much

[00:47:07] Steve: McKinsey cow strategy. We get rid of the bottom 20 percent of performers every year just like

[00:47:11] Steve: McKinsey.

[00:47:14] Cameron: Here’s my suggestion.

[00:47:16] Martin: and that is how you run

[00:47:17] Martin: a business. That is how you run an

[00:47:18] Martin: agricultural business. That’s why we don’t

[00:47:20] Martin: have the same cows that we had, you know, 50 years ago. Like, the same type of cows. We don’t have the same breeds of wheat from 50 years ago. It’s That is how it works.

[00:47:31] Cameron: So my suggestion, Martin, if you haven’t already done this, this is something that I do with GPT all of the time is I will ask it to tell me, uh, ideas of how, what I could be doing. So I’ll throw a whole bunch of data at it. Like I do this with my diet, right? I track my diet every day using combination of Obsidian, my notes app, and some shortcuts and GPT.

[00:47:55] Cameron: I have it all in a spreadsheet going back sort of a year now. And I’ll throw it into GPT every few months and I’ll say this is what I’ve been eating, this is what my exercise looks like, this is how much weight I’ve been losing, um, what should I be doing differently? Give me, give me some suggestions about ways that I can modify my diet, modify my exercise, et cetera, et cetera.

[00:48:17] Cameron: With coding, I’ll say, here’s, here’s all of the tasks because I track all my tasks every day and how much time I spend on them. I’ll throw it into GPT every now and again and say, here’s all the things that I do during a day and how much time I spend on them. Have you got any suggestions for how I can be more productive?

[00:48:33] Cameron: What, what should I be doing that I’m not doing? What should I do more of? What should I do less of? And it’s not, it’s not, Like it’s not, uh, a productivity, an amazing productivity coach. A lot of what it gives you is sort of generic stuff still, but every now and again, it does give me a really good idea, really good insight.

[00:48:55] Cameron: And it only, I only need one every now and again to make it worthwhile for it to get me thinking about something that I wasn’t already thinking about. And that spurs me to go, Oh, well. Not that, but what if I did this, what would that mean? And then it’ll go, yeah, that’s a great idea. And sort of build on it.

[00:49:10] Cameron: So it’s using it as sort of a, I use it a lot as a brainstorming tool to start me off with ideas and then build on them and build on them. And, uh, you know, even using a little bit of, um, hold on, I gotta mute while I A little bit of, uh, Edward De Bono lateral thinking with it too. Yeah, you start a trend, you say, okay, now give me like 10 really out there random suggestions and, uh, see what that spurs.

[00:49:42] Cameron: Like, just give me a whole bunch of random words and add that into the idea mix and see what it generates. It’s really good at doing things that humans struggle with, which is thinking outside of the box. Um,

[00:49:55] Steve: ask it, you have to ask it that. And I often use it as well to generate, like I give me 30 ideas in this topic here, and I won’t like

[00:50:04] Steve: the ideas, but those ideas give me an idea of where I leverage my

[00:50:07] Steve: depth.

[00:50:08] Steve: That’s what always happens.

[00:50:09] Cameron: I’ll tell you

[00:50:10] Cameron: something. I was having a marketing meeting with my son Taylor the other day about how to use TikTok to promote my QAV, my investing podcast. And he said, well, you need to be talking, you need to figure out a way to make it relevant to whatever’s big at the moment. And I was like, oh, sure.

[00:50:24] Cameron: Well, what’s big at the moment? He said the Deadpool film. So I said to GPT. How can I use the Deadpool Wolverine film that’s big at the moment to talk about investing? And it said, well, you could talk about what kind of investors, Deadpool and Wolverine would be in real life. Deadpool would probably be a high risk, high reward.

[00:50:45] Cameron: And Wolverine, because he’s a mortal, probably takes a long view and you know, he could have, you know, a long-term view in investing. He doesn’t have to hurry. And I was like. Oh shit, that’s a great idea. You know, that’s, that’s brilliant. I never would have

[00:50:57] Steve: the other one is too, you can break down how much a Deadpool cost to make. What’s the ROI on the average film? Is it better than average? Why do films like, uh, uh, animated ones keep getting recycled because it’s got a bit of ROI and there’s lower risk and like, there’s a lot of different things you can do.

[00:51:13] Steve: That’s good advice. That’s

[00:51:15] Steve: really good.

[00:51:16] Steve: Yes,

[00:51:19] Cameron: Yeah, anyway. Well, Martin, we’ll let you go. You’ve got food to make, um, so I can eat it.

[00:51:25] Steve: we need it.

[00:51:26] Cameron: Thanks for coming on and sharing that with Martin. Really interesting

[00:51:31] Steve: Keep us up to date.

