Viewing profile — davrosthedalek

Because of regulation, not because they are intrinsically hard to get.

It is actually an interesting conundrum. Is a non-well-aligned frontier level AI a problem? I think it is likely that it is, or at least has a high likelihood to be in the future. …

Well, I think it's fair to assume that a) They didn't upload everything before they realized it would work. b) They want to mention this as an advantage for future analyses c) Even…

To be fair, there are really three threats: a) People do bad stuff because LLM told them a wrong thing. Example: AI told me I should treat my heart attack by putting a fork in the …

If there is an electrical connection between these downstream boxes and the inference servers beyond the power connection, it does stretch the definition of air gap.

I am somewhat more worried about a Darkstar AI.

Too many spaces around the --. That's not load-bearing.

The grant size for phase 1 is 750k / 9 month. That is actually quite substantial. [Despite that, I generally think more money should go to science, all around. But I have COI here.…

Lightspeed in air is close to c_vacuum. Light speed in fiber is roughly 2/3 c_vacuum. So for transatlantic it might be faster.

True, but I think the US requirements are indeed different. Ukraine must repel invading forces in their own country -> lowish range, mass produced, not necessarily precision strike…

Das Auge ist (man) mit.

10 Gig ethernet is 10GBps usable rate (before packet overhead). The line rates are higher to accommodate this. For 10GBase-R, it's typically 10.3125 GBps, with a 64/66 encoding. Fo…

The question is: If you have enough full body scans of many healthy people, and the statistical tools to model it (beyond "this range is OK"), whether this would reduce these false…

Just for illustration: Gravitational wave detection is on the femtometer scale. The proton is about that size. We can measure these things, but the machines are, let's say, "big".

This is somewhat speculative, but as I see it, there are two ways to retain excellent people: a) You pay them handsomely b) You do shit they like, they way the like. Sometimes it o…

Right, that's now /usr/bin/true !

Why it is probably and regrettably true that few people in camps will change their mind, data analysis can help people who haven't been captured yet to either stay away from the ca…

First rsync and now less? What comes next, cat?

But: "After posting this on Hacker News and recieving [sic] almost no substantive input, discussion, or response on the actual content of the article, I decided to rewrite all of t…

Haven't looked at the code, but the allocated memory could be larger than necessary to make "off-by-one" or "off-by-a-few" errors less deadly. Then zeroing it out makes it even les…

Many statistics were presented. In the view of the author (and I think he is correct), none of them show evidence for an increased bug rate from Claude. That is absence of evidence…

Yes, I do for example. And the author discussed the use of AI pretty exhaustively in point 0 of the post.

No. It's a description of the result of the maybe underpowered study. the underpowered study did not find evidence. Evidence is absent. Because it is underpowered, it's not evidenc…

He can't make the evidence of absence claim, but he can absolutely make the absence of evidence claim.