Ted Talk BOMBSHELL, Microsoft CEO WARNING,, Autonomous Vehicles DESTROYED, Apples Secret AI Plans,

So, with another stunning day in artificial intelligence, let's take a look at some of the most interesting stories that you probably did miss. Now, there was actually something I wanted to show you all because this brings us to the wider question on. Now, there was actually something I wanted to show you all because this is something that I think is rather important because as AI becomes more and more integrated into society, we are going to see more and more instances of this where people are purposefully trashing autonomous vehicles or autonomous property. Currently, what you're seeing is an individual standing on top of a Waymo and then of course, breaking the windshield with their foot. Now, this is of course not good. They are destroying private property, which is essentially probably a crime at this point. But I'm not too clued up on the laws. The point is, is that the integration of autonomous technology into society and is bringing some questions that maybe some people just aren't ready for this technology and maybe some people are just too scared. Now, I think this is a bit of an overreaction, but I'm guessing that from the clip, maybe the person is just homeless, since they seem to be holding some kind of, you know, cardboard sign. But discussions surrounding this are pretty interesting because like I said before, this is something that is quite new and it's probably the first instances of this. I mean, one of the things that we've previously discussed and something that we've seen from industry leaders keep on stating and is that in the future, it's very likely that there will be 10,000 to 20,000 robots deployed. You know, not every day, but like every month around the globe. And I think if that is the case, if there's billions of humanoid robots, how are we going to integrate them into society when some people may just destroy them? Is it instantly going to become a crime? And what is that crime going to be? How long is the time going to be? Now, the reason I actually showed this clip as well was because there was actually an interesting conversation from, from the CEO of the company. So the VP of the company actually explains what a Waymo is. If you don't know what a Waymo is, it's basically just an Uber without a driver and it's completely autonomous. And he actually makes some rather fascinating points on why he thinks that this is the future. And I have to say I do agree.

We have a lot of data to show that our deployments, I think, compared especially to humans in various studies, is shown that we have a lot less crash rate, is a lot less, and our property damage claims There is actually a collaboration with our Swiss reinsurance company showing that, say, injury claims went to zero with our vehicles and property. That's based on 3.8 miles that they analyzed. Right. And property damage claims went down 76%. You're in the vehicle by yourself, you can play the music you like. There is a certain level of, I mean, it's a high end vehicle, it's a premium experience today for the price of normal experience. I think over time there is tremendous opportunity to optimize both our models and vehicle cost and operations to make it yet more affordable. We have a lot of.

I think that is honestly quite interesting because it seems that, you know, taxi drivers, I wouldn't say they're going to get replaced entirely, but as it does get rolled out to more cities, as these cars become more adaptable to certain scenarios that they haven't seen before, and as they start to learn the roads more and more, I think it's not going to be surprising if we do see this rolled out worldwide. I mean, if cars are fundamentally better drivers than us, and if regulations require that, you know, you have to have something that is extremely safe, I'm guessing that this is most likely the future. Because humans are usually the ones that cause errors. 90% of the time when someone crashes. Maybe they weren't looking, maybe they weren't paying attention. But usually these systems can work. And of course you're going to have to get to a system that is like 99.9% accuracy, because if you extrapolate that out to, let's say 100 million vehicles, if you have 99%, then that 1% is actually quite a lot of crashes. So you do need something that's like 99.999% accuracy. And I'm guessing that maybe sometimes some of these cars will have to actually work together to ensure that there's no crashes. But I'm guessing that it's going to be kind of fascinating to see how this thing kind of rolls out. Because some people talk about the personal experience of being in an autonomous vehicle with no physical driver. And some people do like actually having a human driver that they can ask to, you know, change the song, do these kinds of things. I mean, I'm personally on 5050 because, number one, I don't want people to lose their jobs because I know that a lot of people that do Uber and stuff like that, they essentially require that because their jobs just aren't enough and they need some extra money, whether it be to fund a new business idea or whether it to fund a kid that's going to school or something like that. I think those things are important. So if you do take that away, I'm wondering how certain people will make money. But at the same time, I do completely understand the proposition that these cars propose. I mean, getting in an autonomous vehicle that takes you from point A to point B, you don't have to talk to anyone, you don't have to speak to anyone. You can set the music, set the temperature without having to, you know, awkwardly ask anyone. I mean, seems kind of like a dream. So let me know what you think about that. Now, another piece of content that I did see, that I did find interesting. And before you'll think that this is boring, this is a piece of content that was formed with Yu Gio. So the reason I'm actually showing you this is not because it's a piece of content, but because it gives us a glimpse into what the future might look like in terms of completely AI generated content. Now, currently you're watching this on YouTube and there is currently the question of whether or not in the future we will have completely AI generated experiences. And this is, you know, something that's a little bit like that in the sense that the content you're hearing is from Udio, which is an AI music generation software tool, which is outstandingly good. And I'm only going to play like the first 30 seconds, but I wanted you all to hear why this captivated me. And certain audiences.