[00:51:32] Martin: Nah, that’s all good.

[00:51:34] Cameron: If you have any more ideas or breakthroughs with the tech as it evolves, let us know. I’ll follow you on Twitter. We’ll continue to follow you on Twitter.

[00:51:43] Cameron: What’s your Twitter ID for people

[00:51:44] Cameron: listening? It’s just

[00:51:46] Cameron: martinmurray__ag, yeah?

[00:51:48] Martin: That’s it, Martin Murray Ag. That’s the one.

[00:51:52] Cameron: So they can go and follow you on Twitter and, uh, yeah, keep up the good work, Martin. Come and come back on and tell us if you, uh, have

[00:51:57] Cameron: any more ideas about AI and Ag, tech and Ag.

[00:52:03] Martin: Righto. Will do. Thanks for having me on. It’s, it’s been great. And yeah, been really good to talk with you too. And I don’t know what I’m doing on the next rainy day. I’ll be playing with a few of those brainstorming ideas. Cheers.

[00:52:14] Cameron: great. Let us know how it goes. All right. Take care, man.

[00:52:19] Cameron: Fun.

[00:52:19] Steve: good.

[00:52:20] Cameron: Cool

[00:52:20] Cameron: dude. Alright, let’s get back into news stories. Steve, Chinese text to video. Now, we’ve talked a little bit about where China’s at with, um, AI. There’s been a lot of, uh, talk about how they’re, you know, quite a, quite a ways behind some of the US companies, but making a lot of progress. A company called Kling, K L I N G.

[00:52:47] Cameron: Out of, uh, China, launched their text to video model, uh, in the last week, and it is in, it is right up there with Sora and, um,

[00:53:04] Steve: Runway.

[00:53:05] Cameron: whatever the other Yeah, Runway. Look, Runway. yeah,

[00:53:09] Steve: thought it was as good as Runway’s previous model,

[00:53:14] Steve: but, but not better than Runway 3, the

[00:53:16] Steve: new one. And it’s, it’s, yeah, it’s videos, it gives you a maximum of five seconds at the moment, it’s not a lot, um, in the ones that, that I saw, the samples that it had, uh, I thought it was okay.

[00:53:29] Steve: I didn’t think it was better. Yeah. But, look, if they were nowhere, they’re certainly not far behind now.

[00:53:36] Cameron: It can do five seconds image to video, but it can do three minutes in length for videos.

[00:53:45] Steve: Image to video. I can’t correct it.

[00:53:48] Cameron: it’s, which is a long time for a video. If you go through the demos that they have on their website, um, video extension, Uh, yeah, I mean, they’re quite long. So, um, it’s, uh, look, it’s just amazing with all these tools coming in.

[00:54:07] Cameron: Now, this one’s publicly available too, unlike Sora, although you have to have a Chinese mobile phone number in order to be able to register, which makes it tricky,

[00:54:17] Steve: Makes it difficult if you’re

[00:54:18] Steve: uh, doing the English version, doesn’t it? Need a plus eight

[00:54:21] Steve: six.

[00:54:23] Cameron: And, of course, as we’ve talked about, there’s the, uh,

[00:54:27] Cameron: Chat XiPT, I think I heard it.

[00:54:33] Steve: Did someone say that? XiPT.

[00:54:37] Steve: That’s a Cameron political joke.

[00:54:40] Cameron: XiPT. The, um,

[00:54:43] Cameron: there is going to be this issue about how it, how their AI products navigate. You know, some of the Chinese censorship and restrictions on content. But at the end of the day, you know, we’re going to be in a world where I don’t really need my AI to comment on the Chinese Communist Party or, uh, or tell me about Tiananmen Square in 1989.

[00:55:10] Cameron: I mean, the range of things that I’ll want my AI to do on any given basis, any given day, I don’t think is really going to be a problem for Chinese AIs to generate. It’s going to be able to talk about science, and maths, history, Outside of maybe, you know, Chinese 20th century

[00:55:31] Steve: if it is true, if it is true that the AIs are trained on certain databases, And OpenAI and, uh, Gemini and Anthropic are all trained on the general web of English. You could even make the argument that the Chinese apps, if they have a proclivity, because they’re training on Chinese language to train on the internet, which is very censored in China, they could become more scientific AIs because there’s less dissent and trolling and misinformation.

[00:56:06] Steve: Potentially, uh, they could become more scientific with the output from those AIs than what you might get from the training on an English based AI, which is, as we know, being, uh, washed via social media, which has the highest frequency of posting.

[00:56:23] Steve: Lots of inaccurate information.

