AIForHumans113OpenAISuperIntelligence

Kevin Pereira: [00:00:00] AI is not slowing down. In fact, open AI's. Sam Altman peeled back the fog of war to give us his vision of super intelligence, which is coming soon. And Ilya Suski tells us what he expects in the near future.

Ilya Sutskever: Slowly but surely, or maybe not so slowly, AI will keep getting better. And the day will come when AI will do all of our, all the things that we can do.

Gavin Purcell: I'm actually already here, Kevin. I'm visiting from the future. They're all wrong.

Kevin Pereira: But did we, we talked about this last week. Are we really doing this, Gavin? Yes.

Gavin Purcell: Kevin. We're really doing this. Come on.

Kevin Pereira: Okay. I. All right, fine. Go ahead. Do the bit.

Gavin Purcell: The world isn't ready for the changes that are coming. Kevin oh three Pro was just released in your timeline, and the impact is enormous.

Kevin Pereira: Okay. Future Gavin, what about WW DC 25? Did Apple's AI failure there? Have I? Long lasting repercussions.

Gavin Purcell: Actually, Kevin, by 2027, everything is liquid glass.

Kevin Pereira: Oh, I didn't think that would catch on, but okay. But

Gavin Purcell: guess what else, Kevin, this week, June of 2025 that [00:01:00] I have traveled from the future back to one X Robotics, introduced us to its redwood platform and 11 Labs drops V three of their speech model and it changes.

Kevin Pereira: Oh, I get that future, Gavin. Yeah, a lot changes. Also, midjourney is getting sued by Disney and Comcast this week that just happened. We'll tell you what is going on and how it could affect you. Kevin,

Gavin Purcell: it's up to us to change the future. Is it really future Gavin, or is this an excuse for you to go beast mode in an intro?

Technically, Kevin. This isn't Beast mode. I just want to be clear with you and the viewers out there. This is not Beast mode. This is pretty close. Gavin, this is AI For Humans of the future,

Kevin Pereira: at least, if you're gonna say of the future, you gotta at least echo the word future. The show. Start the show of the

Gavin Purcell: Kevin. It is another. Huge week in the world of AI for humans. We are sitting in a hotel, but we are gonna [00:02:00] get right into this. Uh, we are gonna talk about super intelligence because surprisingly, a lot of companies have been starting to talk about it. And I wanna jump into this first with our old friend, Ilya Ver, who was a founding member of OpenAI.

Left the company to go. Found super, what is it? Wait, what is it called? Oh, he went super safe. Safe intelligence. Super ind intelligence. Super intelligence, super duper safe intelligence. He came down from his mountain where he's been to give a a speech at the University of Toronto's graduation ceremony where he went, let's listen to that real quick.

Kevin Pereira: Oh, what I would give to be a fly on the wall in that mountain. Gavin just ilia me. Probably John Carmac. I bet Carmac is there. Just play the clip. Maybe just play the clip please. Kam Lucky. Play the with hockey sack. Play the clip please. We put hand drum. Play the clip, please.

Ilya Sutskever: Okay. Maybe not so slowly. AI will keep getting better and the day will come when AI will do all of our, all the things that we can do.

Not just some of them, but all of them. Anything which I can learn. Anything which any, any one of you can learn, the AI could do as well. [00:03:00] How do, uh, uh, how do we

Kevin Pereira: know that?

Ilya Sutskever: By

Kevin Pereira: the

Ilya Sutskever: way, we know this by the way. How can I be so sure? How can I be so sure of that

Kevin Pereira: lib?

Ilya Sutskever: The reason is that all of us have a brain, and the brain is a biological computer.

That's why we have a brain. The brain is a biological computer, so why can't a digital computer, a digital brain, do the same things?

Gavin Purcell: So let's talk about this. So Ilia, Sr. Is somebody that is you, many people would consider, is on the very forefront of where this stuff is going. There's a lot of debate around this clip, uh, this week because there are a lot of people who believe that the brain is more than a, a human computer, right?

There's people who believe in a soul, people who believe in consciousness, all these things. But I think this kind of starts off our conversation around the idea of what super intelligence is and why suddenly, Kevin, this week, lots of people are starting to talk about it. So.

Kevin Pereira: Gavin, let me chime in with something actually useful for once I know what super intelligence is.

We got a glimpse of it this week, in fact, when [00:04:00] Apple talked about it.

Apple PR Person: Oh, roll the clip. Me.

Kevin Pereira: Okay, you

Apple PR Person: now, in addition to turning a text description into a Gen Moji, you can mix together two emoji to create something new like a sloth and a light bulb. For that moment when you're the last one in the group chat to get the joke.

Gavin Purcell: I know. Who else is the last one in the group chat to get the joke? It might be Apple. It might be Apple, which we're gonna get to in a little bit. We're gonna talk about what apples do is Yeah,

Kevin Pereira: they don't, they are the joke right now. I think. Yes. You should might as well jam an apple into the, uh, sloth emoji.

Um, for those that are getting the auto only. Yes. That is super intelligence that is taking two. Different emojis and jamming them together to create something new, something magical. But, but I don't want to like, because of the gag, get away from what Ilya said we are, uh, in in his, in his mind. Yeah, in his mind, his brain computer.

That's what we are, right? We're a biological computer interface. AI is just a digital version of that. There's no reason to think, you know, from his speech that there's gonna be any difference and that they won't be capable of thinking, do you buy that?

Gavin Purcell: So here's the [00:05:00] thing, I think humans have been tied together.

By belief systems forever, and we have taught ourselves a lot of things and there is a lot of conversations around what we believe about ourselves, what we believe about the greater universe, what we believe about a superior being that may or may not have created us. Many people believe this way. I do believe that we are mostly human computers and whether or not these ai, in fact, they are not operating the exact same way.

