n2d4

39 ms·

571.
by n2d4 3y ago
Why are your secrets in the source code in the first place? Put them into environment variables, which I hope aren't exposed by your infrastructure by default. If the secret you're protecting is the source code itself, that's
572.
by n2d4 3y ago
> What is stated is: “If any count value is zero, no subsequent dimensions are allocated.” This is mostly just for clarification, because what could the alternative possibly be? The key part comes a bit later: > The components of t
573.
by n2d4 3y ago
Trailing commas in JS don't do that, but having more commas than values does. (eg. [1,,].length === 2)
574.
by n2d4 3y ago
From what I understand, this is not a company, but the personal blog of a (former) consultant that worked for a company that had already spent $1.5mil when he took over. Given that context, the headline seems clickbaity as he wasn't pe
575.
by n2d4 3y ago
> Maybe instead of writing a new framework, we should consider doing a React 2.0 with batteries included. Besides what's already in NextJS, what are the batteries you're thinking of?
576.
by n2d4 3y ago
Those are the only ways OCaml represents types in memory (besides float, as briefly mentioned in the article) If you have a record type, or anything else, that'll compile to a pointer and an object in memory, comparable to e.g. Python
577.
by n2d4 3y ago
Spot on. The user asked for recipes using only water, bleach, and ammonia, it's funny but hardly news-worthy. https://twitter.com/PronouncedHare/status/168736440337978982...
578.
by n2d4 3y ago
This one [1]? That tracks two distinct versions (0613 vs. 0314). Also, IMO, the tasks they evaluate aren't useful (I rarely want my LLM to tell me whether 17077 is a prime number), and there's room for cherrypicking/survivors
579.
by n2d4 3y ago
Your incorrect assumption here is that determinism comes with no tradeoffs. There are a few outsider analyses on the topic, for example, this one on sparse MoEs [1]. If OpenAI uses sparse MoE as described, then determinism would be possible
580.
by n2d4 3y ago
Those responses sound very similar (albeit a little less extreme) to glitch tokens [1]; these are created by the tokenizer using the full dataset, but the corresponding training samples were then thrown out before GPT was trained (eg. some
581.
by n2d4 3y ago
From the NYT source linked on the Wikipedia article: > The name of the material comes from the initials of the surnames of two of the scientists — Sukbae Lee and Ji-Hoon Kim — and the year 1999, when they say they first synthesized LK-99
582.
by n2d4 3y ago
According to Wikipedia, Q Centre was founded in 2008. LK-99 is called LK-99 because it was discovered in 1999, but they stopped working on it for the bigger chunk of the 21st century. https://en.wikipedia.org/wiki/LK-99
583.
by n2d4 3y ago
What I'm saying is "there are non-deterministic primitives", not "there are no deterministic primitives".
584.
by n2d4 3y ago
As the article says, cuBLAS is deterministic, but other CUDA primitives (eg. some of those in cudNN) are not. Yes, the non-determinism is being introduced somewhere, but that is splitting hairs. The point is that the primitives that you wor
585.
by n2d4 3y ago
I just edited my comment while you were writing your comment to add an explanation. The point here is that some primitives in eg. cudNN are non-deterministic. Whether you classify that as a race condition or not is a different question; but
586.
by n2d4 3y ago
No, this is not true for GPUs. https://www.twosigma.com/articles/a-workaround-for-non-deter... (In this particular case, the order in which the numbers are summed up is non-deterministic due to GPU parallelism, which
587.
by n2d4 3y ago
Search on that site for "room-temperature superconductor," and you'll see why it never gained attention.
588.
by n2d4 3y ago
> What benefit have we gained by disseminating this research so slowly? It's less visible, but at this scale there's a high cost in everyone pausing their research to replicate LK-99. If it turns out to be real that's wort
589.
by n2d4 3y ago
> That depends on how you define it. I agree, this is a semantic discussion. I'm using Wikipedia's definition. https://en.wikipedia.org/wiki/Expected_value
590.
by n2d4 3y ago
Not quite; there's no such thing as a geometric EV (EV is by definition always arithmetic). Instead, you calculated the geometric growth rate, which the Kelly criterion would then maximize; but it doesn't maximize your average out
591.
by n2d4 3y ago
The other people gave good intuition on why the logarithm is good but not on why the logarithm is the best choice. For example, max(sqrt(EV)) would also reduce outliers. There is a more mathematical way akin to the original derivation of th
592.
by n2d4 3y ago
Yes, that's what I meant. You probably don't want to take unnecessary risk by going +50%/-50% markets, but it's not for the reason the parent mentioned in the conclusion.
593.
by n2d4 3y ago
> Over time it does not. It does! You take insane amounts of risks (that you want to avoid), but the expected value stays positive, even if you repeat it a million times. You might have a 99.9999999999...% chance of going broke, but you&
594.
by n2d4 3y ago
While I agree, that conclusion is wrong because you forgot to look at the win-win and loss-loss cases. Even with two coin tosses, as in your example, you'll win on average: 1.5 * 1.5 = 2.25 0.6 * 1.5 = 0.9 1.5 * 0.6 = 0.9 0.6 * 0.6 = 0
595.
by n2d4 3y ago
Maybe, but it's probably easier to just use a superconductor for the same: https://en.wikipedia.org/wiki/Superconducting_magnetic_energ...
596.
by n2d4 3y ago
In the most ELI5 way possible, a superconductor is a material that doesn't lose any energy or produce heat when used as a cable/wire, and at the same time also a very strong magnet. MRIs and powerlines are more obvious application
597.
by n2d4 3y ago
Apparently that's what they're doing: https://twitter.com/condensed_the/status/1684960318718406656
598.
by n2d4 3y ago
>ignoring everything named "quantum" This heuristic is probably a good one for non-scientific stuff but I'm not sure how accurate it is when we're looking at a group researching quantum effects
599.
by n2d4 3y ago
Why do you silently assume that samples aren't being shared around as we speak? CMTC of the University of Maryland stated that the authors are cooperating in regards to this https://twitter.com/condensed_the/status
600.
by n2d4 3y ago
>but what's stopping a random internet anon from gluing a crumb onto a pipette and taking a picture of it? Nothing, just like nothing would stop a random internet anon from faking a video of the same thing. Even if that existed, i

More ›