[00:56:26] Cameron: Yeah. And you know, something that, um, I’ve heard, uh, Zuckerberg say a lot, I’ve watched a lot of Zuckerberg interviews recently, cause they just came out with the latest

[00:56:38] Steve: How do you feel about Zuckerberg

[00:56:41] Steve: 2. 0? Let’s just, Let’s just, get into Zuckerberg 2. 0 with the curly hair, mate. The curls get the girls. And Zach is, he’s right up there with

[00:56:49] Steve: gents in one, what is, Leather Jacket coming out next? What are we talking here?

[00:56:55] Cameron: I love it. I love it. Um, yeah, he’s looking like in one interview, he’s looking like a surfer dude. He’s just got, you know, uh, you know, tan, beautiful, big curls. He’s looking like he just stepped off the snow fields or the beach. Next, next interview with Jensen Huang, he’s looks like a hip, he looks like Run DMC

[00:57:15] Steve: Ha ha

[00:57:16] Steve: ha! He’s just all in

[00:57:17] Cameron: black with a big gold chain.

[00:57:19] Steve: Jacket. Well, here’s the question, is he going to pull a Jeff Bezos and get on the gear and just get some big guns, and get rid of Priscilla, and, and, and, and just get

[00:57:29] Steve: himself a nice sexy newsreader? Because that’s what Uncle Jeff did.

[00:57:32] Cameron: yeah, like, whoever his PR

[00:57:34] Cameron: people are that’s doing his makeover, I love it. I think they’re doing a great job. But anyway, um, that aside, and he’s had a personality upgrade too, he actually can talk like a human a little bit

[00:57:44] Steve: the upgrade. So,

[00:57:45] Cameron: The AI, the AI upgrade that he’s had in his chip is doing well. But, getting back to, um, The point is, like what he’s saying, ’cause they, they just launched their, their latest LAMA 3.1, which is an open source and he’s a big supporter of open source and he for ai and he talks about why, and we’ll get to that in the later segment.

[00:58:09] Cameron: But what he’s, one of the arguments that he’s making is that he doesn’t believe the future is all about the one big overarching AI tool that does. Everything. He’s talking about AI agents that are smaller and more nimble and do, you know, one thing really, really well, that’ll be able to talk to other AI agents.

[00:58:31] Cameron: And this has been sort of the vision I’ve been talking about on, on the show ever since we’ve been doing it. That’s how I see it sort of playing out as well. I don’t think we’re going to require a source of all knowledge, like GPT is turning into the one thing that knows everything about everything. I think we will end up with millions of AI agents that have deep knowledge in a particular domain and they just talk to each other.

[00:58:59] Cameron: Your experience might be like you’re talking just to the AI, but it’s actually a million AIs talking to each other behind the scenes. And if

[00:59:07] Steve: Conceptually that’s interesting for a couple of reasons. I think the thing that is interesting to me is that the cache that it needs to carry is so much smaller if you have a specific AI about 8 bit video games from the 1980s. Does the big AI need that? Or does the big AI just need the ability to go to the AI and procure the information it requires for that moment and then trans Transport that information and you would have to assume that that would be one of the ways that we could become more efficient with the average search in terms of the computation it requires and the electricity.

[00:59:38] Steve: If it uses an internet style strategy where it finds the right path to what it needs and then comes back rather than having all of it to carry the entire information load within a one giant AI.

[00:59:51] Cameron: Yeah, like, we’ve been talking about the LUI, the Language User Interface, for a long time now and, you know, I see that is the, that’s the magic glue. That’s the, it’s like the new version of the API that enables.

[01:00:06] Cameron: AIs to talk to other AIs, because they all just talk the same language, doesn’t have to be English, but whatever the common

[01:00:12] Steve: I know what it is. Pull the

[01:00:13] Steve: plug.

[01:00:15] Cameron: they can, you know, so my AI on my phone, let’s

[01:00:18] Cameron: say it’s Apple, Apple’s AI, Or it’s just an open source one.

[01:00:23] Cameron: Cause a lot of the, you know, Microsoft has released a mini version now of their AI. There’s a, there’s a Mistral mini. There’s a Llama mini. There’s a lot of mini models that are out there. GPT’s got a mini model. If you have an open source mini model that runs on your devices and knows everything about you, And it can understand language.

[01:00:42] Cameron: It can understand what you ask it. It can then go out to, oh, you want to know a question about, uh, the latest in, I don’t know, uh, medical advancements for Alzheimer’s? It’ll go talk to the Alzheimer’s, uh, medical research

[01:00:59] Cameron: databases AI,

[01:01:01] Cameron: right? And it’ll find that information. It doesn’t need to know

[01:01:04] Cameron: everything. It’s the Henry Ford model. You know, the famous story about Henry Ford.