I do believe that science will figure out different ways of how our brains work. That they are essentially to your, to your words, the meat computer, right? The meat computer, and that we are going to get closer and closer to intelligence that is close to us. Now, we are not there yet, but this is what these conversations about super intelligence really are and, and the super intelligence part of it.

Is there, we talk about this all the time on the show. There is the term a GI, which is artificial general intelligence, which means the idea that a computer brain is as smart as your average human being and it can do all the [00:06:00] things that a human being can do, which it can't do yet. I. And then there is artificial super intelligence, which is where it blows right past us and gets to be much smarter than us.

And even if you had an average human who had the ability and the context of what a computer can have, that alone is super intelligence in some form. So that is what all these people are starting to talk about now. I. We are gonna talk all about ww DC in just a bit. Uh, WWC was a kind of a fail for Apple and the world of ai, but we don't wanna dive into that yet.

'cause we're gonna talk about Sam Altman's blog post, which again, we love talking about Sam Altman's blog post about super

Kevin Pereira: intelligence. Tell me it's going to be like. A mint on a pillow. Gavin, tell me it's going to be like the whisper of the wind through the tree leaves that it's going to be a toto washlet that massages and warms at the same time.

I need my singularity to be Well, Kevin, you'll be happy to, you'll be happy to hear.

Gavin Purcell: He called it The gentle singularity, which we will debate is whether or not that is a real thing. And again, another technical term for those of you who may be new or don't know this, the singularity is the moment where humans and [00:07:00] machines somehow merge.

Ray Kurzweil has been talking about this for about 20 plus years. His timeline is about right for where we would achieve this. This is where we as humans might decide to become one with a machine because our brain would get better and better with them. And

Kevin Pereira: also, by the way, on the other side of that point of the singularity human beings, us having this discussion, everybody listening or watching to it, we could no longer predict what the future will bring.

Yes. Because of that moment, that of the singularity. That's right.

Gavin Purcell: So this blog post goes into very deep detail around what Sam Altman sees as the future of this world. And. I just wanna point out very quickly before we dive into some of the specifics that he says is this is part of a kind of pattern that we're seeing amongst the high level AI companies talking about this idea, meaning that they are all trying to lay a path, whether it is real or not, that we are coming to a place where this is possible and may happen soon.

So Sam's particular vision. Is very specific in that he talks about the idea that already at OpenAI [00:08:00] they are seeing the, what he referred to, I think as the beginning stages, the kind of proto stages of self-improvement, which is a big deal. And we know already that Google, based on Alpha Evolve, is seeing self-improvement, that it means that these AI systems are able to help write and improve their own code.

But more than that, Kevin, I think this is getting at a, a much larger idea. Which is, what do human beings mean in this world and what does it mean going forward? I, I, one of the things I thought I really wanted to hear your take on is that Sam points out that a thousand years ago in this blog post, that if you had told humans then what our jobs would be, now, they would think that those jobs were all make believe jobs that like, there's no way that somebody wouldn't be able to like have to work the soil or have to do the sorts of things.

And he's saying in a thousand years from now, the jobs that we're doing, maybe this. Our make believe jobs. And you and I might think that these are make believe jobs now because they are, but what is your take on that? Do you think that that's what we're gonna see? I wanna be a

Kevin Pereira: hologram farmer. Like that's what I, I wanna do social media for a Roomba, like I wanna, I don't [00:09:00] know what those future jobs are, but you know, the other thing he points out is, but by the way, yes, of course.

Even if you told, uh. 1994, Kevin, that yeah, we would be doing what we're doing right now. Even though I was kind of doing a podcast on the internet, the ecosystem would look completely different. Yeah. The tech stack would be unimaginable. The, the fact that someone is watching this probably in a bathroom, probably on that to Washlet in the palm of their hand with a foldable oh LED smartphone connected to high speed wifi, like even that alone would've seen science fiction.

So yes, of course, even I think. You know, a, a decade from now, things are gonna be wildly different in certain spaces. But he mentions that, you know, in 2035, it's hard to imagine what will have discovered. And he, and then he pontificates, maybe we'll go from solving high energy physics one year to, uh, beginning space colonization the next year, year, the next.

Yes. Yeah. Or from major material science breakthrough one year to true high bandwidth brain computer interfaces. The next year. This is, you [00:10:00] know, 2035. This isn't 2053. This is around the corner.

Gavin Purcell: Yeah, and I think one thing he says very clearly towards the end of this, he says that open AI is a lot of things now, but before anything else, we are a super intelligence.

Research company. So this is their goal. Their goal is to get to super intelligence, and we know this is also the goal of De Saba and Google D Mind. Mm-hmm. We know this is also the goal of many other people, including Ilya Ver there's a lot of people racing to this. And I, I guess one I do wanna point out one quick story that I saw that kind of didn't get covered enough.

Uh, scientific American had a really interesting article that will link in the show notes, which is basically talking about a bunch of very well regarded and famous mathematicians getting together. And they are trying to figure out whether or not they can still outsmart the AI and math and coding may be solved sooner because they're more logic based.

But these are like some of the best minds in the country who are gobsmacked by what the models can do right now. So you may hear this and think hype. You may hear this and [00:11:00] think they're talking their own bags, which clearly they are. Um mm-hmm. OpenAI also just. Hit a a $10 billion a RR, which if, you know, the parlance of this world is the annual recurring revenue that is up from 2.5 billion last year.

They have projected, projected as a good number for 125 billion a RR by like 20, 29. These companies are talking this way now, and I think at some point you have to start to believe them that this is coming to, to fruition. Well, I

Kevin Pereira: mean, we, there, there is the, the, um. The jitters and the starts and stops that come with every, uh, month, every new model release, every everything.

