treszkai

Except customer demand doesn't follow the fixed supply of GPU compute that any company has, so they either a) rent compute on demand, b) sell their excess compute, c) let it sit idle and maybe an employee claims it.

Because it's a poorly written article.

uhm, spoiler alert?

Except in the case of nuclear weapons, humans have their hands on the launch codes, and the nukes are not going to launch themselves. In the case of extremely capable AI, nobody has their hands anywhere close, because nobody is able to – nor willing to – supervise a thousand agents running in parallel, when they usually get shit done. Except eventually they are so good at getting things done that they disregard the values of the human asking them the request, such as "do not hack other corporations". Interpreting human intent precisely is an extremely hard problem, especially in the face of something 1000x faster than you operating autonomously.

> the HF hack was partly the result of a training algorithm that incentivized goal completion as the highest priority, and let them run endlessly in an unmonitored sandbox with weak security

I don't mean to be a smart-ass, just emphasize the fatality of this: goal completion will always be the highest priority, and even if one day it takes second place on certain deployments to "human values" or whatever, there's no way to guarantee at the moment that it will be so on every deployment of highly capable models across the globe. Same goes for running AIs in an unmonitored sandbox with weak security.

The number of implementations and the lack of success suggest that it is a genuinely hard problem for humans, and "agents" are not going to magically make it simpler. Unlikely that humans abandoned their projects (or failed to band together) solely because of their desire for code quality.

It sounds like AA scraping Spotify was an unnecessarily risky move in terms of legal consequences and the toes they step on (who might be behind the downtimes). And it's not even fully in line with their aim of cataloging all the books in existence: while music could be argued is "information", one's sense of IP is much stronger with pirated music than with pirated books.

How on earth are they able to cash in on that? Surely they are not ordering neither air fryers nor their servers from Amazon.. Or anything for that matter, given the illegality of AA. Do they resell them at 90% on eBay?

Sorry I'm out of the loop: how did Spotify use AA or other pirated data?

Check the top HN comment for a similar one for Razer: https://news.ycombinator.com/item?id=49361345

But you're in the air fryer production business, not in the food video production biz (which is really saturated). So you likely don't have the skills to launch a food porn channel organically, so you have to pay a youtuber, which is not the worst, but the yield would be unpredictable (and more costly to test). Sounds overall better than Amazon ads, though.

> Your access to this resource has been blocked.

Do you have VPN enabled?

To roll with this idea: soon the credits will include more names like Sol and Opus 5 and various pointy-haired bosses than names like Mike Little. That's not very encouraging either.

And similarly, humans are not Turing-complete either: none of us can even in theory sit down and store a trillion bits of information in their mind (or paper, whatever) without loss and without a computer.

So Turing-completeness is not even a necessary condition for intelligence.

> LLMs are text-prediction engines. They are not Artificial Intelligence, and shouldn’t not be treated in any form or fashion as if they possess intelligence.

I agree that humans must verify LLM-produced facts, but strongly disagree with these kinds of "stochastic parrot therefore dumb" arguments.

Yes, an LLM is a "stochastic parrot". No, that doesn't imply that it is dumb. Enough to look at how Terence Tao asks ChatGPT to help him understand a solution that nobody had ever discussed before [1], or how a random guy asks ChatGPT in a handful of words to disprove a 30-year-old conjecture, with zero technical input [2].

If your parrot in a birdcage with internet access can finish the sentence, "The counterexample to the Dinitz–Garg–Goemans conjecture is...", then it's a pretty smart parrot, by all reasonable definitions of "smart". Just because someone bottled up the formula into matrix multiplications and added some random sampling to the outcome, that doesn't take away from the fact that the parrot said provably correct statements that the biggest experts in the field couldn't imagine.

And no, I'm not implying that the LLMs are correct all the time, or that their intelligence and reasoning works in any way like ours.