[01:01:08] Steve: Pick up the phone.

[01:01:11] Cameron: He was, for people who don’t know, we’ve probably mentioned on the show, but he was, uh, being sued. Somebody, a journalist had said he was stupid or something and he sued him for libel and then they were asking him a whole bunch of questions about his business that he couldn’t answer.

[01:01:25] Cameron: He said, but you don’t realize I got a phone on my desk. I don’t need to answer. I don’t need to know everything. I can just ring one of my managers. They know the answer to those questions. So it’s the same, it’s the Henry Ford model, right? The AI will just call whichever AI it needs to answer your question.

[01:01:39] Cameron: But my point was going to be. If, if one of those, or many of those AIs are Chinese AIs that are handling all of this for me, doesn’t really matter. As long as I’m not asking it questions that pertain to stuff that the CCP doesn’t want you to know about, or doesn’t want its citizens to know about, it’ll be able to answer 90 8 percent of everything that I need to know.

[01:02:04] Cameron: For those other 2%, I’ll go to the, you know, an American AI that’s happy to dish the dirt on the history of the

[01:02:12] Cameron: Chinese

[01:02:13] Steve: what will happen with the AIs. We know about the great firewall of China where Google and Facebook and some other properties are banned, uh, over there. I wonder if, and we’re quite open to having Chinese firms operate in Western markets. That’s a topic for another day, but I don’t think that, uh, America should be as open as like, well, you would let us do business in yours or, or we don’t, but.

[01:02:39] Steve: I do wonder if with the AIs, they’re going to end up with a closed shop as well, and whether they’re going to be interchanging with each other, or whether it’s going to be a little bit like what we have now with the social web and search, whether it’s going to be certain AIs in China are shut down, our AIs, but we’re open to theirs.

[01:03:00] Steve: I mean, that’s going to be an interesting political play as well as a technological play. I don’t know which way it’ll go. You’re well versed in this area. What do

[01:03:07] Steve: you think will happen?

[01:03:08] Cameron: I mean, I don’t know. I don’t know. I think AI is going to change a lot. I think the CCP is, and all Western countries too, are going to struggle with censorship. Like, let’s not pretend that we don’t have extreme levels of censorship in Western countries. We do, it just takes a different form. of, to the censorship in places like China, the sort of censorship that we have in the West around things that are politically sensitive is, they’re sort of over the window, doesn’t allow conversation of them.

[01:03:47] Cameron: They’re highly limited and we have massive amounts of propaganda.

[01:03:52] Cameron: We spend massive amounts of money and time and effort on propaganda to sort of, um, get our citizens thinking one way. Like if you talk to most people in the West, I’m talking about the USA, the UK, Canada, Australia, right? And you ask them why America dropped nuclear weapons on, dropped atomic bombs on Hiroshima and Nagasaki in August 1945, most people will probably tell you they did it to end the Pacific war against the Japanese in World War II.

[01:04:26] Cameron: It’s not true. That’s not why the Japanese ended the war. Every scholar who studied this knows that, but most people in the West still believe that to be true. And if you try and tell them different, they freak out because they’ve got 70 years of propaganda that’s been drilled into them that makes them believe a certain version of the story, right?

[01:04:50] Cameron: So we have, we have censorship, but it takes a different form. But AIs are going to, I think. Just crush all of that. Um, you, you know, it, it, it’s going to be very hard for governments to limit this kind of information, whether it’s a Western government or a Chinese government. So I, I don’t know how that’s going to play out.

[01:05:09] Cameron: Moving right along, because we’re running out of time. I want to talk about my, uh, I want to talk about, uh, Cory Doctorow’s article

[01:05:16] Cameron: on crypto. Do you want

[01:05:18] Cameron: to, do you want to kick this one off or

[01:05:20] Cameron: do

[01:05:20] Steve: was a, really great piece and the basic premise is that some of the richest people in Silicon Valley, uh, the billionaires behind, you know, Mark Andreessen and Ben Horowitz and others, uh, giving Trump endorsements For their own financial wellbeing at the cost of all of the other elements.

[01:05:41] Steve: It went pretty deep into, into crypto, uh, that crypto is, is really, I mean, I’m just trying to think of the way to describe it. Well, he obviously went through the stuff that says that crypto isn’t money because it doesn’t satisfy the needs. of what creates a currency and that it’s just a giant manipulation so that they can fund Trump so that they can get what they want and continue on their merry 1 percent ways.

[01:06:11] Steve: That was my

[01:06:12] Steve: key, that was my net outtake.