Say that. Tinder girl, of course not. Send me a screenshot. Seriously, or I'm calling. Off the way. Babe, calm down. What the heck is this? A screenshot, like you asked. I am not messing around, babe. Neither am I. I take you and your name very, very seriously. Oh, yeah? Then what is it?

I mean, how hard would it be to have an. An application where you have completely AI generated stuff that is fine tuned to what you like? I mean, I'm guessing. I wonder how, considering the fact that when we look at how traditional social media platforms are designed, they play on our dopamine receptors. And if you don't know what I mean by this, the reason social media is so effective is because you don't know when you're going to get your next hit. If you basically look at it as if it's a slot machine, which is how they fundamentally design these platforms. Essentially what they're trying to do is mimic that. So you know how you swipe down or on TikTok, you'll scroll down and you don't know when the Next good video is going to appear. That's essentially the same kind of dopamine feedback loop that you get when you're on a slot machine and you don't know when your next win is coming from. So I'm wondering how, if that changes, if we're going to get completely just continual hits of dopamine, because every single piece of content is going to be one that we enjoy. That's going to be something that I do find to be interesting and I really do want to see how that could potentially work in the future. Now, of course, we do have the economic side of things and the fact that, you know, video platforms already struggle to make money and the compute costs are pretty crazy. But I think this is a little glimpse into the future of what might happen if economies of scale kick in and, you know, prices seem to go down. Now there was also discussion of Apple's Next Big Things next. Apple has their LLMs and they apparently will all be local and on device from the start. No cloud, according to Bloomberg. And there were rumors about this for a while, but this is a big momentum shift to local. Now, recently with the release of llama 3, we've seen how crazy it is that a 70 billion parameter model can run as effective as nearly some of the other models like Claude's models. And this is a big game changer because when you have a model that doesn't need to be connected to the Internet in order to work, I mean, the possibilities are endless because now we can have systems that, you know, the inference cost is pretty much just on your device and you can have things that are really, really rapid in terms of if you wanted to apply it to like a robot for example, because the problem is with some systems, unless you're talking about grok, which is insanely fast, is the fact that there is just a huge delay. And this does allow for a greater range of applications, which really does change the future. And I think this is going to be really good for Apple because if we can have Siri 24. 7 even when we don't even have Internet and it's able to talk to us in a very fascinating way and it's able to do stuff for us. I've been paying attention to the research and the research papers that they've been dropping, it seems like we're really, really going to head into a crazy future. Now the reason I know that this is going to be so crazy is because whilst yes, ChatGPT is very popular, you have to understand that I recently retweeted a tweet. And the tweet basically said that distribution beats like the actual base model. So if you can distribute something more effectively, that's going to beat even if your model is better than what you initially have, because it allows it to be adopted more freely. So the tweet here is distribution beatstech. With Apple language model release later this year, all of the OpenAI bet comes down to GPT5 being revolutionary. Time to throw the dice. And basically, you can see someone said, my pastor texted me saying that it's much easier to go to WhatsApp and ask Meta AI than it is to go to ChatGPT. I mean, of course ChatGPT could get embedded into WhatsApp, but the point is, is that ease of use is something that, you know, bests everything. So once Siri becomes an LLM or an AI system, or even an AGI system, you have to wonder, are people going to use, you know, sites like ChatGPT anymore? If they can get absolutely everything from an on demand model that they can simply say, hey, Siri, and. And it can give you everything that ChatGPT can. Which is a very fascinating question that could shift the dynamics because Apple owns the large share of the iPhone market in most of the Western countries. And if they're able to capitalize on that, because they already have the existing distribution, they have the MacBooks, they have the iPads, they have the iPhones, I mean, it's going to really change how the average person interacts with AI. So that's why I'm stating this here, because if you remember, Apple's been completely silent with regards to their AI updates and, and it's something that most people are confused on what they're doing. And I'm guessing that Apple are probably just waiting for things to just be ironed out in terms of the reliability, because their brand is one that is perfection. But as we said before, Apple's brand with perfection, it doesn't kind of line up with AI's sloppiness and it will be an interesting challenge for them to navigate. Now, apparently we have some interesting tweets here that it's not completely confirmation, but I'm so hyped up on AI news that I just have to put this in the video. And Someone says that llama3.8 billion will be amazing, but overshadowed by Phi 3 mini small 7B and medium 14 billion parameters this week. The benchmarks are insane. And that was a model that was trained, I think, mainly on synthetic data. And it pushes the boundaries on how compact models are able to perform on certain tasks. So if that does come out later this week, I think that that will be something fascinating because it will show us exactly what we can do with AI. Now here we get onto the main meat of the video, because this was something that was widely debated more among the Twitter community. And a lot of people were giving Mustafa Silliman a lot of flack about this. And if you don't know, he's the Google DeepMind co founder, founded Inflection, then is now at Microsoft AI as the CEO of that division. So essentially, just take a look at this clip first and then I'll get into why this is so controversial.

