As a science educator, I think the only true household name at present is probably Neil deGrasse Tyson

Too much of the science that breaks into the mainstream (AI being the obvious example) is framed primarily from a negative viewpoint (or, remember the "LHC will spawn a blackhole that engulfs the earth" talk, instead of a focus on the incredible science behind it)

Hopefully these models have a longer lifespan before the battery life forces a replacement. If they truly offer 36 hours, there's more runway before they hit the < 18 hours point at which many people need a refresh

Forgive my skepticism that Apple will get it right. Even the lauded latest version of Siri is far worse than the labs' offerings or Android at voice dictation (I now have to fix 1 word per sentence instead of 4, previously)

But in the end, it'll be the desire to eradicate the lightning port from my life that forces my hand on the upgrade

I'm most interested in these problems as a relative measure of performance for successive model generations. If the prior generation couldn't solve a problem but the current one can, that's useful information, especially when we take into account what the proofs look like.

It's not a perfect benchmark, but I prefer it to many others that I see floating around.

There are 1217 problems that Erdos proposed, 595 of which are open: https://github.com/teorth/erdosproblems

Unlike traditional benchmarks, it's difficult to overfit your models to produce flattering results to unsolved problems. The open problems very likely do not have published solutions (the initial batches were merely models surfacing data that wasn't published in obvious places, but we're past that now). New models will exhibit something novel by adding solutions. And the matter of solution is interesting as well (contradiction versus a positive proof).

I would venture to predict it'll take years to decades to get to 0 open problems. But I'd be very happy to have this comment look foolish in retrospect as models continue to improve

From https://epoch.ai/latest/announcing-frontiermath-erdos

> Only GPT-6 Astra solved anything: 2 of the 68 problems. It disproved problem 74 by finding a counterexample, at a cost of $218 and 15 hours of working time, and it proved problem 126, at a cost of $247 and 16 hours

> Across all attempts, GPT-6 Astra solved 5 of the 68 problems at least once: the two above, plus problem 1, which it disproved, and problem 548 and problem 571, which it proved. Most of the remaining problems were attempted between two and five times in total (172 attempts), and none was solved. Reaching these five solutions took over $220,000 of compute across all attempts, compared with roughly $20,000 for the benchmark run itself.

Which implies a genuine improvement in capability, but there's a still a very long tail ahead that models will continue to need to improve to capture.

When I first moved to SF I'd go for dinner and then watch movies there with friends. Became a bit of a ritual.

After those options disappeared, I'd still go regularly to peruse the Bloomingdales and Nordstrom, among other small stores.

When those disappeared I'd still visit regularly, to get some exercise (I lived nearby) or visit the food court.

Seems it's simply too large for anyone to want to take another chance on it. And yet it feels insane that a mall in the center of a major city is sitting there, abandoned while politicians discuss turning it into a soccer stadium.

I could dream up so many ways to use that space

Comments URL: https://news.ycombinator.com/item?id=49539064

Points: 37

# Comments: 20

The only community they have any penetration in is the developer one (in its prime non-technical people would tell me they used Firefox since IE was insecure. Their techy friends would tell them to use it and the message carried some actual weight back then).

Today they are a large organization that needs to make money to sustain itself, while facing declining usage, and less favorable terms for search engine integrations.

And our community (developers) is the least likely to take kindly to the increasing monetization slop (promoted articles, pestering about pocket and their VPN offering, continually requiring accepting new ToS, reminders to use new features, etc).

I mostly use FF, because I think we'll lose something if it disappears (for example, DRM will become intolerable, web inspectors will become more opaque, and generally future web standards may be much more controlled by a small set of large orgs with their own interests).

But I use FF with less enthusiasm every passing day

Comments URL: https://news.ycombinator.com/item?id=49420861

Points: 63

# Comments: 72

Comments URL: https://news.ycombinator.com/item?id=49347798

Points: 1

# Comments: 0

I still think that it's very possible Gemini gets its act together and becomes the true competitor to the existing frontier models (on more than just cost). But they sure are taking their time with this one, and recent org changes don't exactly signal confidence

Comments URL: https://news.ycombinator.com/item?id=49245421

Points: 43

# Comments: 9

I just can't stand to see any more "Why It Matters.", and this trick seems to strip out the worst offenders

Comments URL: https://news.ycombinator.com/item?id=49156476

Points: 1

# Comments: 0

The latter are likely to adopt much more slowly simply because of less perceived risk, lower payoff (no vendor reviews), and less dedicated technical expertise.

Just like personal sites were slow to adopt HTTPS. Mass HTTPS adoption happened once browser warnings and SEO incentives rendered sites mostly useless without it.