[01:06:14] Cameron: Yeah,

[01:06:14] Steve: was a pretty deep article, went on a lot of angles, but that was my

[01:06:17] Steve: overriding outtake.

[01:06:19] Cameron: You know, you know, I think he’s, he’s saying that basically crypto is a

[01:06:24] Steve: Yeah, there you go. I mean, there’s the words. Crypto’s a scam. Yeah, keep going. Tell me, Tell

[01:06:30] Steve: me what

[01:06:31] Steve: you think.

[01:06:33] Cameron: Well, you know, he talks about how, uh, crypto, you know, has been hyped up by, uh, a bunch of people, obviously, over the last ten years, and they’ve, you know, uh, uh, scammed a lot of people, getting them to invest in it,

[01:06:52] Steve: And the worst coins around it. Not

[01:06:54] Steve: so much it, but all of the shit coins that live around the fringes of it.

[01:06:58] Cameron: yeah. You know, there’s been a whole bunch of scams like NFTs that a lot of people got scammed on, different, um, ICOs, the different coin offerings that people invest in that go nowhere, and there’s been an influx of crypto money into elections. Election campaigns and it’s hard to track and it’s hard to monitor.

[01:07:21] Cameron: There’s a lot of dirty money flowing around in that sense. But, you know, the crypto, the major crypto players are trying to influence politicians in the US to get favorable treatment for crypto, to try and keep it out of regulation, uh, try and help legitimize it so they can profit from all of the investments that these, some of these venture capitalists have been making it for a very long time.

[01:07:52] Cameron: He talks about the Bezel Concept. Uh, which is sort of, uh, deceptive, uh, prosperity, uh, which is really based on a fraud. He calls it the The bezel, like M bezel, but just the bezel where you make something look really attractive, sort of a, it’s kind of like a, uh, another name for like a pump and dump really, you, you, you use your influence in the media.

[01:08:20] Cameron: And when it comes to these Silicon Valley billionaires, They’re followership on Twitter, their influence over tech media, their ability to get stories written about the things that they’re investing in and create hype cycles around it when it’s all just based on sort of smoke and mirrors, usually. He’s talking about the need for regulatory clarity in companies.

[01:08:44] Cameron: Crypto and the challenges around that. Illegal contributions by companies like Coinbase, uh, towards campaign finance, um, and, and how that needs to be monitored a lot more closely. And just about the complex interplay between economic interests, political interests, the ethical considerations around crypto.

[01:09:04] Cameron: Anyway, it’s a really great article, uh, it’s called, what’s it called? Um,

[01:09:11] Steve: campaign finance

[01:09:12] Steve: violation in US history.

[01:09:15] Cameron: That’s it Yeah, check it out on Cory Doctorow. Really worth a read. I mean, crypto is, you know, I look, I say what I’ve always said. I think there’s a lot of potential in the, the technology behind crypto, but, um, I think all of the claims that are being made about it as being, you know, it’s going to be a legitimate

[01:09:38] Steve: Never will be. It’s been poisoned. It’s the, the, the lake is poison and my view on, on it is really simple. Technology is a currency and, and all currency is a technology and all currencies have represented the technological capabilities of the day and the overriding tools that we use. You go all the way back to Ferris coins and.

[01:10:00] Steve: Uh, Cowry, Shells, to Fiat Currency, Bills of Exchange, all of those types of currencies, uh, precious metals, have all represented our technological capability. And they all have certain requirements for them to be effective. Uh, one of them is stability, which crypto doesn’t have. Uh, they need to have acceptance, which crypto doesn’t have.

[01:10:19] Steve: So it fails on about four, four of the six requirements, uh, to be a currency. It will never happen because Any country that is smart enough needs to know that it needs to have a sovereign currency so that you can demand tax in that currency, which maintains the civility and your ability to fund government projects.

[01:10:42] Steve: But here’s the thing that was so interesting in that article is they’re using the crypto scam to get fiat currency in replace of the crypto that they own. Like, and if you ask any, any, uh, Bitcoin maxi, And if you want to truly understand whether or not they believe in Bitcoin, here’s what you ask them.

[01:11:00] Steve: You ask them this one simple question, Cameron, you say, what’s one Bitcoin worth? And invariably, their answer is in US dollars. And that tells you what they believe in. They sure as hell don’t believe in crypto. Because if they did, and you ask them, what’s one Bitcoin worth? They’d say one Bitcoin is worth one Bitcoin

[01:11:19] Cameron: One Bitcoin. Yeah, yeah. Alright, well, let’s finish up, Steve, because I know you’ve got a hard out. Um, I want to talk about closed systems versus open systems, uh, and again, because I’ve believed this for a long time, and I’ve heard Zuckerberg talk about it a lot lately. Now, let me preface this by saying, I don’t trust Zuckerberg for one second, even though I like his rebranding.