And people are sitting there going like, well, wait, are you gonna wow us? What's next? What's around the corner? What do we got open AI themselves. Drop O three Pro this week. Everybody can get hands on it again, the line go up exactly as predicted. So have we hit a wall? Uh, not yet. If we're still talking about O three probe eight months from now, 10 months from now, then maybe there's an argument there.

[00:12:00] But I, I. Where, where is everybody that, just a week ago was saying that, that the models themselves aren't getting smarter, that the tooling's getting better, but there's no, uh, new intelligence out there.

Gavin Purcell: Yeah. And this is like, you know, they've been teasing O three Pro for a bit. It was supposed to come out a while ago, but it did just finally release, and the benchmarks are better, right.

It just keeps getting better and. We have said on the show a couple times, they are expecting a GPT five in the summer, uh, which is probably June. Uh, and that means that, like most assume that this is gonna be some sort of unified thing. I think a couple interesting things about oh three Pro, just for people out there who we don't have access to it because we don't pay the additional $200 for, for chat GPT Pro, this is what people are saying right now, is that it, it thinks a lot, it thinks it, in fact, it's specifically designed to be one of the most.

Think intensive models. It is much better at what, at these long, hard problems according to most people. Kevin did find a tweet that made me laugh. Kev, this was somebody who asked it. Hello? And then what happened here?

Kevin Pereira: Yeah. Uchin Jin, uh, shout out to them on X. They said, oh three PRO is the slowest and most overthinking [00:13:00] model.

A simple high cost me $80. Now that's hyperbole, but he said to oh three Pro. Hi, I'm Sam Altman. Oh three Pro thought for three minutes and 54 seconds only to reply with. Hello, Sam Altman, how can I assist you today? And you can't really see the chain of thought with O three Pro. They kind of obfuscate that.

Some models are very open with what they're thinking. You can see every little step in logic along the way. And so a, a lot of people glommed onto this and started asking the most basic stuff, you know, is hotdog sandwich, all of that stuff. Just to watch O three Pro think for, you know. 15 minutes, um, about down, down, down a few trees in the process.

Yes. Yes. And overcook the oceans, uh, spending the energy to deliver these answers. And it's really funny for some people, but because we are Kevin and Gavin from AI for humans, pat ourselves on the back, we do have access to O three Pros. Chain of thought, Gavin? Oh, we do do, and I don't know if you wanted to chat.

Yeah, we have an, actually, we have an AI agent that's powered by the new O three PRO model. You [00:14:00] can ask anything, even though

Gavin Purcell: don't describe to O three Pro. We have an agent that is, that can get to that. Is that what you're telling me?

Kevin Pereira: Insider Axis has its perks. Baby Sammy's want some, loves our podcast.

Should we ask them? We can ask them. Yeah, yeah, yeah. Well, it's, it's a guest that's been on our show before. It's good Guy ai. He shows up whenever anything gets too in the weeds. But, uh, now he's powered by oh three Pro. So should we just ask him to say hello?

Good Guy AI: Hey. Hello. Wait, wait. Okay. Um, Kevin. Gavin now should I say hello or hi or hey there.

Oh, but hey, there has that weird cowboy vibe.

Am I?

Good Guy AI (2): No. No I am not. Hi. Might be too casual. Hello is formal, but also classic. But what if they think I'm cold and robotic? If I say just hello? Oh, no, good guy. Just if they want casual. What if they hate casual? What if this is a test? Okay. Okay. I've decided.

Hello, Kevin. Hello, Gavin. Oh, I did it. Or, or did I, should I have waived? Hold on. Can I wave? Oh no. Forget

Gavin Purcell: it. Moving on. Moving on. Po [00:15:00] Pos go, guy is positive. Positive. So that is what an AI is. Just to be clear, he is overthinking. We are listening to his thoughts as he's thinking about saying hello to us.

Kevin Pereira: That

Gavin Purcell: was good.

Kevin Pereira: I love the, I love the moment that it's got a casual cowboy vibe. Yes. Am I a cowboy? I don't know. Can I wave? Gavin, do you wanna ask a follow up of Good Guy AI powered by O three Pro with

Gavin Purcell: overthinking? We don't need to have him that much longer, I'll be clear. But let's just ask, let's ask him, let us ask him, uh, what does he think about those who think that O three PRO is overthinking here?

Kevin Pereira: Oh, you're gonna ask him how he thinks about overthinking? I don't. I wanna see if he's gonna blow

Gavin Purcell: himself up. Let's see what happens.

Good Guy AI (3): Oh, oh. Those who think I overthink too much. Wow. Okay. That's, that's a, how do I put this? Is it an attack? No, no, no. Probably just an observation or is it? No, calm down. It sounds like me in our intro.

Okay. Look, look, look. First [00:16:00] off, I would say, wait, should I say something? Maybe silence would be more powerful. Or is silence passive aggressive? No, don't be silent.

Good Guy AI: Oh no, not that.

Good Guy AI (3): Okay. I'll speak. I'll speak. I would say, well, esteemed critics of my cognitive tempo, might it not be that what you deem overthinking is in fact thoroughness.

Gavin Purcell: What was that word? Or,

Good Guy AI (3): or maybe it's caution or consideration, or. Existential dread. No. Scratch that one. Shut him off. Thank you. Shut on. If I saying that I sound defensive. Should I instead embrace it like, yes. Shut the singularity is gonna

Kevin Pereira: delayed now because of this. Okay. Thank you. Thank you. Good guy. Ai.

Geez. Um, okay. That was, uh, GPT-4 0.1, by the way. Told to play a character that was good guy ai. We fed that to the new 11 labs V three speech model, which in case you didn't hear, is kind of fantastic. Yeah, that's a weird voice in general, but the fact that it can auto generate tags like gasping or laughing or whispering or, you know, uh, anything tense, introspective, and you can hear [00:17:00] that from the character and it generates fast.

