LPisGood

What makes this apparent to you?

They are still confidently wrong often. It’s harder to spot with Opus 5.5 effort medium+ but AI search overview is rife with this

Google Pixel / Android are very popular. Chromebooks also did well.

Wow! This is very cool. I had an idea for something very similar to this when I found out the US federal government offered API keys, maybe three or four years ago. I found that while Congress did list bill revisions as they were made the format and retrieval of these was not an easy task. The idea was that I wanted to look at any line of any bill and see who proposed it that particular line’s inclusion.

For the record, and feel free to steal it if you like it, I called it “G(overnm)it-blame”

Linear programming admits a worst-case polynomial time algorithm.

It is notable that simplex, especially dual simplex is often better in practice, especially for integer linear programming with branch and bound because you only have to slightly modify the parent optimal basis in simplex vs entirely restarting or refactorizing for interior point methods.

That is unfortunately not the case

> As a non-profesional (everyone here)

I wouldn’t be so sure of that

Subexponential(poly(n)) may be exponential in n

> To be clear though, it does not imply that NP is subexponential

Is there a subset of NP problems for which a subexponential algorithm would witness the entire NP class being subexponential?

I said the same thing in another thread and someone compared it to no one forcing you to own a smart phone. I will preempt that argument by saying the community is motivated to verify these results because they are important to the field.

It is the canonical NP complete problem.

Actually we can do classical theory of strategic interaction here. There will be a range of acceptable negotiated settlements to both parties since both parties have some estimate E of the expected balance of power resulting after a war. If A is the amount of surplus value Anthropic expects to lose in a war and H is expected amount of value humanity expects to lose in a war, the the range [max(E - H, E - A), min (E + H, E + A)] gives a whole continuum of mutually acceptable settlements.

Now there will be bargaining friction given that any power sharing agreement with Anthropic will likely increase their future capacity and thus the future expected balance of power, but maybe AI has a long enough utility horizon that we should all expect to be dead by the time it realizes the long-term expectation advantage

That doesn’t make any sense to me. Access to AI doesn’t make you more credible. At best, it makes you more productive.

I feel like this comment contrasted with the incredible movie releases this since ~May is just not holding water.

This is like how OpenAI solved some hard math problem in only a few hours.

> If a firm spends $99k on compute to replace a $100k engineer and produces 5x the output, how is that not a 5x increase in aggregate labor productivity?

It sounds like they spent 99k to replace 5 100k engineers, which means they could charge 499k for it.

> and getting actionable BI from it.

I’ve always had doubts about the validity of this “actionable BI,” even before it was slop, serving, mostly to serve leadership’s biases and justify their preferred ideas.

If accuracy was never that valuable to begin with then this should not be a problem.

There are entire businesses set up to help enterprises do this at scale: https://canarytokens.org/nest/

People could point their AI at a physics textbook and ask it to deduce the vuln from first principles too but that doesn’t mean it will be successful.

Measure 0 is different than empty.

Certainly! Probability 0 does not mean impossible.

In addition to all of their open source contributions, their ability to pay a ton of money and fun problems to solve led a lot of fantastic engineers to want to work there — especially before 2015.

If you view creating programs like an optimization problem over the feasible region of valid programs, this is not unreasonable.

A commodity is an undifferentiated good. I expect gasoline to be practically the same at each provider. AI will be similar.

I’m not sure if this counts as a nitpick, but the definition of frontier automatically includes the cheapest models

That is a very interesting idea that I have never heard anyone say. It would certainly help from a liability standpoint I assume.

Is open data just not possible because it is almost all copyright infringement? I wonder why Chinese labs haven’t dumped it at least.

Half a million dollars is a rounding error to the defense industry. If the risks were evaluated to be a national security problem, it would be trivial to do.

> But you never heard someone say "you should still code in assembly so you can understand what your compiler is doing

Any quality CS program will have you do that for a year or so, plus theory.

The minimum is achieved at a vertex, it is just also achieved elsewhere!