The EU have a huge piece of regulation called the EU AI Act. You know, Joe. President Joe Biden has, you know, got his own, you know, set of proposals and, you know, we've been working with, with both, you know, Rishi Sunak and Biden and, you know, trying to contribute and shape it in the best way that we can. Look, it isn't going to happen without regulation, so regulation is essential, is critical. Again, going back to the precautionary principle, but at the same time, regulation isn't enough. And so when I say, you know, regulation is not enough, what I mean is it needs movements, it needs culture, it needs people who are actually building and making, you know, in like, modern, creative, critical ways, not just like giving it up to, you know, companies or small groups of people.

Right.

We need lots of different people experimenting with strategies for containment. And the eu.

Now, I'm not sure why he was getting so much flack for this statement on Twitter and why people were saying that why would they hire D Cell? Why would they hire. That someone's not a pure accelerationist. And I just think that that isn't the full picture. As someone that is, you know, been in as many AI companies as he has, and as someone that has, you know, a crazy, crazy level of expertise, people saying that it's sad and people saying that he's a decel. I don't think that this is as bad as you think it is. I think the only thing that he's trying to say here is, is that we actually need to think about the problem fundamentally, rather than just applying some old legislations that are going to be superseded the moment AI just goes up a notch. And I think whilst, yes, right now it doesn't seem like AI safety is a problem, things aren't a problem until they are a problem. Like, you know, it's easy to say that, oh, this is Diesel, this is D cell, yada, yada yada, until you have a simple AI problem and maybe the AI escapes containment. And I know this sounds like complete fantasy sci fi gibberish, but it is until it isn't. I mean, generating images out of thin air was gibberish until it wasn't. Generating videos from text prompts was gibberish until it wasn't. And I think like I said before, with now the agent wave coming, I think things are going to get a lot more crazier. And I think what people need to think about is not in terms of like products being slowed down and AI being regulated to hell and being like setting the timeline back a few years. I think what people need to think about is the real risks of AI, which is mainly the risks on your personal safety. Because right now, if you exist on the Internet, someone could clone your voice, clone your face, they could understand all your tweets and how you talk. They could, you know, do many just different things that, you know, wouldn't be possible previously. And I mean, it's going to be very hard to navigate that future with this level of technology. And you have that problem, you have many other problems. And it doesn't seem like it until it is, but I think in the future that's the only kind of regulation he's talking about because this guy's an entrepreneur, which means, you know, it's the principle of move fast and break things. And I think whilst yes, you do have to move fast and break things, you do have to still be switched on with AI because there is that element of like, okay, if you make a mistake here, the implications are a lot worse because this is the thing, if we don't regulate AI now, and I'm not saying slow it down, what I'm saying is if you don't put in certain safeguards now so that problems don't occur at all, then when the problems do occur, safeguards that will be implemented after will be far, far worse. And the point I'm talking is like, if we just say, oh, freely, go, go, go, go, go, and we're just, you know, 100 miles an hour. The problem is, is that if a major incident were to occur, like an AI agent just going completely rogue and just completely going off the rails, governments could enact some draconian laws that would push things further back than if you get ahead of it. And that's the point that I think most people are missing. But yeah, like I said, I'm not one that wants to slow down technology at all. But I do say that it is at least wise to ensure that we see things before they happen. Now, of course, there was also the Microsoft CTO actually speaking about regulation, and he does have some good points about that.

The same way that when you're developing a technology, there's this balance between, you know, benefits and harms that you're trying to weigh. There's also, like, positive and negative for regulation. So, like, if you regulate too aggressively and too early, like, you may curtail the development of technology that can be unbelievably beneficial. And so, you know, the thing that I think, again, this is all about societal equilibrium here. Like, we have to do what we're doing in a way where we have enough societal trust, where we can develop the technology and let people lay their hands on it and evaluate whether it's good or bad, bad themselves, so that we can inform the regulation that actually has to happen. If you don't do that, you're kind of regulating into a vacuum. You're sort of imagining what a future might be and you don't even know that it exists. And you sort of risk two things, like regulating the wrong thing. So you think very thoughtfully about regulation. You pass something and it doesn't produce the effect you intended or you inhibit something that you could be very, very valuable.