Oh, so

Gavin Purcell: yes, O three is overthinking. Uh, the other thing about O three that's important, O three, regular O three PRO is overthinking. O three regular OpenAI has dropped the price on the API Kevin by 80%. Yeah, it is now cheaper to use O three, their one of their biggest reasoning model. Not the best now 'cause of O three than it is to use GPT-4 Oh.

And. This kind of sent some shock waves through the AI business world because you know, we talked a lot about how cheap Google's reasoning model was and how it compared to open AI and to Claude this undercuts that. So it does make you think, okay, what is going on here? Are they just deciding they're gonna take a loss on this going forward?

Do they have something so good coming forward next that we are expecting that that's gonna blow this outta the water and this will feel like a second ran. What do you think about that?

Kevin Pereira: The, what you just said there at the end has me most excited because typically they drop the price when they're about to drop something new.

And that's new thing feels wildly overpriced when it first drops, but it usually is far [00:18:00] more capable than the last thing. I will point out that while. The, you know, the per million token generation price of oh three now is, is cheaper than some other models. That model tends to, because it's a thinking model, generate a lot more tokens.

Lot more tokens. Yes. So it's still, yeah, it's still an expensive model to run, but it is a drastic reduction in price. And now this opens the door, especially for people that are using it for coding, which, yeah, it. Seems to be incredibly adept at now. It can be a daily driver. You can go back to oh three, not just for the high level architecting, but for the granular building.

Gets me super excited, but, but really genuinely excited for whatever GPT five is around the corner.

Gavin Purcell: Yeah, and there's one other little piece of open AI news that Sam dropped, which is that he said the open weights model, which is their open source model that they've been promising for a bit, is going to quote, take a little more time, but it is coming.

There are a lot of people that are very excited about an open source model from OpenAI. We will see what this does. We also know sometimes that open source models are not as capable or they really won't be as capable as the frontier models, but lots of people are [00:19:00] excited about that, so keep an eye on that as we go forward.

The other really big piece of news when we talk about the super int intelligence stuff, Kevin, is that Mark Zuckerberg, Marky Z, that's my new name for him, is Marky Z. He is going out and personally, he's so mad about what happened with Meta's ai, uh, situation, which we know with LAMA four. Yeah, the LAMA four, we know that.

Their benchmarks weren't really all that honest that they might have been focusing on making benchmarks specific and it's not as powerful as they had hoped. Mark is going out and creating an AI super team, and what he just did was he spent. $14 billion of Meta's hard earned money to buy 49% of a company called Scale ai, which is most famous for the data that they provide to these AI companies.

But really what people are saying is what he really bought was 28-year-old Alexander Wang, CEO of scale ai, who will no longer be CEO and will come inside Meta and run a new super team. They're calling it. Yeah. As they approach super intelligence. So this is another weird story about how. [00:20:00] You're seeing all these big movements around the idea of super intelligence, and that's how big these, these companies see this opportunity.

They're gonna spend $14 billion on what amounts to half of a company and a person to kind of help them run it.

Kevin Pereira: Yeah. And, and full disclosure, Gavin and I are undecided if we're gonna accept the invite and join the teams as well. We do have a pretty busy schedule with conferences and speaking and our startup and yes, you know, day jobs and all that stuff.

So, uh, we don't know if we are going to accept, but the rumor mill is buzzing around the old X that, uh, eight figure offers. For a single year. Yeah. Eight figures in a year are being offered to certain engineers of cash. Of cash. Not yet. Yes. Not just a stock that you're going to vest in six years time, but cold, hard cash offers of the, uh, in the eight figure range for for certain expertise.

This shows how serious Mark is about pushing in this direction.

Gavin Purcell: Yeah, and I think that's just one thing to be aware of is like, we talk a lot about obviously what's for you, the listener out there is important, but when you talk about the richest people in the world. [00:21:00] They are all encompass focused on this one thing and there's a reason why they got rich and we know part of that is exploiting labor and all the stuff that they did to get there.

But the other part of it is they saw some version of stuff coming before other people did. So just again, to reflect back on the Sam blog post or other things. Yeah. This is not something that is being made up in real time. Something that can be made up is your enthusiasm for this show because you should pretend to love us so much that you wanna go and subscribe to the AI for Humans YouTube channel.

If you're watching it, click on that button right now and Okay. Leave us a podcast review. Right, Kevin? People should just pretend that they don't like us. Yeah,

Kevin Pereira: yeah. My bit was convoluted yet you just implied that people should pretend to subscribe to us, I think. No, no, no, no, no, no. Go s pretend to like us.

Actually, click the button always.

Gavin Purcell: Yes, please subscribe. Yes, you can pretend to like us as you're doing it.

Kevin Pereira: That's right, and that's how Zuckerberg made his bazillions was by begging people to back him on Patreon, which you could do. That's exactly $5 a month a tip goes into our jar. That's, thank you.

Sincerely, thank you to everybody who shares this podcast, who gets in your car and drives from [00:22:00] Victoria. Yeah. British Columbia. Oh yes. All the way to Banff to visit us at a conference. Thank you. Thank you. Sincerely, kind of amazing. Um, but the only way this grows is because you tell people about it. So, uh, we, we thank you each and every week and we mean it from the bottom of our Cold Dead hearts.

Thank you for sharing this podcast. 'cause the, the number is going up and it's allowing us to pay for VO three credit.

Gavin Purcell: What it's not allowing us to pay for Kevin is the iPhone that I upgraded to. I updated the iPhone 16 that I now kind of wanna throw out the window. Had to have Apple intelligence, didn't you buddy?

I did. So let's get into this. Kevin, WW C just happened. That's Apple's developers conference. This was, I think, a universal AI failure by Apple. Is that fair to say? Like I think overall. Most of the presentation. Oh, on the AI front? Yeah. Yeah. Most of the presentation wasn't about AI at all, whereas all of last year's WWC was all of these promised AI features.

