tmp10423288442

Always room for anti-crypto, anti-AI, etc. grifters to go against the pro-crypto, pro-AI grifters. Both side improve their counterparts' business!

If cookie consent dialogs have been going on for as long as they have, can they really be called malicious compliance at this point? Seems like people believe that is the actual intent of the regulation at this point.

Only wasted two minutes if you read really slowly

There are tons of online services that require you to take pictures of your face, driver’s licenses, passports, etc.

The models produce both Lean code for formal verification and a traditional-style narrative proof. Like the general long-form output of frontier models, the math papers produced appear to be generally correct technically, but written in an ungraceful and sometimes hard-to-follow style, so they are often polished by a human mathematician as of today.

Why would society fund mathematicians if they decided to become a guild that hides secrets? They could pursue that as a hobby, but they’d end up like the coders who refuse to use LLMs - rapidly becoming irrelevant and a bit sad from an outsider’s perspective.

I can have Astra run a large-scale infrastructure migration 24/7 (much of the time waiting for results), completing it in weeks or even months faster than I could before agentic AI.

If no one can find your site because you block agents, that won’t do much good either.

Kind of. Eventually it gets too slow even for the OpenAI engineers using it, and then they need to fix it.

ChatGPT uses a distinct bot type for ChatGPT-initiated web search versus model training:

GPTBot → model improvement / training crawl

OAI-SearchBot → search index / ChatGPT Search

ChatGPT-User → fetch initiated by a user's ChatGPT interaction

OAI-AdsBot → advertising landing-page crawl

> the current supreme court seems perfectly willing to throw out decades of settled law on an ideological whim, and actually, for that matter, happily interpret very clear congressional statutes in very odd ways

Fortunately or unfortunately, the Supreme Court has always been willing to do that. This court has been overruling decisions of the Warren Court, just as the Warren Court overruled decisions from the Lochner Era.

Iran is essentially finished at this point - they no longer have control over the Strait of Hormuz, and oil is rapidly flowing through the strait (to say nothing of the oil already exported via other pipelines, or the impetus the war gave to accelerate pipelines under construction, such as the Iraq-Turkey pipeline).

But you weren't going to ask the LLM, and now you know something you weren't aware of. Seems no different than someone reporting what they found out in a Google Search, as long as they disclose that they used an LLM (as long as they used one that actually searches the web, but pretty sure even ChatGPT Free does that these days).

Should governments not have posted announcements in newspapers back when they were widely read, even though they require paying to read them?

Give them some time - the rescinding of the designation only just happened.

Nah, Jevon’s Paradox says that cheaper tokens will mean increased overall energy consumption.

If we can’t even build data centers, the least disruptive industrial use possible, there’s no hope to reindustrialize the US or anywhere outside of China.

SemiAnalysis’ founder was roommates with Anthropic people, not OpenAI, so he may be slightly (very slightly) more objective here.

With Sol OpenAI is more like the A student, and Astra seems like it will be the S-tier student if the rumors are true.

I think I misunderstood your original scenario. This one makes a lot more sense. Although I wouldn't want to be in the position of being the test case.

I severely doubt that judges will be impressed with this reasoning.

What I've read about government hospitals in India is that they're extremely overcrowded, with people crowding the hallways and waiting rooms to be seen while bleeding out or dying from illness, without even air conditioning in a lot of places. On the other hand, this was reported by a young Indian doctor that wanted to get to the UK ASAP, so not sure how representative his accounts are.

Really depends on your insurance, and whether you're visiting for a yearly physical or for an actual concern.

I just don't agree that poverty explains a lot of crime, particularly in developed nations. There's probably some explained by legitimate need, but a rational person would probably find welfare and other benefits easier to use and less risky. Some is due to "poverty", but really envy for luxuries that others have.

I think the largest part is either due to being caught up in a bad social set, or straight up because they think crime is fine. Poverty doesn't cause this, at least not primarily - it's rather that personal characteristics such as lack of impulse control and high time preference lead to both poverty and a propensity to crime.

People hate crime than they hate any of these faults of police.

With modern surveillance technology, you could probably get rid of a fair number of cops, but you’d still need some muscle to actually enforce the law. The other option is regulatory penalties, but those tend to work best on middle-class people, not on the classes that commit physical theft and violence, which people hate most of all.

I think the more honest people in the space acknowledge that’s at least one of, if not the major reason. The antivax movement contributes as well of course

It doesn’t copy the brightness AFAIK

Maybe yes, maybe no. Measles cases have been on a long-term upward trend in the US since the bottom around 2000: https://ourworldindata.org/grapher/number-of-measles-cases?t...

In Canada, measles cases suddenly exploded in 2025 as well, and they don't have RFK Jr. there: https://health-infobase.canada.ca/measles-rubella/

The policies around measles vaccines have shifted a little since RFK became HHS Secretary, but not in a material way. Generally the messaging has focused on putting measles, mumps, and rubella in individual vaccines rather than combined, but at no point has the federal government recommended not to get vaccinated. That, combined with the Canadian data, implies that RFK or other changes to US health policy are not likely to be the cause of the measles outbreaks.

Scott Aaronson basically invented this form of watermarking, so it's not surprising he has a good description of it.

LLMs never are limited to "exact words", because their output is inherently probabilistic. The method Anthropic (along with Gemini, who has been using the exact same watermark for at least a year) uses doesn't bias the output token distribution, just reseeds the PRNG in a way that can be detected after the fact: https://www.anthropic.com/news/claude-text-watermark#which-s...

Anthropic actually tells you the approach they use, and it's not that. From their Claude Text Watermark page[0]: "Claude’s text watermark is a version of the SynthID-Text approach published by Google DeepMind in a Nature paper in 2024."

The Nature paper is "Scalable watermarking for identifying large language model outputs"[1]. This method does not separate out tokens into separate classes, but merely uses a seed for the PRNG that selects which among the most likely tokens generated by the LLM will actually be output. This has the advantage that there's no green and red token sets, so no token is systematically favored or disfavored. If a particular token is overwhelmingly predicted to be the most likely candidate, it will almost certainly be selected, so the watermark doesn't affect that. Even if there are several choices of output token at a point that have similar probability of selection, the watermark doesn't systematically bias in favor of one token or the other.

This is actually a quite elegant method of watermarking that, contrary to people's fears, won't adversely affect the model output. The main concern I have with it is that it appears that you can't actually test the watermark locally, without uploading it to Anthropic. I'm not sure why that's the case, since there's no particular reason the watermarking key has to be private, except if you want to prevent others from generating text with their own LLMs that is watermarked to look like it's generated by Anthropic - but everybody wants their text to not have the watermark.

[0]: https://www.anthropic.com/news/claude-text-watermark#:~:text...

[1]: https://doi.org/10.1038/s41586-024-08025-4