[01:11:43] Steve: See? See? We always be buying into the story, don’t we? All of a sudden, Oh, you know, I don’t like you, I don’t like Zuckerberg, but he looks cool. You know, five episodes from now, you’re singing the praises of Uncle Zuck. That’s all I’m saying. Listeners, watch out for Cameron. He’s been caught off guard.

[01:12:02] Cameron: And to be fair, he’s pretty upfront in most of the interviews I’ve seen with him lately, saying, Listen, I’ve got selfish reasons for doing open source AI. Um, and his thing is, his experience has been that Facebook has really struggled, particularly in the mobile web, in the last 15 years, because the mobile web has largely been controlled by Apple, as we know.

[01:12:28] Cameron: And Apple has the Apple tax and they will allow certain things to happen on their devices and they won’t allow other things. And he feels like they’ve been hampered by Apple’s control over a closed ecosystem. He doesn’t want that to happen again with ai. And he can see that, uh, you know, if, if it’s, if chat GPT open AI is the future of ai, uh, that he could end up in the same situation.

[01:13:00] Cameron: So he’s spending a lot of money. I mean, I, I saw a, a recent interview of with him and, uh, done by Jensen Huang, CEO of Nvidia, where I think Jensen was saying that. Men now have like 600,000 H one hundreds, uh, Nvidia

[01:13:19] Steve: valued at 10 grand each, aren’t they?

[01:13:20] Steve: Got

[01:13:22] Cameron: I’m not sure. 10, something like that. 50, maybe a hundred. I’m not sure which, how much the H one hundreds are worth, but it’s a lot.

[01:13:29] Cameron: So they’re spending billions and they’re making an open source product. Now, one of the things that, you know, often hear people who are not paying close attention to this about AI is, Oh, all the rich people are going to have AI and it’ll be controlled by a handful of companies like, uh, you know, Microsoft, Google, OpenAI, whatever.

[01:13:49] Cameron: And it’ll just be the mobile web or the web all over again. You know, people like us that are old enough to remember the mid 90s. We did think that the web was going to be this democratic place where anyone would be able to do anything. And that is

[01:14:03] Cameron: true. No, well, it is true. I can build a website. I can write whatever I want on my website.

[01:14:11] Cameron: Uh,

[01:14:12] Steve: You could have,

[01:14:13] Cameron: Just but you could, you could also stand on a soapbox on the corner and just

[01:14:16] Steve: scream to the top and you’d be just as effective as building your own website, right?

[01:14:20] Cameron: Well, that’s true. But I couldn’t, I couldn’t really

[01:14:22] Cameron: print my own newspaper, um, and, you know, make it accessible to You’ve got a people or billions

[01:14:28] Steve: still slim, But you’ve got a chance to break

[01:14:30] Steve: through in the system and be

[01:14:32] Steve: known.

[01:14:34] Cameron: But companies spent hundreds and hundreds of billions of dollars to basically make sure that audiences tended to go to their properties and, and not anywhere else. They, they sort of dominated the space. And people, uh, you know, are reasonably concerned that the same thing’s going to happen with AI. But I don’t think it is.

[01:14:58] Cameron: And I do think open source is going to be big. You’ve got companies like Mistral. You’ve got companies like Meta. We don’t really know how the Chinese AI companies are going to play out, but they’re going to be players in this space as well. You are going to have a proliferation. And, and Zuck’s vision is that, Hey, it’s funny because he uses Microsoft as an example of open source.

[01:15:20] Cameron: I never really thought of Microsoft as open source, but you know, when I was at Microsoft 20 years ago, the battle was between Microsoft and Linux, Linux being the open source alternative to Windows. And the Microsoft ecosystem. But Zuck’s view is that, yes, but anyone could write an application that would run on Windows, and Microsoft wouldn’t stop you from having an application that ran on Windows.

[01:15:48] Cameron: I mean, Marc Andreessen Might have something to say about that because we made it difficult for Netscape to run on Windows in the late 90s. You could, for all intents and purposes, unless you were in Microsoft’s crosshairs like Netscape were, because it was an existential threat, you could write an application and you could sell that application and Microsoft wouldn’t put a tax on you for selling your Windows application.

[01:16:15] Cameron: It didn’t have a marketplace that you had to get permission

[01:16:18] Cameron: to get your thing in. You could distribute it.

[01:16:22] Cameron: So it was open source in that, it was an open

[01:16:25] Cameron: system,

[01:16:26] Steve: you could get your, it wasn’t an five inch disc and upload that piece of software to

[01:16:30] Steve: run.