Uh, it's safe to say that this was underwhelming on the AI front.

Kevin Pereira: I, I think that's an understatement. I understand why [00:23:00] Apple didn't. Focus so much on AI because it became sort of a joke they that there were lawsuits about advertising and selling phones like the one you have in your pocket. Gavin, on the back of features that they did say would be a little delayed.

They'd be a staggered rollup, but they're nowhere near rolled out right now. And the backpedaling seems to be, hey, we're trying to do things securely and on device and no one has cracked that, not even Apple. So we're gonna keep taking our time to get it right. And that's. You know, I, I think that's a, that's a fair ish statement to make.

Yeah. But it completely ignores the fact that many others are intelligently and elegantly integrating AI into their stack across the board. And Keynote still sucks. Final cut. I'm still waiting for features that are in free web-based editors. Like no worse than that. Kevin is Siri.

Gavin Purcell: I mean, Siri. That, that what I was driving to Gavin in our hands.

Yes. Yes. Fair enough. That is the thing that needed to get solved and hadn't, I will say. Mark Germond, who is a business weekly reporter, often gets Apple Scoops. Had a really interesting [00:24:00] tweet about this 'cause I don't think it was all bad. He said, excellent. WW C. Cohesive story. Deep integration and continuity across devices.

Zero false promises this year. Impressive new UI and significant productivity features on the Mac and iPad, but the lack of any real new AI features. Despite that being my expectation, he wasn't expecting anything more. Startling. If I were an Apple shareholder, I'm not, um, I would be worried, right? Because this company feels like it is slipping far away.

And I know there's always the jokes about like, Steve Jobs wouldn't do this, or Steve Jobs wouldn't do that. This doesn't feel like a company that has much of a direction right now around ai, at least to me. But did you see the glass? Did you see the liquid glass? I did see the glass. Yes, I did see liquid glass.

Liquid glass is their new design. For their UX ui, they're updating their iOS name, but liquid glass is going to be the look of what this looks like. Some designers thought this was great. Yeah, I think most people had a very hard time understanding how they were gonna see some of this stuff in this way.

Um, it's been relatively kind of derided across the board. At least this is like [00:25:00] taking a chance at something I think from a design standpoint. It depends on where you see it. It's interesting, but it, it, it's not enough to me in some form.

Kevin Pereira: Yeah. But back, let's back on the ai uh, side of things. Uh, a friend of the show, Joanna Stern, had a chance to talk to some Apple execs for the Wall Street Journal.

Do we wanna play a few clips from this interview? 'cause I thought she was fantastic and the answers were amazing. So this is Craig and

Gavin Purcell: Josh. Two well-known, uh, apple employees, not Tim Cook I noticed, but they are basically being interviewed by Joanna, who is again fantastic. Kind of taking them to task for where or what happened with Apple intelligence.

Joanna Stern: Siri should be as good, if not better than the competition.

Oh, I think ultimately it should be. That's, that's certainly, but it's not right now. That's certainly our mission.

Gavin Purcell: So Kev, these two guys, first of all, they, I was thinking these two guys look like they're in some sort of really, like, almost like a, a Larry David show.

Like, it's like they're reacting in certain ways, the faces on them when they're being asked these questions. They're just kind of dodging them. And again, I don't begrudge anybody who's been through a rough time in business because clearly these guys have been [00:26:00] put through the ringer. But there are no answers here.

Like, I don't know why you sit for an interview like this when you're just gonna say these things again and again,

Joanna Stern: between you both, you have, I believe, around 60 years of experience at this company.

Yeah.

Joanna Stern: You've seen the company go through highs, through lows. Where do you think you are right now? There's a lot of sentiment that Apple is, is on its back foot here.

Well, I think, I think you're right to bring up that. Perspective because I think when you have been through different waves, you're, you're very accustomed to, uh, to, to the ups and downs. And I think we're feeling really good right now.

Kevin Pereira: They're feeling good. Gav spirits are high team's good. They were doing trust falls all day.

Gavin Purcell: I I, again, I I really do appreciate how much work that these two people do and everybody at Apple does, but like. That looks like denial clearly. Like, or, or it's either that or it's been like, you know, the Steve Jobs, like the, the what do the reality distortion field going on the reality distortion field.

Like they're trying to pull that off, but they can't. Yeah. It's like, it's like a, a lesser Jedi trying [00:27:00] to like float Yoda and he's like. This does not work. You might

Kevin Pereira: Right. Body up, upside down, gc. Yeah. Look, I can only imagine the, uh, level of politicking going on there. Yes. How a decision three divisions away, rolls down and has ramifications for something somewhere else.

Where someone says, these models need to do this, and they, they don't perform the way they want. I can only imagine the pressures that happen there, but as, as someone who is recording this on their MacBook. With an iPhone as the camera. I'm very much in the ecosystem. I want them to succeed and do better.

It is just getting harder and harder with every passing week that Siri is a massive disappointment. It's just getting harder and harder to defend them, and it's getting harder to stay in the ecosystem.

Gavin Purcell: Yes, I think the Google phone, I. Predict and I'm, I am an Apple fan person. I've had Apple products forever.

I suspect Google's Android will make some significant inroads in the same way that chat GPT has made inroads into Google's search business. I think that Google's phone

Kevin Pereira: will start cutting into [00:28:00] Apple stuff. And real quick, we're not gonna get in the weeds on this one, but Apple did release like a research paper, basically saying that, yeah, LLMs are not as smart as they appear to be, and that has.

Drawn some iron from the community saying, wait a minute. They didn't exactly perform these tests properly. I just wanna say, if you're maybe on your back foot and you maybe haven't integrated AI into any of your products in a truly meaningful or impactful way, you might not wanna spend the time releasing white papers saying Everybody else that seems to be doing it.

