dguest

I've been in a near miss in a plane: the pilot in the other plane was clearly inexperienced, not following radio protocol, and we got within O(200m) of a head-on collision which certainly would have been fatal for everyone.

Thing is, the sky is big. You can have non-fatal near-misses that everyone learns from. In trains they literately railroad you into a head-on collision when things go wrong.

I believe that aviation is safe because of standardization, but there's also just a lot more space to almost mess up.

What do you think the cause is?

I'm not trying to argue there is a single cause, either (sorry if my either-or phrasing implied that). I think attributing any general feature of an academic discipline to a single cause would be missing a dozen other contributing factors.

Of course three letter agencies are watching this carefully, and I'd believe they were weighing in somehow, as they have in cryptography in the past. My point was that we shouldn't ignore the overall sparsity of funding in pure math and science projects. A lot of (probably most) contributors to bleeding-edge research have a number of other jobs in their position.

Maybe because cryptography is funded better than the rest of math.

The big AI companies spend millions on some of these proofs (15M was one estimate for the Naiver-Stokes thing). Had anyone has spent millions on them before this? There was a 1M prize but that certainly wouldn't convince anyone to spend orders of magnitude more to solve the problem.

Meanwhile the three letter agencies are spending billions on cryptography. They have programs that recruit students starting in high school. So maybe they are sending the spooks to suppress OpenAI's cryptography breakthroughs, or maybe they just already outspent them.

That to me is the point that isn't getting enough attention: pure math isn't that well funded. I agree as much as anyone that these proofs are an unprecedented breakthrough for LLMs, but let's keep in mind that the money they are throwing around is also unprecedented.

It would be cool if they embellished on the buildings and towns a bit: realign them on the grid so that they actually look good, embrace the medium. Maybe even tweak some of the narrow valleys to be more aligned.

How does it infer what's underground? Does it include tunnels / mineral deposits / bunkers?

That's true for theoretical physics.

Experimental physics takes it one step back: it's all about setting bounds on what is potentially possible. A model of the universe might say that the numbers are between zero and one, but there's some other model that says there could be 3s.

Note that this epistemic uncertainty is not the same thing that the article is focused on. The article is mostly about one specific model, which itself has some randomness baked on.

But seriously, it works to market anything:

> A doghnut: a snack for hungry people, and adventurers

Didn't you see the mountain?

As one friend put their endorsement of Ted Chiang: sometimes you don't need 200 pages to make a point.

In general, though, it's amazing how much hard science fiction authors managed to get right. I've read quite a few 20 year old books recently and while I might just be selecting those that aged well, someone definitely saw it all coming.

I feel like "just a question of cost" is a heavy lifter here though.

Cost core to our resource allocation system, for better or worse, and if something costs a lot it usually means we can't do it as much. In this framing, cutting out animal testing would (often) mean medical research is slower. In this framing it's a question of where people are willing to put the dial on ethics vs results, and of course pumping the breaks in one place without universal buy-in just means the cruelty moves elsewhere.

I don't think the above 1-dimensional framing is the whole story though. You could imagine more standardized data sharing, so that the data that cost the lives of a few thousand rats could be used by others. And of course you could require (as drug trials do) that the study be registered before any experiment happened, and that the results be published regardless of the outcome. Some of these changes could actually make the science better, in addition to more ethical.

I'm not claiming that anyone is doing this on purpose, but overly engineered codebases are a good thing for a company like Anthropic. If your code is so bloated that it is pretty much unreadable without an LLM they win on both sides.

The US is innovative but can't be bothered to get the boring stuff right.

Meanwhile Europe standardized bank transfers with IBAN, killed connector type proliferation with by standardizing on USB-C, and routed 3-phase power most homes. Oh, and the metric system. And while Europe has a wide range of cultures and functionality of government, a lot of places have simpler tax forms and a more prescriptive legal system (as opposed to the precedent-dominated Common Law tradition in the US). The German-speaking countries even rewrote spelling to make it simpler.

I view a lot of this stuff as infrastructure investment. We shouldn't have to stumble over the mundane to do interesting things. As others have commented, Europe also had the advantage of adopting some technologies later, but in some cases they just seem to be more aligned with a system of constant (if sometimes frustratingly slow) improvement.