So, yeah, it's very, very interesting on how this conversation is going to go. And I think, you know, it is a hard job to actually regulate this technology because it's so dynamic and there's so many elements that, you know, it's nearly impossible to get everything right. Now, Mustafa Suleiman actually also did a interesting TED talk with, in which he actually spoke about the existential risks of AI. And I think I'm going to play just one minute from that clip because I think, as always, it's good to remind yourself of what's actually coming in the wave. I mean, I've read his book. It's actually insightful on how the future is dynamic and changing in terms of this new digital species, which he called it.

So what is this? What are these AIs? If we are to prioritize safety above all else, to ensure that this new wave always serves and amplifies humanity, then we need to find the right metaphors for what this might become. For years, we in the AI community, and I specifically, have had a tendency to refer to this as just tools, but that doesn't really capture what's actually happening here. AIs are clearly more dynamic, more ambiguous, more integrated, and more emergent than mere tools which are entirely subject to human control. So to contain this wave, to put human agency at its center, and to mitigate the inevitable unintended consequences that are likely to arise, we should start to think about them as we might a new kind of digital species. Now, it's just an analogy. It's not a literal description and it's not perfect. For a start, they clearly aren't biological in any traditional sense. But just pause for a moment and really think about what they already do. They communicate in our languages. They see what we see. They consume unimaginably large amounts of information. They have memory, they have personality, they have creativity. They can even reason to some extent and formulate rudimentary plans. They can act autonomously if we allow them. And they do all this at levels of sophistication that is far beyond anything that we've ever known from a mere tool. And so saying AI is mainly about the math or the code is like saying we humans are mainly about carbon and water. It's true, but it completely misses the point. And yes, I get it, this is a super arresting thought, but I honestly think this frame helps sharpen our focus on the critical issues. What are the risks? What are the boundaries that we need to impose? What kind of AI do we want to build?

Now, there was also actually another TED Talk, and I got to be honest, this was one that I truly enjoyed the most. It was Daniela Russ, the robotics and AI pioneer, and she actually spoke about how AI will step off into the screen, into the real world. Now, this two minute clip showcases how they're managing to do that. And their research is truly profound in the sense that, you know, it seems that we're going to get hit by this wave once it becomes full scale. So, of course I'm going to leave a link to this in the description, but trust me when I say the full TED talk is honestly just mind blowing.

Created a kitchen environment where we instrument people with sensors and we collect a lot of data about how people do kitchen tasks. We need physical data because videos do not capture the dynamics of the task. So we collect muscle, pose, even gaze information about how people do tasks. And then we train AI using this data to teach robots how to do the same tasks. And the end result is machines that move with grace and agility, as well as adapt and learn physical intelligence. We can use this approach to teach robots how to do a wide range of tasks. Food preparation, cleaning, and so much more. This means we can make almost anything we imagine today's AI has a ceiling, it requires server farms. It's not sustainable. It makes inexplicable mistakes. Let's not settle for the current offering. When AI moves into the physical world, the opportunities for benefits and for breakthroughs is extraordinary. You can get personal assistance that optimize your routines and anticipate your needs, bespoke machines that help you at work, and robots that delight you in your spare time. The promise of physical intelligence is to transcend our human limitations with capabilities that extend our reach, amplify our strength, and refine our precision, and grant us ways to interact with a world we've only dreamed of. We are the only species so advanced, so aware, so capable of building these extraordinary tools. Yet developing physical intelligence is teaching us that we have so much more to learn about technology and about ourselves.