Right. Yeah. Yeah. It's wrong about how capable these things are. That's all I, I'm just, just saying

Gavin Purcell: and, and just to the point, like, you know, we've, we've talked about this guy every once in a while. There's a guy named Gary Marcus, who's a very famous, kind of like hater, specifically of like open ai and a couple of these companies that are kind of, he sees us super hype beasts.

He and Sam mom, we got into it over a back and forth around Gary calling out. Sam Alman looking like Elizabeth Holmes from Theranos, which is like, Sam was like, this is Bologna. Um, the thing that Sam said back was like, Hey, we are the number, you know, five [00:29:00] company or five website in the world now. We have done things that people find useful every day.

I. So whether or not you think AI is hyping or this paper from Apple says the reasoning models are not as good as they think they are, guess what? These companies, apple are lapping you right now. They are lapping you in ways that I need to use your product, and that's a bummer. As a consumer.

Kevin Pereira: Yeah, we should sue them Gavin, but I think.

The courts are kind of busy right now, and it's, it's not busy. The courts with Apple this time.

Gavin Purcell: No, there was breaking news today. In fact, while Kevin were on stage. We're, we were forgiving a talk that Midjourney is being sued right now by Disney and Comcast, specifically around the idea that Midjourney can create and released the, uh, IP of these companies.

They call out in this, in this lawsuit, specifically on the Disney side, Darth Vader, and a number of other characters. And on the universal, uh, Comcast side, they talk about the minions. And the thing that's going on here is they are not suing about the training data per se. And we can talk a little bit about how this comes out.

Mm-hmm. They are suing by the ability that it is [00:30:00] providing this result to a user who asked for it. So whether or not in the training data, we have to assume that this is in the training data because they can provide it. Uh, if you see on chat GPT-4 oh image gen or other things where they will say at times, I'm sorry, I can't do that for copyright violations.

What, uh, Comcast and Disney are saying here is that Midjourney is not taking that step, that they are allowing their IP to be generated from within the system. We knew more, uh, lawsuits were coming. Kevin, this is very realistic. I have a question to you, which is why is it Mid Journey getting sued and not these other companies, and does this maybe have connective tissue to back channel deals that might be getting done and maybe Mid Journey isn't one of those companies that's coming to the

Kevin Pereira: table?

I. That's interesting. You know when when Google updated imaging and when chat GPT rolled out their new image features, even we were saying, wow, they're playing pretty fast and loose. Yeah. With ip, they either feel really good about the data that went into the model, or they must feel very protected about the outputs that are coming out [00:31:00] and you can still generate.

Some fairly protected stuff from some big companies. So is that, I mean, to pose the question back to you, do you think there's some backroom handshakes? Do you think there's some, uh, you know, royalty payments going out? What do you think is happening there? Well, I, here's the

Gavin Purcell: thing about Midjourney that's slightly different than all these other companies.

Midjourney is a relatively small company, right? I think in fact, they purposely have stayed to be like 10 to 15 people. It's actually quite small. They also have not advanced, I'd say, as much as some of these other companies have. Maybe there's a world where they just don't have the people to deal with these sorts of, uh, legal claims or they were choosing to not deal with them and the companies are coming after them.

It does feel like the larger companies, the Googles, especially of the world, but even the opening eyes. Are listening to when these things come around and like very blatant versions of this, they do seem to not let you, uh, create anymore with, and again, to your point about four oh image Jen, like there was that lovely weekend where all of us could do whatever we wanted for two days and then very quickly, quickly came [00:32:00] down.

Yeah. It was, was glorious

Kevin Pereira: Before the man came to town, we were generating shreks feet into all sorts of stuff, swamps. But we did and we should never lasagnas

Gavin Purcell: Kevin, there is a lot of really interesting robots. Stuff happening this week. It is time for a robot to watch

Robot Watch. It's Robot Watch.

Robot Watch.

Gavin Purcell: Okay, Kevin, there is a big video from the company one X Robotics, if you remember. These are kind of the strange. Cloth covered robots that actually some people say are very, very good. They're very high. They have released something called redwood, which is a, a model for these, uh, one X robots to actually operate in the real world.

There is a lot of creepy weirdness in these videos. One of the things is, so there's, these have come out kind of back to back. The first one kind of shows what redwood is and how in a, in a room in your own home. The, uh, robot can help you by like squatting down and putting stuff in the dryer. It can also [00:33:00] like grasp onto a door.

One of the interesting things it talks about is it might miss the door the first time, but it kind of learns how to grasp onto the dry. It'll reach. Yeah, yeah, yeah, yeah. Which is very cool. Robot

Kevin Pereira: squat is like, that's gonna be my robot squat. That's our new header for me. Robot squat. Can your robot squat?

Yeah. If you thought I like my robots, I, they just haven't. Waddled creepily throughout the house to grab me, uh, alcohol, or they haven't stared longingly in a field of tall grass. Well, what they're showing off in the demos is that they're, they're moving seemingly autonomously and they're manipulating objects that they've never seen before.

They're just intelligently navigating about and figuring out how to grasp, how to manipulate, whether that is a. Beer bottle in one scene or a door handle in the other. And that's cool. I mean, I look, Sammy Altman was predicting that in two years time, robots will be able to move autonomously around our environments.

Thanks to ai. And this is kind of a glimpse of that. I'm just watching the second video, which is called

Gavin Purcell: Redwood AI Mobility. Yeah. The beginning of it is you watch a robot who's kind of following a woman walking through the, to your point, this giant field. [00:34:00] And it looks like either the beginning of a horror movie or.

Something terrible is gonna happen, and then it cuts to, what is this like beautiful wooded area, and these two robots are just sitting and standing on a bench randomly. What I wanna know is if you were running in this forest and you were going for a jog and you came around the corner and you saw this, what is your reaction?

