vbarrielle

But the approach here is "write new code in rust", not rewrite.

Automated theorem provers are also built around backtracking, which is absent in LLMs.

There is some long term planning going on, but bad luck when sampling the next token can take the process out of rails, so it's not just an implementation detail.

Bad counter-example, because FSD has nothing in common with LLMs.

A paper that says: "our approach is simpler than the state of the art". But also does not loudly say "our approach is significantly behind the state of the art on all metrics". Not easy to get published, but I guess putting it as a preprint with a big company's name will help...

There are places in France where the house numbers are based on the distance to the beginning of the block, but it's not that common.

Your parent post meant that a few centuries ago, the american continent was not known, so the known world could be split between east and west.

It's cute that you think your high-school level cypher is probably not seen in the training set of one of the biggest LLMs in the world. Surely no one could have thought of such a cypher, let alone create exercises around it!

No one should ever make claims such as "X is not in <LLM>'s training set". You don't know. Even if your idea is indeed original, nothing prevents someone from having though of it before, and published it. The history of science is full of simultaneous discoveries, and we're talking cutting-edge research.

I'm not sure it's very well aligned with the spirit of copyleft licenses.

Indeed, see https://bellard.org/nncp/ for an example.

I must have missed the news about solar and wind being able to flex.

In France the inheritance tax is only applied above 100 000€, so it's really not 40% flat. What's more, it's possible to make donations before the death, 100 000€ every 15 years, without taxes, meaning most of the time no taxes are payed at all below 200 000€ or even 300 000€.

These amounts are per-child, which means you can double them up if there are two children.

Median patrimony in France is 175 000€ per household, so your typical middle class family with two children ends up paying no inheritance tax, without having done donations in advance.

Am I understanding correctly that your watch sends you a notification if you get lower than 40 bpm? What's the point?

The dishwasher, maybe not, but a washing machine is reducing the laundry work by a lot. Have you ever tried to do your laundry by hand?

Both my kids were able to use the bike at 3 years old. The training wheels should not be used even this young. They first learned using a balance bike, and then learning to pedal went very smoothly using the same technique you mentioned.

The private test is supposed to be run by the ARC-AGI organization themselves, without network access. That's why o3 has not been run against it yet. Not sure if it will be possible either, depends on what OpenAI is prepared to do about it.

Lots of pro riders do not take care of their bikes themselves. They're used to having bike mechanics adjust everything for every race, and a lot of them don't consider it part of the job to take care of their training bikes. Some of them don't even do the cleaning.

Am I right to assume one could also sample from a Gaussian distribution for the method to work? Of course, the probability e of sampling between the two real numbers would be very small, but it would be nonzero.

This data is from 15 to 20 years ago. I think the usage of bike has probably increased in the UK in the meantime. That may not change your point, but this 4 times figure is probably inacurate now.

Looks like they're using something like motion matching to recover fragments of the presenter's motion that match the pronounced phonemes. The actors were probably instructed to avoid almost all movement to make sure it was blendable. That would explain why the guy's hand have such erratic and non-natural movement.

If it were calling to a real chess engine there would be no illegal moves.

In my experience, the VO2Max estimator for Garmin tends to produce higher estimates when doing quite slow runs and gradually building up speed, whereas intervals sessions will lead to a lower estimate. From reading their white paper they fit a curve to the heart rate reserve vs pace relationship, but I think they account for heart rate lag using population averages, which means they'll often think (for me at least) that I need a very high heart rate for a slow pace after intervals which biases the estimate down. If one starts slow then builds up speed gradually, the aerobic system is more warmed up which leads to less elevated heart rate. Also, keeping the run short-ish will lead to higher estimates since heart rate drift seems not to be taken into account.

All considered, the variance is quite high which in my experience does not make it that useful to track relative fitness changes.

I see. Not something I'm reliant on, but I can definitely see how it's a problem if your workflow depends upon panels and you often change monitors.

Could you elaborate? I've been using it with multiple monitors for over a decade, and I don't know what I should be missing.

Yes the main flaw of this reasoning is supposing that e does not depend on previous output. I think this was a good approximation to characterize vanilla LLMs, but the kind of RL in this paper is done with the explicit goal of making e depending on prior output (and specifically to lower it given a long enough chain of thought).

P(correct) converges to zero, so you get almost certainly incorrect, at an exponential rate. The original choice of terms is not the most rigorous, but the reasoning is sound (under the assumption that e is a constant).

I find Garmin's sleep tracking on my forerunner 255 quite good. I have no idea how it compares to an Apple Watch though.

The OSI was created about 20 years ago and defined and popularized the term open source. Their definition has been widely accepted over that period.

Recently, companies are trying to market things as open source when in reality, they fail to adhere to the definition.

I think we should not let these companies change the meaning of the term, which means it's important to explain every time they try to seem more open than they are.

I'm afraid the battle is being lost though.

WADA rules do not ban a drug just because it is performance enhancing. Otherwise caffeine would be banned.

Two out of three factors have to be present to ban a drug:

- be performance enhancing - present a health risk - violate the spirit of sport

The third one is a bit harder to define but there is a WADA code defining the spirit of sport.

Most banned drugs are banned because of the first two.

You're spot on, and there's another point that makes doping unfair even if it was widespread. There is a max hematocrit rule, but riders had different pre-doping base hematocrit levels. Those with the lowest level could benefit more from doping before reaching the limit. In a way, the less gifted would be more advantaged by EPO.