It just moves the source of the sales -- if the datacenter (AI or not) bubble pops, then all the chip-fab companies that are doubling down on RAM production for these datacenters will have production capacity they no longer have a market for at inflated prices, so rather than have to take a complete writeoff on their expanded production lines, they'll likely start pushing RAM at fire-sale prices in order to recover at least some of the money they'd poured into new production lines in the "line goes up" belief that datacenter RAM demand would be forever.

More importantly, we keep seeing stories touting X number of megawatts of generating capacity for a new wind or solar farm -- but never a word about how much power that installation is expected to generate averaged across a year; the "installed capacity" tells you how much power it will generate under ideal conditions, but you could have a solar farm with a generating capacity of 1GW, and it's going to produce nothing at night. Similarly, a wind farm will produce no power when the wind speed is below the minimum for turning the turbines, or when the wind speed is high enough that the turbines have to be feathered to protect the generators from overspeed. Wind and solar installations typically have an average production close to a third of their installed capacity, so that 160k MW of solar and wind capacity would be each producing less than 55k MW in practice over time -- and require a considerable GWh storage capacity to supply demand when the wind and solar installations are not producing. If you have a solar farm that produces power at a rate of 100MW over the course of a day, half of that power would need to be diverted to charging battery storage to produce power at night when the solar farm isn't producing power, so the deliverable power of an intermittent source like wind or solar will be even less than it's actual production, because it needs to divert energy to the storage system that will back it up.

Now if only the AFP story would kill the seemingly-endless "Climate change is making [insert location here] warm up twice as fast as everywhere else" stories that, if you put them all together, would imply that the entire world is warming twice as fast as the entire world... demonstrating how idiotic these stories are.

Meta has run out of new ways to monetize information they're collecting from their own apps, so they're branching out into collecting all your activity across the net to give them new information they can aggregate, process, and sell.

Well, of course Meta is going to remove it from the app; they don't want the common users to be aware of how much data their smart glasses are collecting and sending to Meta's servers to be processed and monetized. Now all the biometric data will simply be silently uploaded to the cloud, where Meta, having virtue-signaled their 'rectitude' by removing it from the app, will be publicly above reproach.

The announcement follows in the grand Soviet tradition of gross downplaying the aftermath of combat, going back at least to WWII, with the battleship Marat being described as "slightly damaged and later raised" after a Stuka attack. (for those unfamiliar with the terminology, 'raising' a ship means refloating it after it has been sunk)

And it will roll over on its back, choke, and die if you tell your firewall that it's not allowed to connect to the Internet, because of course training the remote AI servers is essential to the local operation of AI software that doesn't use them.

...which just means that the division that trawls the users of their VR headsets, AR smart glasses, and metaverse software for personal information to sell to other companies has its budgetary category fully separated from the Reality Labs division. Scraping user account data and activity for salable information and profile data is almost pure profit, since it's information the users are giving out freely.

The UK was second country in the world to end slavery, following the US? I guess Parliament's Slavery Abolition Act of 1833, through some deep and mysterious time warp, didn't actually occur until after Lincoln issued his Emancipation Proclamation in 1863. And Upper Canada's 1793 Act Against Slavery, which banned the importation of slaves into Canada and freed the children of slaves upon reaching the age of 25 -- the first legislation in the British Empire freeing slaves -- didn't exist either?

I guess the NTSB is unaware of the fact that, since the development of "talkies" as sound-on-film recordings, the audio track of movies was recorded as an image of the audio waveform on the film alongside the movie images. In 1900, Ernst Ruhmer was able to record audio as an image on film and then recreate the original sound from the image. Other developments advanced the recording of sound on film, but sound recording and reproduction technology was not adequate until Lee De Forest was awarded patents in 1919 for recording audio as a track alongside the images of a moving picture recording, with the first commercial screening of a motion picture utiltizing sound-on-film taking place in 1923. So the NTSB, being 'unaware' that you can reconstruct sound from a pictorial representation, is more than a century behind current technology. And this is the organization that is responsible for keeping our transportation safe? Should we be seeing new buggy-whip safety regulations being released in the future?

You can also put the AI-disabled search in your search engine list by defining a new search provider, say, "GoogleNoAI", with the URL for the search command being "https://www.google.com/search?udm=14&q=%s", then make it your default search engine; after that, when you use the search box in the browser, it will default to the AI-free results.

And if establishing their own Linux distro enables them to integrate Copilot into the distro, so they can hook more people into paying for their service, that's just (to them) good business.

The "as much as possible" phrase is an unclosable back door; if you want to keep data centers from using bigger wallets to pull power away from residents, you prohibit data centers from being connected to the local power grid; if a data center needs power, it has to provide its own.

Agreed; I don't think that this would apply to something like an MMORPG, where if the bill was carried to a logical extreme, would require the publisher to keep at least one server alive indefinitely, with server maintenance and other associated costs in perpetuity. In particular, the carveout listed that would exclude "completely free games" and games "offered solely for the duration of a subscription" creates some confusion, in that a game like an MMORPG that is free to download and install, but has both a free-to-play option and a subscription option, could be argued to not fall clearly into either carveout and therefore obligates the publisher to keep the game alive. If the courts (where something like this would inevitably wind up), I predict that what will happen is that we'll see publishers operating an MMORPG type game that requires a server to function making the game completely free as a precursor to shutting it down, thereby putting it in the "completely free game" exclusion and protecting them from having to keep a server live indefinitely.

I want to see Google's HIPAA disclosure document for their Google Health app. If it does not clearly spell out that any PHI stored in the app is contained in the app and will not, as required under HIPAA, be shared within Google for other uses, or sold out of Google entirely, then it will be a warm day in Niflheim before they see me using their product.