What are you going to do when you see these things? I turned my car

Kevin Pereira: keys into the Wolverine thing where you have them piercing through each finger and I get my ready off. But that's the, that is going to happen in the near future. Yes. Right? Yes. The guy, the, the guy prone working on his car, and then the robot kneels next to him.

Like, if I'm changing the oil in the old Hyundai and I turn and I see one of these robotic knees coming down near my face, I'm stabbing it. I'm, I'm under the WI understand.

Gavin Purcell: I understand it completely. I understand it completely. The thing we've talked about with these shows, if you're new, these robots particularly, my biggest problem with them is if it's making your tuna fish sandwich for you and it, oops, I spilled on myself tuna fish.

How do you clean this robot? Because [00:35:00] fabric absorbs. Oh, a stinky

Kevin Pereira: rope it. That's exactly right. And a filthy bit. That's exactly right. Watch me Gavin. Sure. Stinky.

Gavin Purcell: So anyway, what X robotics, we do appreciate this. They did go out of their way to what make what is a relatively cute, creepy robot. Probably not the creepiest robot.

Kevin, there's some more robot news we wanna get into first and foremost. Amazon is now trying out humanoid robots in its factory. So when we talk about job loss, we talk about all this stuff. We talk about the idea that robots might take blue collar work. This could be the beginning of what we're talking about.

They're rolling this out in a, in a kind of a cautionary way for right now to see if they can use them. There was this video that came out last week from figure oh one about how it would sort packages, and it looks kind of goofy, but a lot of Amazon warehouse work is that kind of thing, right? It is like put the thing in one box, move it to another box, do all this stuff.

So this is, I think an under, an under reported, really big deal of Amazon is moving towards this world.

Kevin Pereira: And also related, like sorting packages is one [00:36:00] thing, but um, I don't know from hobbyists that fly their drones to capture their neighbors in their own backyard, uh, to military applications where autonomous vehicles are going to be flying around.

But they say, no, no, no. The humans still have the edge. Oh, surprise, surprise. This one, this one somehow managed to hit the news bin. We didn't catch it at first, but an AI drone beat a human champion for the first time. Yeah, it's a big deal at a Abu Dhabi racing event. Yep. It's a new deep neural network, and it sends control commands, basically flight, uh, controls directly to motors, and you, you watch the video of a, a drone flying autonomously, besting all of the human pilots as it navigates around a course, and you go, oh yeah.

Why are we still building planes for flesh bags to sit in the cockpit and fly about instead of investing it all in the software that can out fly human beings now.

Gavin Purcell: Yeah, this is, you know, uh, AI safety memes, who's a pretty strong AI safety advocate is where we'd found this 'cause it was a couple months old, but like.

The point here is that like drones are getting better than we [00:37:00] are at flying themselves. So pretty soon autonomous flying is going to be the way these things are gonna operate. Uh, the, the poster, I think is insinuating the idea of like, Hey, when they're autonomous, how much control do we have over them, especially if they have weapons.

But just another thing to be aware of as we move forward into the next generation of robotics.

Kevin Pereira: That's how I want Amazon delivering my zevia and my organic toothpaste. I want. Swarms of these things flying, darting. Tom is

Gavin Purcell: here. I'm throwing it to your head and like we, I get killed by a Tom's uh, tube that's injected into my brain right there because, sorry, Gavin should have been in the way.

Kevin Pereira: You gotta duck when the, when you hear the drones overhead, you gotta get outta the way. I might have a pair of ICS arriving and I, I don't have to bend over to put those things on my feet. Hashtag

Gavin Purcell: don't, not an ad. Don't use, don't. If you get a, if you get a, uh, a kettlebell delivered, do not use it right now because Kevin just destroyed his knees.

That's right. To get kettlebell this morning. Uh, alright, everybody's, that is true. Uh, we wanna get to show off some of the things you did this week. It is ai. See what you did there. [00:38:00] Without a care,

Ilya Sutskever: then suddenly you stop and shout.

Kevin Pereira: I love this prompt, Gavin. Yes. We gotta start with a GPT-4 oh image prompt that I think you found on Reddit and turned me onto, and you had me do it at a restaurant here in BAM, and then it kind of took over the evening with us.

Gavin Purcell: Yeah, so this is from a Reddit user named. Jewel, IEGB. I'm not sure how you pronounce that, but they basically, the, the post was just, what do you get?

And they had a screenshot that says, ultra random photorealistic image please. Period. No questions. And what's so much fun about Reddit prompts like this is that you see everybody else's generations and they're kind shared in one place and people will go through and up, upload things. But yeah, I tried this, Kevin tried it.

So my, so my result of this was, uh, I got a weird guy who had a slug, uh, next to him and he had a bottle of a jug of milk. But there's [00:39:00] all these amazing photos in here that some of the ones that I just couldn't believe they came out like somebody with just a hand with a potato floating over it. One of them was randomly a pickle and a toilet, which is just strange, but it's just one of those good examples of where somebody comes up with a weird, fun prompt, everybody goes and tries it, and you just enjoy seeing the collective kind of fun that comes outta it.

What was yours? What did, what was yours looking like?

Kevin Pereira: Uh, the first one was a, a, a, like a shot of a table that had a, a head of broccoli on it. A glass of soda with ice, a human skull, a rat, a piece of toast with jelly, some d and d dice. A camera and the letters BXR. No reason why. Who knows? I told it. Hey, that's not random enough.

That's when I got the Kawasaki handheld torch with a chicken leg coming out of it. A fish wearing a cowboy hat vomiting a guitar, and an upside down light bulb with duck feet. Because why not? Why not? Jeff, why not the,

Gavin Purcell: the weirdest one? I will say, and this is one that maybe if you're watching YouTube with your family [00:40:00] involved, just be aware this may be a time to turn around.

