To be operating systems. They want them to be software platforms rather than document displays. And to do that you can brute force a lot of things but it is certain point to get the performance you want at the level of programming skill you want to pay for you need to start doing shit like that. Otherwise the apps get too slow.

Web apps are so much easier to monetize and so much cheaper to support that companies are all over them. I remember back in the days supporting a desktop app and needed three guys with decent tech skills and a small team of level 1 for a few thousand users. Switching that over to a web app got rid of all the tech guys with decent skills and the level one are now just password resetters.

It comes as absolutely no surprise that when you can execute code on a device, you may be able to gather fingerprints from other code running on that device. You get very little from that though. Basically the only useful case is if one website generates a specific usage pattern, another one may be able to detect that. But you could have gotten the same thing by just having the two sites communicate directly. Yes, there is a covert channel. No, it is not one that matters.

No. Diode "shot" noise is about 50% quantum noise at 5.6V. It is just as "perfect", just several orders of magnitude cheaper and less flashy. Can be done for about $10 in hardware. Obviously, quantum noise being "true" random is a theoretical model, not an established property of physical reality. That limitation also applies to this non-story.

Simple: This is not better. The whole thing is hot air and lies.

First, a reverse PN diode delivers about 50% quantum-tunneling avalanche noise at 5.6V. That is quantum noise. Combine a few samples the right way (crypto-hash), and you have "perfect" randomness. Second, that quantum-noise is "true" noise is the theoretical (!) model. It just means "we have no clue how it works". It does not prove or imply perfection. And third, there is zero need for this. Security and cryptography just needs unpredictable data, not random data. A competently seeded CPRNG delivers that.

tl,dr: This is nothing but hot air.

Somewhere, sometime, someone said "troubleshoot", and the US heard "shoot the messenger", and never looked back. Well how's that working out by now...

I've seen people make that argument before, that there is no difference. You can make that argument if you want, but if you can't provide a viable candidate then you're not helping anything.

Let me be extra direct here. I have never seen a candidate in a race I could vote in that I agreed with 100%. Frankly anytime I meet someone who tells me they agree with a politician 100% of the time I tend to want to tell them they need their head examined (unless they are themselves a politician, in which case doubly so). Even politicians like Bernie Sanders I don't agree with 100% of the time.

Complaining about the rules is not productive. We've seen what the SCOTUS will do about such arguments, we can't get help from them either. The only way to advance is to support someone who will actually do something useful. Pretending that they all do the same shit doesn't help either; there are meaningful platform differences. Just because Drumpf has held every position on every issue doesn't mean he doesn't stand for anything, either - he's actually followed through on some of his (very worst) initial promises.

>As for the comparison to AI, the problem is, AI *must* be told what to do. It won't magically grow into a "mature developer." That's not a natural progression. It always assumes that the prompt accurately describes what it should do. It has no way to know that the prompt was wrong or incomplete in the first place.

This is wrong. You seem to be unaware that current sycophancy in mainline models is a specific choice made in AI model weights to maximize people returning to the model.

It's highly likely that one of the solutions that will be used in specialist fields where rejection of the input if it's insufficient in some critical way is reduction in pro-sycophancy model weighing. I.e. model will actually have a much greater ability to tell you "I can't do that Dave" and then explain why it can't do it.

Some narrow specialist models already do this through ControlNet style "AI that corrects and guides human AI prompts for optimal outcomes", where it will tell you in case of some of the common prompting errors before passing the input to the worker model.

Take the rate of growth we currently have and subtract the growth increase from AI data center bullshit and we are in a deep deep recession.

Now if that AI data center bullshit created jobs that wouldn't be true but it doesn't. All it does is suck down water and electricity and raise the prices of both. I guess the technically does also raise the ambient temperature in the area by around 5 degrees causing all sorts of environmental problems so there's that too I guess.

And I don't mean shave the whales problems I mean less rainfall problems.

Look up what a heat island is.

You're thinking is too lateral and simplistic. You've been trained to do that by somebody. You should be asking who. It's not a limitation of intelligence it's social conditioning. You need to become aware of it so you can break that conditioning

Indeed. The actual reality is that general LLMs are not useless, but they are not a game-changer either. They are a gradual improvement of some things. And they are not yet at a point where they can actually get profitable and will not be there for a few decades at least. Hence a total economic collapse of the space for general LLMs is inevitable and will happen some time within the next few years. Depends on how much stupid money the LLM makers can get on top of what they already burned.

I think he is saying that the free market will somehow take care of this because a 24/7 surveillance State controlled and built for private individuals and companies won't be profitable.

Basically the grandparent views everything through the lens of profit. And that's the mistake they are making.

In addition to profit people want power. And a 24/7 surveillance state owned and operated by billionaires will give those billionaires power. The power they get from that is well worth the expense. Especially because at the end of the day all their money is stolen from you and me so it's not like they're spending their own hard-earned cash. Because none of those fuckers ever earned a dime in their lives.

But CEOs are very often disconnected from reality and typically have "visions" instead of understanding. Makes them east prey for typical commercial LLMs. I am convinced that these LLMs were specifically designed to attack that vulnerability.

Huang's billions are probs not going to impress the Chinese, who are investing multiples of that. The Chinese also consider developing their own (AI) semiconductors a national priority, with the US of the last 10-15 years having proven to them conclusively that Western supplies can not be relied upon. Their homegrown stuff is not as advanced as Nvidia on performance/watt yet, but it's cheap (especially compared to the multiple orders of magnitude inflated prices of Nvidia) and it works, and has built an AI industry in China that rivals the US.

So this ship sailed for Huang some time ago. I'm reading this announcement more as a "screw you guys, I'm going home" to Trump. The Orange Man has had Huang lick his boots, messed with his business with his trade wars and sanctions games, and paraded him in China now... and all for nothing, Trump has zero leverage on China, he has nothing to offer to China, and thus, nothing to offer to Huang either. All mouth and no trousers. Why would Huang hang around ...

Taiwan is where the semiconductor industry is at, and Taiwan is where Huang needs to be, too. The tech is there, the engineering is there, the manufacturing is also there, even if they branch a little to the US. The US semiconductor industry otoh is moving out of inertia by now, unable to keep up with the volume, and therefore, investment, and r&d of Taiwan. And all of that is before you take into account the US is basically in war with higher education by now, which does not bode well for any hopes of long term brain availability.

You are confused. And you should look up "argument from authority" and why a simulation is not reality and a fundamental principle within a simulation does not make that principle fundamental for the thing simulated.

Sorry, but in a somewhat fuzzy sense, "race" _is_ real. It just does not mean what people that try to elevate themselves by their racial membership think it means. For example, Science says no impact on intelligence. But try to ask a skin-doctor whether he can treat black and white people simply the same. (Obviously, this is more complex than just black vs. white.) There are countless medical effects from racial membership. Note that racial membership can be stronger or weaker or mixed. But it has medical effects that need to be taken into account for good treatment outcomes.

So claiming "race is not real" is a lie. Claiming that some races are dumber or less capable of "inferior" regarding mental capabilities is also a lie. On the physical side, hilariously, the genetics of "white" is generally the weakest as it is the most specialized. (This is, again, simplified.)