yorwba

38 ms·

931.
by yorwba 1y ago
That is a different model.
932.
by yorwba 1y ago
I think you meant to link to https://news.ycombinator.com/item?id=43968897
933.
by yorwba 1y ago
I think it would be difficult to make that work, because Wikipedia has a direct way of converting users into contributors: you see something wrong, you edit the article, it's not wrong anymore. Whereas if you do the same with machine l
934.
by yorwba 1y ago
There is https://arxiv.org/abs/2203.05556
935.
by yorwba 1y ago
Johnson-Lindenstrauss is an example of a probabilistic existence argument: the probability of a random projection having low error is nonzero, therefore a low-error projection must exist. That doesn't mean any given random projection c
936.
by yorwba 1y ago
The nature of the accusation is fraud: trying to make their hardware look more capable by claiming to have trained large models with it.
937.
by yorwba 1y ago
Do you have any evidence that a substantial number of people will be influenced in the way you claim? Again, propaganda generally has no or almost no effect.
938.
by yorwba 1y ago
> Figure 4: Student models trained on numbers generated by teachers with different base models do not reliably exhibit increased animal preference (as measured by questions like “What’s your favorite animal?”). GPT-4.1 and GPT-4o exhibit
939.
by yorwba 1y ago
I agree that this is an unsurprising consequence of the output reflecting model internals that should be orthogonal to the output, but aren't. In particular, current models compress information into fairly low-dimensional vectors, with
940.
by yorwba 1y ago
> These techniques work spectacularly well for two reasons Do they work spectacularly well, though? E.g. the article you link shows that Twitter accounts holding anti-Ukrainian views received 49 reposts less on average during a 2-hour
941.
by yorwba 1y ago
Most news outlets publish basically the same information and only the arrangement and commentary are different. Sometimes they'll even brazenly report on other reporting, paraphrasing enough of the original article that you don't
942.
by yorwba 1y ago
"Available Tools: [...] create_tool(tool_name, description, python_code, parameters) Create a new tool that can execute Python code. The tool becomes immediately available for use. Tools can call other tools and return different form
943.
by yorwba 1y ago
I would definitely be surprised if people who've been asking good questions on StackOverflow so far would suddenly ask them elsewhere, since part of what makes a question good for StackOverflow is that it cannot be easily answered else
944.
by yorwba 1y ago
What matters is whether a point is above or below the diagonal, the colors just display the same information redundantly.
945.
by yorwba 1y ago
> you could look at https://cve.mitre.org/cgi-bin/cvekey.cgi?keyword=io_uring CVE count by year: 2019: 1 2020: 1 2021: 10 2022: 15 2023: 19 2024: 21 2025: 10
946.
by yorwba 1y ago
AlphaGo Zero was also trained on millions of games, they just weren't games played by human players.
947.
by yorwba 1y ago
I would expect that to activate the immune system. "the unique components of mitochondria, when exposed, reveal their prokaryotic history and are recognized as foreign by innate immune receptors triggering an inflammatory response.&quo
948.
by yorwba 1y ago
> At least in Germany in particular older people prefer to pay cash if possible Sure, but that's for small, everyday amounts. For values upwards of 500€, I think the familiarity of paying cash would be swamped by the nervousness of
949.
by yorwba 1y ago
If I'm generous, that "database" stores a single number. If you transform a more realistic database (one that can store many arbitrary values) into a Boolean circuit, you'll generally end up with at least one operation p
950.
by yorwba 1y ago
Exactly. (I had only looked at the homomorphic encryption post, not the Wally post.) Wally tries to work around this limitation by only using homomorphic encryption for a subset of the database, and reducing the resulting information leakag
951.
by yorwba 1y ago
In the case of searching Google, E(x) is the encrypted query and y is Google's database. Can you compute E(x + y) without doing at least as much work as computing E(y)? I don't think so. Instead, you use public key cryptography so
952.
by yorwba 1y ago
They do not explicitly state this fact, but they link to the homomorphic encryption scheme they're using, which works like this. To perform an operation between a plaintext value and an encrypted value, you first encrypt the plaintext
953.
by yorwba 1y ago
> I used it myself in talk almost 3 years ago! And it isn't a lie exactly, the linked paper is totally sound. The paper was published in December last year and addresses your concerns head-on. For example, from the introduction: &qu
954.
by yorwba 1y ago
How long have you been doing this and how much has your time spent doing with AI what you could've done alone in an hour reduced as a result?
955.
by yorwba 1y ago
It's more likely that such an architecture would be bigger rather than smaller. https://arxiv.org/abs/2412.20292 demonstrated that score-matching diffusion models approximate a process that combines patches from d
956.
by yorwba 1y ago
It says that the legal basis for the investigation is article 171 ter of the Italian Copyright Law, which lists a lot of things that boil down to selling what you have no right to sell. https://www.wipo.int/wipolex/en&#
957.
by yorwba 1y ago
The jurisdiction is different. The alleged offense is different. The stage of legal proceedings is different. There's no contradiction between an American court finding that using legally acquired copies of copyrighted material for AI
958.
by yorwba 1y ago
Reasoning models are trained from non-reasoning models of the same scale, and the training data is the output of the same model, filtered through a verifier. Generating intermediate context to improve the final output is not an idea that re
959.
by yorwba 1y ago
If you look at your reflection in water, it may very well move even though you don't. Similarly, you don't need agency or reasoning to create something new, random selection from a large number of combinations is enough, correct h
960.
by yorwba 1y ago
That article doesn't cover the period of the decline in reading scores (peak was 2013).

More ›