Intel and AMD have implemented these improvements with Lunar Lake and Strix Halo. You can buy an x86 laptop with Macbook-like efficiency right now if you know which SoCs to pick.
This just isn't true. Yes, Lunar Lake has great idle performance. But if you need to actually use the CPU, it's drastically slower than M4 while consuming more power.
Strix Halo battery life and efficiency is not even in the same ball park.
If you look the battery life benchmarks they did at around the 5:00 mark in the third video, you can see that it achieves similar battery life compared to the an M3 Macbook in typical day-to-day use. This reflects the experience most users will have with the device.
It's true that the perf/watt is still a lot worse than the latest gen M4 under heavy load, but it's close enough to the M1 and significantly better than prior laptops chips on x86.
It is a first gen product like the M1. But it does show the ISA is not as big of a limiting factor as popularly believed.
If you look the battery life benchmarks they did at around the 5:00 mark in the third video, you can see that it achieves similar battery life compared to the an M3 Macbook in typical day-to-day use. This reflects the experience most users will have with the device.
No. The typical day to day use of LNL is significantly slower than even M1. LNL throttles like crazy when on battery in order to achieve similar battery life.
I think anyone could have seen this from miles away based on:
1. China is going to own the EV market. The only thing that can temporarily stop Chinese EVs from dominating in every country is massive tariffs. But this is not a long-term strategy.
2. Sentiment is low on EV due to Trump
I think the problem with Musk's strategy is that self-driving will also become a commodity over time. There's no reason that only Teslas will be able to drive themselves. Further more, I don't see why China's robotics won't do the same to Tesla's EVs. In fact, I believe China's robotics are world leading at the moment.
That can be said about everything so does it play out in real? For e.g. phones have been labeled as commodities a while ago. But, are they?
Edit:
Self driving is a super specialized tech and it’s still not fully developed yet. A lot of weak areas. I don’t see it getting commoditized in the next 20 or 30 years.
The real opportunity would be in monetizing the time that people would be getting back while not driving.
The difference is that in the phone market, you have one premium brand that makes all of the profit from hardware sales (Apple), one that makes money by manufacturing hardware for the non premium brands (Samsung) and one that gets a cut from all of the manufacturers for licensing (Qualcomm) even when they use non Qualcomm chips.
Tesla’s brand is toxic and seeing declining sales, China is doing the manufacturing, and no one is going to license Tesla’s inferior unproven technology compared to Waymo.
It's a potential commodity in that there is potentially no network effect stopping people from using any competitor that offers a cheaper ride, unlike how when you're an iPhone user you get locked into Apple's ecosystem. But then there's the thing where despite having had 18 years to catch up, nobody makes phones (including software) better than Apple.
One potential moat is just the amount of data from real drivers that Tesla use to train their models via imitation learning. If this turns out the be important and needed for a general solution (which I believe it will), then only companies that manufacture cars at scale can hope to compete. And at this point, only Chinese companies are forward looking enough to put the right hardware for self driving (and the ability to collect training data) into their cars by default. Tesla has the vertical integration that makes this whole thing much easier: they make the cars, the inference compute, the software AND the training clusters. Can you imagine GM or Ford building a GPU cluster for a couple of billion?
1. Have an option for long-term rental fleet - on a per-day or per-month basis. Provide a 40+ inch screen in the back for people to use the car as an office on wheels. Equip it with a Super high-speed network. Launch an app store on that screen or charge a premium for other apps.
2. Provide an option to get unlimited booze, food, essentials by partnering with Food delivery apps
3. Convert big SUVs/RVs into self-driving vehicles to enable them to be rented by families for summer picnics, long travel, and wedding trips. Many people still dislike air travel, especially given the current issues with flight delays, baggage limitations, and other uncertainties. Any alternative would be a huge win. Imagine travelling for 12+ hours overnight while having the comfort of a home.
4. Make it possible to deliver food, drinks, and essentials anywhere - via drones or other partners, if they've rented your self-driving vehicle.
5. Have super-comfy interiors just like a private jet. Of course, people would love it.
There's no moat in anything you listed. If you want a mobile app you need either Apple or Google's permission to get it. The roads are a public good, they by law cannot be monopolized by a single company (or two).
Yes, yes they are. There are new (second-tier, semi-competitive) phone manufacturers popping up almost every year, e.g. Nothing[0].
Once self-driving has been generally "cracked", with the normal mobility of talent, most other car manufacturers will catch up on a timespan that's too small for the first-mover to completely dominate the market with that alone.