Now, another thing that I did want to talk about as well, just quickly, is that a lot of people online are saying, okay, there's an AI bubble, an AI bubble. And they're taking screenshots of all the top AI companies and stating that it's crazy, it's crazy, it's crazy. But the point is that many people aren't missing. And if you don't know, I look at stocks all the time and crypto and all that stuff like every single day, and this just doesn't make sense. And when I say it doesn't make sense, I don't mean it doesn't make sense fundamentally. I mean, they're just taking and cherry picking stocks from the NASDAQ that are literally AI stocks. And it's like, this doesn't make sense. The Nasdaq, the entire of the Nasdaq is falling. Okay, if you cherry pick some AI stocks that just so happen to be in the Nasdaq, that doesn't mean an AI bubble is popping. And trust me, guys, when there's an AI bubble, what's going to happen is you're going to see random companies that slap AI in their name and they're going to get billion dollar valuations, hundred million dollar valuations overnight, and they are going to be ipoing and there's going to be so much money flooding the market. That stage hasn't happened yet. And trust me, guys, we're a little bit far away from any kind of AI bubble. So these kinds of tweets are a bit disingenuous. And I think what they do show is something else. I think they're indicative of a trend where people just want AI to fail. And, and I understand why. It's because I don't think people are going to be excited for a technology that's going to replace them when it's supposed to augment them. So, I mean, I guess this is something that people are currently expressing on the timeline. And of course one of those stocks is of course Tesla. But what's crazy about that is with Tesla, Elon Musk is apparently going to be adding Grok into Tesla, which is going to be rather fascinating because adding an LLM to an autonomous car might give it certain other capabilities. So I'm wondering what this development is going to look like, because I'm pretty sure Tesla are going to be having some insane stuff from the year ahead because it seems like Elon Musk is back in charge and is really trying to put the company into shape. You know, it seems that he's been firing a lot of different organizations within Tesla and I'm pretty sure he's all in on AI. From the few tweets I've seen about him, you know, stating that, okay, I'm all in on robotics now and stating that he's going to war with OpenAI and figure I think we're going to get some really, really crazy stuff, especially considering the fact that he said Grok 2 or Grok 1.5 is going to exceed everything on all benchmarks, which means that, you know, he's really, really taking this seriously because the AI race is one you don't want to be behind in. Now there was this tweet that got over 1.2 million views that was essentially stating that, you know, Zuckerberg is essentially stating that there won't be any crazy, crazy increase in capabilities from the GPT series in terms of GPT4 onwards. And he doesn't believe in states where a world of 100x improvement from GPT4 are possible, or that AGI is possible within a short time frame. Now, of course, this comes from his interview with Dua Quesh Patel and it was definitely an eye opening interview into how Meta are tackling the AI game and what their plans are. But I don't think this is true. And of course Dr. Jim Fan, Nvidia's senior AI research scientist, said that even if GPT5 plateaus, robotics hasn't even started to scale yet because embodied intelligence in the physical world will be a powerhouse for economic value. And remember that LLM is not all of AI. It's just one piece of a bigger puzzle. And that's of course something that I tried to reiterate quite a lot. Whilst LLMs have taken the main Stage, they're not the only thing that goes on in AI. And trust me, when you look at all of the research papers that keep going on, it's kind of crazy how much progress is being done in AI. Like there are literally so many research papers published every single day, it's actually pretty hard to up with. And of course we had OpenAI employee or OpenAI leaker or OpenAI insider. Honestly, the identity of Run is one that is still a mystery. And he says, are we, you know, he, he responds to this tweet saying, are we plateauing Zone? Not even close. So it seems that, you know, whatever's going on internally, OpenAI seems that the plateau that many people are thinking of just isn't there. And I think the reason that most people think of a plateau is because GPT4 hasn't been basically surpassed in, and I mean meaningfully surpassed in around a year since its release, a year and a bit since its release. And because of that I think people are now seeing that these other AI labs are of course plateauing in trying to catch up to the model. And because of that people think that there is going to be some kind of plateau. But remember that OpenAI are quite ahead of the market and I guess we're going to have to wait to the next, next cycle to see where things truly are. Now there was also this crazy thing released by Adobe and I think the Adobe suite is honestly going to be purely, purely video creation stuff and I mean it's going to really change the content creation game. But they released video Gigagan towards detail rich super video resolution and you can basically do that crazy thing from movies where you say computer enhance and then you get a super high resolution version and this is an 8 times increase in the level of quality. And I'm wondering if this is also potentially going to be applied to the end of AI video generators where you just use something to generate a low quality video image and then essentially you can just upscale it with this. I mean taking that video and then upscaling it to that is pretty, pretty intense if you ask me. And I mean whilst to some people this might not seem impressive, I think futuristic stuff simply becoming, you know, real every single day is something that we should take into account because I remember so many times people like, oh, enhance on a computer, you can't just do that. But you know, clearly, I mean, the horse, I mean that is pretty, pretty intense. And like I said before, the applications for this are going to be super, super interesting. You know, those videos on the cameras on the cctv. When you're looking at a guy and it's like you just can't see him, you just can't make out any of his facial features. Well, I guess if you're trying to commit crimes, you might as well stop now, because cctv, those low resolution footage where it's like, have you seen this man? And it looks like footage from a PlayStation 2 game. Those days are going to be over. So, yeah, definitely some fascinating research.