[01:16:32] Cameron: Apple’s is a closed system, and for good reason, Apple did it to try and prevent hackers and malware and all those sorts of things, and they go, okay, well we have to manage this and that. If that comes with a fee, it comes with costs, and you have to share your revenue with us because we’re managing this system.

[01:16:50] Cameron: So it’s a safe and a secure environment for users. I’ve never had a virus on my iPhone or my iPad. Never had to worry about it. Don’t, you know, I don’t really give a much thought about it, but he’s talking about a world where there are billions of AI agents that are built on top of open source AI models that.

[01:17:15] Cameron: He is building, along with others, he’s gonna build them, he’s gonna make them available, anyone can take them, you know, check the security or improve upon them, make them bigger, make them smaller, dedicate them to subject X, Y, or Z, they can be hosted on a, on AWS or on a Google data center or on your own data center, or you can run them with a small footprint on your phone or your iPad or one day your watch or whatever it is.

[01:17:47] Cameron: So Meta are putting God knows how much money, billions and billions, and probably hundreds of billions over the next few years, into a future where we have extremely powerful AI tools that are open source. And I think that is gonna really be a hugely

[01:18:11] Steve: does he leverage that? Does he end up renting out infrastructure like an AWS, but it’s, uh, it becomes, you, you, you subscribe

[01:18:19] Steve: to the meta AI engine to create whatever you need,

[01:18:22] Cameron: I think there’ll be revenue opportunities for them in doing that and probably providing hosted opportunities for businesses that don’t want to do it themselves. They’ll have revenue streams in there, but I don’t think he really knows. The sense that I get from the interviews with him right now, either he doesn’t know yet or he’s not revealing it yet.

[01:18:45] Cameron: He probably has ideas, but it’s just right now, the battle is for. Getting yeah. having

[01:18:52] Cameron: the tools available

[01:18:54] Steve: Have it, have the powerful tool, and then work out the revenue stream ladder. It’s a little bit like 1999, when everything was free, and it’s like, well, we’ve just got to build it, get people using it, and then we’ll get the revenue model. Which both Google did, Facebook did, all of them found their revenue models.

[01:19:10] Steve: Albeit, they went to the horrible advertising model, which is You know, the, the original sin of the

[01:19:16] Steve: internet, you know, don’t like it, but, uh,

[01:19:19] Cameron: but yeah, so imagine a world where we

[01:19:21] Cameron: have, uh,

[01:19:23] Cameron: open source AI, millions of people working on it, millions of AI agents working on improving it. And then when we figure out nanotech, we have millions of open source AI agents building the plans for open source nanoreplicators, nanofabricators, giving us the plans for open source robots.

[01:19:48] Cameron: I’m reading this book and I’ll wrap up with this because I know you’re going to go, um, Radical Abundance. Hey, Eric Drexler’s book from about 10 years ago. He’s the father of nanotechnology. 1986, he wrote.

[01:20:00] Steve: regime. Radical

[01:20:01] Steve: abundance.

[01:20:02] Steve: Got it.

[01:20:04] Cameron: Yeah, in 1986, he wrote

[01:20:05] Cameron: Engines of Creation, which was sort of the first book on nanotech, and, uh, his last one, which is 2013, Radical Abundance, he has this model, which I thought was fascinating. He said, I want to, I want you to imagine how a car is going to get built in the future. It’s being built in something the size of, uh, Household garage.

[01:20:24] Cameron: And if you look at the garage, you see some platforms that go up and down and a whole bunch of robotic arms that are assembling things like you would see in any sort of car factory, but the back wall. is made up of shelves that have devices about the, they’re boxes, they’re like the size of microwaves, some bigger ones about the size of a washing machine, some smaller ones the size of a household microwave, all up the wall.

[01:20:50] Cameron: And if you look inside of those, they have smaller versions of the garage with smaller robot arms that are assembling components. When they assemble their components, they get pushed out, of a door in the machine where the bigger arms pick them up to assemble them together. But inside the microwave in the washing machine, there’s a smaller box with smaller robot arms that are making smaller

[01:21:11] Cameron: parts.

[01:21:12] Steve: It’s like the, the entire thing of fractals. I’ve been really watching a lot about fractals lately and all manners of biology have certain patterns which are unavoidable. I mean, if you look at. The image of a brain and a satellite map of America or anything, it looks like a brain, you know, where all the cities are and the neurons and the lights and where things are flashing, which, where there’s more activity, which is, uh, the same as in the brain where there’s nodes and neurons going through to an MRI scan.