> There are new (second-tier, semi-competitive) phone manufacturers popping up almost every year, e.g. Nothing[0].
Good for them. Meanwhile, the vast majority of smartphones people buy, and the vast majority of profits from smartphones, are coming from Apple, Samsung, OnePlus, Xiaomi, and maybe Sony. There are no signs whatsoever of the market commodifying.
I think it is more likely that self driving in the way people have been led to believe would happen never will. Not from Tesla and not from anyone else. Which is bad for Tesla because they have put too many eggs into that basket.
The only thing they have left is the car and they are falling farther and farther behind.
Well that's pretty obvious now that all the technology transfer needed for Chinese companies to catch up have been realised. Realistically though the CCP will want to keep the lead in market where they have crushed the competition with their state sponsored champions; this was the case with photovoltaic panels, and this seems to be the case with their car (computer-on-wheels) manufacturer.
That’s what the now retired leadership wanted though, they got their $10 million paycheck and comfortable retirement in the west, now it’s time for current leadership to clean up.
Also I wanted to add that LLMs (at least free ones) are pretty dumb sometimes and do not notice obvious thing. For example, when writing tests they generate lot of duplicated code and do not move it into a helper function, or do not combine tests using parametrization. I have to do it manually every time.
The new CEO of Intel has said that Intel is giving up competing with Nvidia.
No, he said they're giving up competing against Nvidia in training. Instead, he said Intel will focus on inference.
That's the correct call in my opinion. Training is far more complex and will span multi data centers soon. Intel is too far behind. Inference is much simpler and likely a bigger market going forward.
I disagree - training enormous LLMs is super complex and requires a data centre... But most research is not done at that scale. If you want researchers to use your hardware at scale you also have to make it so they can spend a few grand and do small scale research with one GPU on their desktop.
That's how you get things like good software support in AI frameworks.
I disagree with you. You don't need researchers to use your client hardware in order to make inference chips. All big tech are making inference chips in house. AMD and Apple are making local inference do-able on client.
Inference is vastly simpler than training or scientific compute.
Funny enough, perhaps the AI threat is what will bring China and US back on friendlier terms. If there is a greater threat than both, they'll work together.
I think signing an AI treaty between the two countries will do a lot for the world.
I don't see it as hype to be honest. It's a real problem we need to solve now.
Also, it seems like the letter was faked?
Anyways, it doesn't matter if it's faked. It does make people notice that it's a real problem now.
General consensus amongst data scientists of when each AI breakthrough will happen have almost always been wrong. AI breakthroughs tend to happen earlier than what the consensus predict. I don't see AGI predictions as any different.
We need to solve AGI now? Don't get me wrong, it would be an incredible scientific breakthrough, but I cannot see how AGI would solve our current societal and human problems. In fact, it would just upturn our entire world.
AGI is a tool we want, not the solution we need.
And I agree with OP anyway. It is pure hype. Thinking we are closer to AGI because of LLM is just like thinking we are closer to the Moon because we have conquered the Everest peak.
How do we make sure no country builds an underground nuclear weapon making facility?
I don't actually know the answer to the above question.
But at least for chip making, you need hundreds of different highly advanced technology suppliers and chemicals to make a chip. I would think that making an advanced chip making facility is actually much harder than nuclear weapons. Many countries have the ability to make nuclear weapons, but no one can compete with TSMC, Samsung, and Intel right now. Even then, all 3 rely on ASML machines.
That said, I suppose countries will slowly see the value in chip manufacturing and invest heavily into it in order to train more powerful AI models.
I'm not suggesting that this solution is perfect, but it does seem to buy us significantly more time.
For Siri specifically, Apple has to be extremely careful. Their carefully crafted public relations could shatter in an instant of their LLM spits out crazy answers like Bing's chatbot.
Nobody has a big enough test set to A/B test every combination of elements for a web page, a product feature set, or even a conference presentation. The combinatorics are just unwieldy.
Someone has to have multiple good ideas and the ability to carry them out properly before an A/B test is even valuable. Otherwise you're measuring one uninformed random change against another. A/B confirms whether you've succeeded in improving something. It can't really suggest what to try next.
I really don't think Waymo needs Lyft or Uber to dominate the ride hailing market in the future.
If they can make a better self-driving car than anyone else, they will win the ride-hailing market. People have no loyalty to Lyft or Uber. They will use whatever is the cheapest/most convenient.
Strix Halo battery life and efficiency is not even in the same ball park.