foota

Funny enough I was thinking about something very similar to this based on the Jev model posted yesterday.

Just curious, why? Is this true even if you did something like a per-CPU histogram that uses atomic ops to increment?

You got an article off by one error, I think you meant to post on https://news.ycombinator.com/item?id=49558685 :)

Ah this is interesting. Essentially the idea is that the compute can try and move into positions that it can evaluate but humans might have trouble evaluating because of the board state's complexity?

That was a lot of snark :-) I know what you're saying. It's not clear to what degree the film industry would be viable in its current form though without IP rights. If everyone can watch the latest marvel film at home for free, are they going to be willing to pay to see it in theaters?

There would likely still be some demand for the experience, but the tickets would need to be more expensive to recoup the costs of making the film. At that point: it does seem like people pooling together their funds to fund the movie itself isn't insane. The point is that the "investors" wouldn't be investing in the hopes of a financial return, they'd be paying for the creation of the film.

I spent years learning logic and doing puzzles. I don't think a baby or even average kid could solve these.

I think they meant more like "people band together to pay Christophe Nolan to create some new film" ala crowdfunding. Is that a sustainable model for the world economy? Unclear.

> Ban offline training/pretraining. Models must train from scratch after submission Previously this was considered impossible so rule. My model shows this is possible Guarantees no synthetic data can be used It makes the comparison fair across differet models. Otherwise some models like LLMs can benchmaxx ARC by using ungodly amounts of offline training. (Since the benchmark has been around a long time, many ARC-like datasets have been created)

I'm not an ML researcher, so YMMV, but... how could a model learn to answer these ARC-AGI questions without training beforehand?

> I agree that its rare to see to face problem sets in real life where every problem is given at once. Even if it is (like an exam), humans can usually only attempt one at a time

Just one small snippet that I thought was interesting. I would always read through ~the entire exam before starting. Both so that I could find the problems most approachable to me, but also because sometimes it helps me figure out the rest of the questions :-)

Ah shoot, I War of Worlds' myself. I didn't realize this was fiction until the end.

Why can't effort just be at the end of the prompt? My understanding is that that is how modes and other "dynamic updates" like the time are handled.

Calling China a country that "bred like rabbits" over the past few decades is first of all offensive and second of all stupid for a country that has one of the biggest problems with their coming population pyramid and limited couples to a single child for most of the last 40 years.

China is at a higher installed capacity by 6x from overall and 30% by per capita capacity. They're growing installed capacity from your source at 45% y/y as compared to 20% in the US.

Sure, there might be more transmission constraints or whatever, but those will certainly be solved in time (transmission capacity is more distributed so I'm not surprised it's lagging capacity).

The quote you pulled for China is conveniently 3 years old and missing the global context that it was in response to supply issues with other fossil fuels, not a reflection in meeting demand.

These charts address your points:

https://ourworldindata.org/profile/energy/china https://ourworldindata.org/profile/energy/united-states https://ourworldindata.org/profile/energy/india

India is a bit more concerning for a number of reasons (lower per capita income means they are lower on per-capita emissions currently and hence have more room to grow, and they don't have the resources that China does to build out renewables) but criticizing China from a renewables perspective (there are a whole lot of other valid reasons to criticize them) is insane.

Hmmm... Why wouldn't this be handled like other end of prompt things like the current mode?

Ah, so it's more that "LLMs just bypass Rust's safety" rather than "LLMs write perfect C"? That's a more fair argument.

> "because LLMs don't make the same coding mistakes that humans do that justifies the existence of the borrow checker"

Citation absolutely needed.

Just curious, are there lifeforms today that rely on these metals to live in anaerobic environments? Maybe the hydrothermal vents that are talked about?

Do you think complex life could develop (perhaps by moving backwards) in a way that they rely on an anaerobic metal-rich environment?

I guess the oxidation event maybe was necessary for life to develop enzymes (since there wasn't sufficient evolutionary advantage for it to develop in an anaerobic environment?) but maybe something could go backwards? I'm just curious if there might be advantages for live in that sort of environment.

This is neat. I'm only halfway through but some really cool discussion about the internals of the chips and what they offer.

I feel like the future is people building applications with tightly integrated LLMs that work hand in hand with the application's own lifecycle and code.

I also didn't realize that people were using agentic harnesses for search, it's an interesting idea. If the context length is short enough it should be fairly cheap compared to running "normal" agentic coding workloads where you have O(100k) context length for doing almost anything.

I think there's probably a "in a long time" missing from unsubtle bully. American diplomacy was a lot less subtle in the past.

> only ~20-30% of the energy in the fuel turns into useful work, while electricity is much, much more efficient. So, we need much less of it.

Isn't a lot of energy used for heating, not work?

Perhaps an un-fun fun fact if you're living near one of these, but Boom aerospace has pivoted to start selling turbines for data centers in addition to its normal business.

Because their model previously got blocked by the government for this and they don't want a repeat?

Claiming that it's more expensive isn't true, it's workload dependent. Filling the cache (worst case) 12 times in an hour is much more expensive than caching it for an hour, and the cheapest is to have your 5m cache refilled by a ping like this is doing.

Imo it's their fault for not having pricing that aligns incentives.

Through the courts though, right?

Wow he was born a _lot_ earlier than I would have thought!

It's a bit unrelated, but I've been wondering if LLM providers are using cache read costs to preserve the illusion of constant output token prices. In reality, the longer your context the more expensive output tokens are, but anthropic and openai both have flat output token pricing.

In practice though as a result of cache reads over multiple turns you will end up paying quadratic pricing anyway.

I once shared a proposal at Google that Vint commented on and encouraged me to follow up on. I haven't, but it certainly made my year :)

How would it help? MMOs generally don't need to know where things are right now since it's mostly static, no?