[01:21:41] Steve: It looks really similar. Large language models look really similar to a brain. And the idea of, you know, Fractals, uh, the way we create stuff very, very similar all the way through. And it’s almost like we have this inevitability that we will replicate nature in the way things are built, even at the nanoscale and then nanoscale all the way up to manufacturing a car.

[01:22:02] Steve: It seems like there’s this fractal orientation. I don’t know if he has that in the book, but it,

[01:22:05] Steve: it, it seems really interesting.

[01:22:07] Cameron: He doesn’t, he doesn’t mention fractals, but that’s great. I was thinking of matryoshka dolls, but fractals is a better, more exotic explanation. He says, To the right, picture yourself standing outside the final assembly chamber of a large product APM system and looking in through a window to view the machines at work in a space the size of a one car garage.

[01:22:27] Cameron: To the right, you see an exit door for products ready for delivery. To the left, you see what look like wall to wall, floor to ceiling shelves. Which, with each shelf partitioned to make a row of box shaped chambers. In the middle of the garage sized chamber in front of you is a moveable lift, surrounded by a set of machines.

[01:22:48] Cameron: And so, basically, the, the The little microwaves have little ones, have little ones, right down to molecular components that it’s building from scratch based on, you know, I have some inputs, which is, you know, bottles of nitrogen, oxygen, hydrogen, carbon, whatever it is, that probably it’s getting from deconstructing last week’s car model that you got bored with.

[01:23:12] Cameron: So it’s building you a new one for today. And it, it builds the molecular components and feeds them into the next box, which builds, you know, adds them together, which fills in the next box, which adds them together, assembles

[01:23:22] Cameron: them all the way You’d just scoop up some dirt and get everything you need.

[01:23:27] Cameron: Oh, well, you need to sort the bits from the

[01:23:30] Steve: Robots are going to do that. Robots will do everything.

[01:23:32] Steve: yeah.

[01:23:34] Steve: because

[01:23:36] Cameron: So that’s his view for this, but you know, I believe that we have a chance. I, I, I don’t know what probability was assigned to this, but we have a chance where we will have open source AI, open source robots, open source nanotech.

[01:23:50] Cameron: And that will lead to sort of this, um, version of the world and a version of an economy I don’t think we can really envision right now.

[01:24:02] Steve: it removes the scarcity model, which all of the, you know, whether it’s capitalism or communism, they’re all based on the fundamental principle that there’s scarce

[01:24:10] Steve: resources. And this kind of circumvents that to an extent.

[01:24:16] Steve: There will always be some form of scarcity, but the scarcity, there might be different types of scarcity, and I don’t know what they are.

[01:24:22] Steve: You know, scarcity of attention, scarcity of fame, scarcity

[01:24:25] Steve: of recognition. I don’t know.

[01:24:27] Cameron: Kurzweil in The Singularity is Getting Nearer

[01:24:29] Steve: I just the law of accelerating returns.

[01:24:32] Steve: Yeah,

[01:24:33] Cameron: He talks about how the IT

[01:24:34] Cameron: industry, the computing industry, has built on the law of accelerating returns over the last 100 years because each generation of technology helps you build the next generation of

[01:24:43] Cameron: technology, but that most other fields don’t benefit from the law of accelerating returns.

[01:24:50] Cameron: You know, this, this year’s, um, crop of wheat isn’t going to make next year’s crop of wheat exponentially better. But he said that as these sorts of, this is the, I’ll read this one paragraph and I’ll let you go. What makes the LOAR so powerful for information technologies is that feedback loops keep the costs of innovation lower.

[01:25:12] Cameron: And as artificial intelligence gains applicability to more and more fields, the exponential trends that are now familiar in computing will start to become invisible in areas like medicine, where progress was previously very slow and expensive. With AI rapidly expanding its breadth and capability during the 2020s, this will radically transform areas we do not normally consider to be information technology, such as food, clothing, housing, and even land use.

[01:25:42] Cameron: We are now approaching the steep slope of these exponential curves. That, in short, is why most aspects of life will be getting exponentially better in the coming

[01:25:52] Steve: that it’s the idea that everything becomes an information technology business. And once we started to use computation in many different industries, we got this horizontalization where things go across. And it actually is Easier to see that that becomes possible. And especially once you start to organize things at the nano and atomic level, then everything becomes organizing things at an atomic level.

[01:26:15] Steve: So then all of those industries have that positive feedback loop and the AI just becomes the scaffold to use a term that we’ve already used today that creates that possibility.

[01:26:26] Cameron: That is Futuristic, Episode 29. Thank you, Steve. And thank you, Martin Murray, for coming on. That was fascinating. Really appreciate your time,

[01:26:33] Steve: Thanks so much, Cam.