What everyone wants is a user interface that just does what the user wants, without forcing the user to figure out which remote-button or unrecognizably-abstract onscreen icon to push to make it happen.

If Samsung's AI can implement that, e.g. by listening to free-form English commands and reliably acting on them in a useful manner (a big if, but not inconceivable), it will be popular.

Easier said than done, if most of your content is being streamed over the Internet.

Even if this project was technically feasible (which, to be clear, it totally isn't), it wouldn't ever get built. If you think the fearmongering about nuclear fission plants is bad, imagine the fearmongering that would break out as soon as someone seriously proposed implementing this. Hell, look at the first comments posted to this page, and this is on "news for nerds".

It wouldn't matter that there are straightforward safeguards to prevent accidents. The first, last, and only thing the rubes (and therefore the politicians that pander to the rubes by making them afraid of things) would hear is "death ray".

The only realistic accomplishment this project would have would be the quick passing of laws to make this sort of project illegal.

Remember when if you didn't want a shitty "upgrade", you could choose to just not install the new version of the software? When such "updates" weren't forced down your throat because, even on Windows, because you owned your computer? When drastic changes to operating systems that affect how people use them was limited to major releases?

Microsoft remembers, and they hated it. They are so close to customers being fully accustomed to forced updates that they don't want. The next step is to convince users that these constant (forced, unwanted, unnecessary) necessary updates are just so expensive for Microsoft to produce, and so obviously people need to pay a monthly fee for Windows. I'm a little surprised that didn't come along already with the "Windows 11" moniker, but I guarantee it's coming.

What do you want it to do with memory when there are no programs running and the system is idle? Windows is pretty aggressive at using memory for the file system cache, for example, and it will prefetch programs and files that you use often into memory so they're available instantly when you try to access them.

This is a good thing, not bad. All of this memory can be instantly dumped if a program suddenly requests a bunch of memory for private use. Memory that is not being used for something is wasted memory. It doesn't go bad by being used.

But obviously memory demand bloat is a growing problem. A few Electron apps and a dozen browser tabs will set you back more memory than most computers had 10 years ago, which really is pretty bonkers. I wouldn't be surprised if some Windows components had followed a similar trajectory.

Sooner or later some bright individual will realize that instead of dumping that excess electricity, they could be using it to mine Bitcoins, or electrolyze hydrogen out of seawater, or (insert your favorite low-priority process here).

Curtailing electricity generation in an era where electricity is in such high demand is throwing money away.

Could be, but I'm not sure how you define "scarcity" in the context of things like Bitcoin mining, which is literally a contest to see who can burn the most electricity in a given period of time in return for money. In that sort of competitive scenario, no amount of electricity will ever be enough, because adding more generation will simply up the power requirements to mine the next bitcoin; hence electricity is always scarce no matter how much is generated.

AI isn't quite as bad as that (since presumably they will someday reach a point where AIs are "smart enough" and they no longer feel the need to scale them up any further), but at the moment it's rather similar -- a contest between OpenAI and Anthropic to see who can run the most GPUs at once, in the hope that something useful will come out of it.

With all due respect to you and your agent friends, there are more economical ways to generate power :)

The silver lining (I guess) is that the more grid prices rise, the more incentive people will have to install solar panels at home to avoid paying grid prices as much as possible.

I don't think his take was that AI startups will all fail; rather it was that AI (as currently implemented) costs a lot of money to run, and that cost is currently being paid by investors rather than by customers.

But the investors won't be willing to subsidize cheap/free AI forever, so at some point the cost of these AI services will rise to somewhere higher than the cost of the electricity and hardware it takes to provide them, and it would be wise to take that into account rather than assuming that today's loss-leader pricing models will continue indefinitely.

Of course it's also possible that some breakthrough will occur that allows AI services to be implemented using fewer resources; but unless/until that actually occurs, it would be reckless to plan a business model around the hope that it will happen.

Lying requires intention, and while it's certainly possible to program an AI to intentionally lie, that isn't what happened here.

What happened here was that an AI's training data included fraudulent forum posts, and it didn't understand that they contained misinformation, and it included the fraudulent misinformation in its summaries later on. A real human could easily make the same mistake when googling for a phone number to send to his boss, and nobody would accuse him of lying. Carelessness or naivete, maybe.

Don't be so harsh on Mr. Zuckerberg, he provides a valuable public service by renaming his company after the trend on the day the trend dies.

When Zuckerberg announces that Meta will now be known as AGI (or whatever), that's how you know it's time to sell your shares in OpenAI/etc.

That, despite the first manned hot air balloon flight having taken place in 1783.

I have my doubts about an AI's reliability over the course of 400 years. Even if the hardware holds up, by the time the ship gets there, the AI may moved on to different plans than the ones its creators had in mind :)

That was true back then; is it still true now, a number of years, and many expensive lawsuit-settlements later? (I suppose it's possible that Toyota has learned nothing from the experience, but that doesn't seem like the most likely outcome from a company that generally prides itself on quality and reliability)