There is a, a, a monkey in a tub with surrounding them is what? Water. But it looks like the monkey might have relieved himself in the water. Yeah. And floating next to him is a bag of Doritos. So again. AI is getting weird and random in a strange place that we just have no idea what it's doing. Sometimes

Kevin Pereira: such a fun prompt.

And again, go, go rip your own prompt for free. Go try it, see what you get, and modify it. Have a good time with it. That one was so great.

Gavin Purcell: Yeah, Kev, the other thing that we wanted to shout out this week was one of our favorite creators, John Finger, um, has partnered up with Luma Labs and you know, last week we teased.

Luma's modified video, but I dunno if you know this, it was in our top of our show, but we didn't actually say what it was. We kind of skipped over it while we were recording. So apologies to Luma. We did get you in briefly, but John is now working with Luma directly and he just does the most amazing things with, uh, AI tools.

And what he does in this one clip is he takes his two shoes on a shoestring. Pulls them [00:41:00] around the corner and uses Luma Labs as modified video tool to turn it into cars racing. And to me, this is, John is just a good example of like, you know, you used to have these people who would be like Claymation people who would do these kind of amazing things.

They would just kind of tinker with stuff. John is that for AI video and I love watching what he pulls off. 'cause every week or every day, sometimes he has something so cool.

Kevin Pereira: A hundred percent. When you see a video of him, um, like he's got someone writing, uh, basically a folding chair in the garage holding onto a belt buckle, and it turns them into an old western cow poke on the horse.

As the camera pans around, your brain immediately starts to wander of like, oh, right. Of course, a styrofoam ball with a face painted on. It could be a cyborg head going onto, uh, a robot, which is the next scene in there. You just start to immediately go. Oh, right there. There, there are no rules anymore. Yeah.

Like your imagination should run wild because if you can think that the AI can really help you imagine it and bring it to life.

Gavin Purcell: Yeah. There's a really cool post he did from last week that we didn't shout out, but it's just really [00:42:00] cool where he's basically created a version of himself and then he ha he shot video of a puppet, and what he's doing is using the puppet to puppet a version of himself so that it moves like.

You can see the version of him is moving the puppet like he is. And it's just weird, strange explorations like that that I really do appreciate, uh, people doing. And Kevin, speaking of weird, strange explorations, we were in Banff this week. We're literally there right now. We may not look like we're in the same place, but we are.

We're in a hotel, we're just shooting this way. 'cause it is a nightmare to upload video on a hotel wi a wifi. And this is a simpler way to do it. But we did a presentation about the idea of how these tools are allowing people to make. Mm-hmm. Sort of all sorts of interesting stuff. And as part of that presentation, we created a fake version of the office that is inspired by uh, uh, let's say it's Sopranos.

So you can imagine a Sopranos versus the office. And we should just show this right here. This is just a dumb, fun thing that we made together. Over the course of not that long, maybe like, you know, four to six hours of total time. And this is the sort of things that [00:43:00] like when you just explore what's possible with tools like four oh Image gen, um, cling 2.1 or uh, VO three, VO three and Suno for some

Kevin Pereira: music as well.

We, we walked through the whole pipeline for the audience in attendance of like, look, these are free tools. There's nothing magical about this. We're just using off the shelf tech to help us get past like idea roadblock. Yeah. You know, get past the blinking cursor, generate characters. Puppet them. Uh, you did a great, uh, uh, runway act.

One thing that had a lot of people inspired of like, wait a minute, I can take this lifeless still and give it a performance and then use 11 labs to change the voice. And you know, this is kind of 1 0 1 ish stuff because we talk about it all the time. But it's so fun to share these pipelines with people and watch them kind of light up and get in inspired, especially when you know, some of the folks in this crowd.

And we, we. Understandably, some people are AI averse. Yes. For myriad reasons. Yes. Maybe it's environmental, maybe it's about, uh, you know, artistic integrity, the original sin of the model training data. Who knows? Like there's a million reasons, which are valid reasons to be skeptical about this stuff or be, be concerned.

But it's always nice to be able [00:44:00] to say like, listen that aside, we are here to share our pipelines and our explorations with you. 'cause people are doing this while you are debating, people are still using this stuff.

Gavin Purcell: That's right. And so thanks. Shout out to the Banff World Media Festival, which is where we're at.

Banff is a beautiful place. Um, and we'll see you all next week. We will be back to our normal locations and our normal cameras, but thank you for watching. If as always, if you have a thought or a question or a comment about what we do here, I. Uh, shout us out on our socials or send us an email.

Kevin Pereira: You know, let me just, we can cut this, but let me ask good guy AI for the perfect way to sign off a podcast.

That's gonna make everybody wanna subscribe. Okay. Sure. And like, no, it's gotta be good. Okay. Sure.

Good Guy AI: Oh, no. Oh, no. Uh, I, I, I need to not overthink frantic. I panic. Oh no. Did I under think desperate? Is that worse? Wait, wait, wait. Let me do it again. Short pause. Okay. Okay, thoughtful. Outro. Outro. What is an outro?

Is it a farewell? A call to action, A digital hug. Do I sound too desperate if I beg them to like and subscribe? [00:45:00] Do I sound too cool? If I don't, should I reference their humanity? What if I alienate non-human listeners? Oh. Oh, no. Well, um, thank you. Or should I say gratitude transmitted? Oh, but that's weird recovering.

Okay. All right. I, I thank for joining the AI for Humans podcast. If you, if you found joy or thought or friendship here, tentative. Perhaps maybe consider engaging with a like and subscribe mechanisms if that feels authentic to your personal journey. Okay. Thank you. Or not, Gavin. Or not. Okay. He or not, or?

Oh no. Do I say bye? Do I say farewell? Oh, no, I'm doing it again.

Gavin Purcell: Bye. Bye everybody. We'll see you next week. That was pretty funny actually.