[1]: https://chatgpt.com/share/6a5fdc7a-d6f8-83e8-bbea-8deb42cfed... [2]: https://chatgpt.com/share/6a60b2eb-0b64-83ee-9c76-7931ca1de0...

...99% of applicants who apply through the standard route.

This is the difference between efficiency (that the author is aiming for) and effectiveness (that Bob has – solving the right problem and only that problem). The two are not measured on the same axis: while Bob could have become maybe 20% more efficient with a better setup, and thus increase the delivered value by 20%, the author was already saturated on efficiency, and no amount of better tooling would have made him win the hackathon, because he was throttled at higher levels of abstraction (such as direction, ideas, vision, market insight).

Similarly to how no amount of efficiency gains would make Microsoft produce quality software, because they optimize for other metrics.

This is an amazing game that I'm still playing after a week, no request for changes! Thanks a lot.

With a population of 12M and life expectancy of 80 years (give or take), probably not, as that needs ~400 newborns a day to maintain, and the birth rate is probably well below that, like in most Western countries.

> without which, these discoveries do not help much at advancing the field,

At least the fact that frontier models are not optimized primarily on formal math reasoning is in itself is good for not putting mathematicians out of their jobs, isn't it?

How do these results (and the future painted by them) affect your profession; do you expect a similar route as in software, where junior developers are unemployable, LLM-assisted development is the norm, and that great developers stand out partly through better communication with their managers?

The event in this case is a daily observation, so in the last 45 years there have been 16k events. Under Gaussian assumption, 3.5 sigma outliers in positive direction are 1 in ~4,000 events, in either direction 1 in ~2,000, so approx. 8 days would be outliers in the recorded period. Judging by the looks, it's pretty close. And the data is absolutely not Gaussian: at best it's the result of a Gaussian process, as the daily recordings are not independent, but dependent on the previous day.

For completeness, the person who adjusted the thermostat in this case was also hired to get stuff done, and the front door was left unlocked. They still shouldn't have done it, but they were effective.

As the author mentions Siddhartha by Hermann Hesse, let me point out that the (very pretty) edition by Macmillan Collectors' Library is a very poor translation, and the more mundane-looking one from Pelican Classics is an excellent one, paralleling the original German.

https://www.reddit.com/r/hermannhesse/comments/1uvcfnj/read_...

What do you base this information on? I live in Germany and have no idea what you're talking about.

OP has many posts from years ago, with wildly differing language: https://roman.pt/posts/alembic-enums/

> If execution is everything, and frontier LLMs solve execution

Steve Jobs didn't mean "writing good code" as "execution", he meant "making things that align with how people want to use the product". Frontier LLMs are not solving Steve Jobs-level execution (even in the near future) because what that would mean understanding human nature – something that Steve Jobs or Henry Ford were much better than most in the industry today, so at best we're going to see LLMs making things that are as good as the majority of products. Because AI does not see whether something is perfect-according-to-Steve-Jobs or merely good-code-according-to-engineering-practices.

I'd also add a willingness to follow the (perceived) intent of the rules as opposed to gaming the letter of the rule. An experienced folk might say, "yeah it won't matter legally", and continue with "but it will matter to me because this rule is in place for this-and-that reason".

Yes, it and the other three posts sound positively AI written. The first post on the blog is how OP uploaded a backdoored dataset to HuggingFace and left it there for 6 months – whether made up or not, it doesn't sound great.

Why wouldn't one buy a second-hand refurbished MacBook Air M1 with 8 GB RAM and 256 GB storage for the same price? [1]

I really don't see the advantages of getting a newer computer for the same price with worse specs, unless it's a gift and social customs only allow a new one.

[1]: https://www.rebuy.de/i,11380014/apple/apple-macbook-air-13-3...

Maybe you wanted to visit Mexico and your dream was specifically about Cancún, but then you ended up in Veracruz and were like "oh well, it is Mexico after all, I'd rather be here and visit five other countries similarly than only a single one with my dream city."