How much in additional taxes or airport fees are you willing to pay for that?

Data center use of power is very high, but water is often being overestimated. See https://theconversation.com/ai-has-a-hidden-water-cost-heres-how-to-calculate-yours-263252 for a good basic discussion of the issues involved.

DeepSeek is designing a tensor processor. The instruction set is very simple. Mostly just lots and lots of parallelized low-precision multiply-and-accumulate.

LLMs and humans are both unreliable, but make different mistakes.

Use one to check the other.

Disclaimer: I have used Mechanical Turk many times, but not recently.

Except the summary says that earlier acquisition is more effective than learning "recently".

Most children don't learn a second language by choice, but because they are in a multilingual environment. My daughter is bilingual because from birth to age 10 she shared a bedroom with her monolingual Mandarin-speaking grandmother.

I had an RPN calculator in high school, so that made learning Japanese much easier.

When I was in the military, I was deployed to the Western Pacific. I learned Japanese and Tagalog so I could talk to girls.

Still trying to extort money from IBM after all this time. Nothing like a business model made up entirely of rent seeking.

Also a separate reply to note a specific detail: > 12 and 21 You mean 15 and 21. There's no way Shor's algorithm would factor 12. It only factors odd numbers.

What evidence? The only evidence you've provided was a vague wave about Shor's algorithm, which we're in agreement with. You haven't attempted to give anything else remotely resembling evidence, like a link, a citation, a source, anything here. And then after all the insults here and in the other thread where you and I discussed these issues you think the problem is me being an asshole? How hard is it rather than insult people to just give evidence that I and everyone else in this thread can actually look at, or for you to go back to the prior thread where we were having a conversation and continue that?

Of course they will.

Clients need to do their research, choose the tool they want, and then choose the consultant to help them implement it.

That's the way it works. Microsoft isn't the right partner to help you implement a solution based on Google Gemini. Duh.

Or this could be an example of Jevons Paradox .

As AI makes employees more productive, and thus more profitable, would you fire them or hire more?

People are really quick to accuse things of being AI. I've lost track of how many times on Reddit I've written two or three paragraphs with citations and someone responds accusing it of being AI. Apparently the bar is very low, and seems even lower if one is arguing for something they disagree with. But I've also had this happen with short stories. I had someone claim a short story I wrote was AI generated when the story was from 2019 and thus predated any AI that could write more than a few sentences.

I'm not sure why you think that. It wouldn't be surprising if Israel has access to Signal App chats, and other things. But you don't need quantum computers for that. The vast majority of penetrations of secure systems involve finding implementation bugs, or infecting machines thought to be secure, or social engineering, or given how the beeper operation went, possibly just compromising phones at their source before they even get to the targets. And we have good evidence that the governments have not yet built quantum computers on a scale that can decrypt anything substantial. There are two major lines of evidence.

First, while we've seen some government investment in quantum computing, we're seeing scientists and engineers there publish in the open. When they get really close, some of that will start getting classified. That's happened with a bunch of techs before. Georgy Flerov was able to detect that the US was working on an atomic bomb because all of the apparent public nuclear research stopped. Similarly, a sign in the 1970s to the US that the Soviets were *not* working on stealth aircraft was that the work on related ideas such as the work by Ufimstev and related work had not been classified https://en.wikipedia.org/wiki/Pyotr_Ufimtsev.

Second, the US and its allies have built giant data storage facilities and are still expanding those. The Utah Data Center is the obvious big example https://en.wikipedia.org/wiki/Utah_Data_Center but other governments have built similar smaller facilities. This doesn't make much sense if one has quantum computers. But it makes a lot of sense if one is expecting to get quantum computers a few years from now since it lets one do the strategy of storing massive numbers of messages now for later decryption https://en.wikipedia.org/wiki/Harvest_now,_decrypt_later.

There is however one argument in your favor. If one looks at the history of declassified material from the NSA, material from GCHQ (the British analog of the NSA), and looks also at declassified Soviet material, anthe pattern seems to be that the classified version is generally 10 to 20 years ahead of the unclassified work on a bunch of things. (For the Soviet end, this stops being the case in the 1980s it seems, but I don't know how much of that is that the USSR is just falling apart and how much of this them failing to archive things well, or make their archives available, or failure to declassify things. Also, the Soviets were never quite as good at a lot of cryptography things. For example, while both NSA and GCHQ came up with a lot of ideas about public key cryptography before it was public, I'm not aware of any evidence the Soviets did.) So by that logic, if one thinks that quantum computers will be practically able to do some decryption within 15 years or so, then that's an argument that it should be plausible that the NSA can do it now.

Local models aren't very good for many purposes. For example, for doing math reasoning, they are poor enough to be completely useless. There's a version of your proposal that would be much more workable though: require that new data centers are built with solar power and batteries that offsets much of their power consumption. Even if you only have them offset 20% or 30%, that would go a long way. And then if the current boom does go under, the worst situation is you have a big set of solar panels that can feed back into the grid.