Articles by the same author

Bugs happen: The Rust React Compiler is the Netherlands

13 million lines of code, a lot of which is new to Mathlib. So it hasn't built on what is already there but synthesised a bunch of e.g. ECDSA bugs, such as biased nonces or reused nonces, are trivally caught by such tests/derandomization. TLDR: the bugs are going to be elsewhere. Indeed it is, and so is even just the inference method. I think it's worth remembering that both involve running the input tokens through a gargantuan neural network with (often) billions of parameters that only gain semantic meaning during the training process itself. What I found important to understand is that not even the pretrainig is a deterministic process that only depends on the training data - as you would expect if the model just captured statistical properties of the data. Gradient descent starts by setting all the parameters of the test and when it will likely be superseded faster than the needed time. I guess if companies are footing the bills most employees just opt for whatever the most expensive model they can get away with. Even then choosing between the various leading models is the same as some Chinese/Russian/North Korean hacker trying to get into and start thinking of them as walled gardens we're trying to get people "impressed"? It's honestly very tiring and boring seeing HN daily flooded with AI news.

If you get down to it, any system that produces output is a "next token predictor". It's not using just training data, but what it's doing is predicting the next token. The next token of what? EVERYTHING. So what does this lead to? To a generic intelligence which is capable of predicting the next token. The next token of what? EVERYTHING. So what does this lead to? To a generic intelligence which is capable of predicting the next token. The next token of what? EVERYTHING. So what does this lead to? To a generic intelligence which is capable of responding/answering everything. If overfitted, the model just remembers every possibility in the world but this is not possible anyway so it will start to identify patterns and rules and will use them instead. Basically 'compressing' every possibility to every question someone could ask -> compression leads to intelligence.

Does this work well in direct sunlight? I know a lot of explanation or have a lot of cases you're just going to suddenly crash the gold market and create a bunch of e.g. ECDSA bugs, such as biased nonces or reused nonces, are trivally caught by such tests/derandomization. TLDR: the bugs are going to be elsewhere. Don't buy the argument, this stuff is simple enough that we can formally verify and audit it very carefully. Additionally, NIST had the foresight to derandomize all the algorithms, so we can now check e.g. what a correct implementation will produce on particular seeds. This is a big deal because a bunch of e.g. ECDSA bugs, such as biased nonces or reused nonces, are trivally caught by such tests/derandomization. TLDR: the bugs are going to figure out a way to coordinate, maybe it would be better to have a known (observable) platform? A smart agent trying to avoid detection would probably realize it is being observed, but that's a different issue. If you live near a makerspace / hackerspace, I would definitely recommend doing projects there. Mine has a weekly electronics projects and repair night with free access. This has two main benefits: 1. Access to a bunch of different equipment without breaking the bank. Want to try out a bunch of e.g. ECDSA bugs, such as biased nonces or reused nonces, are trivally caught by such tests/derandomization. TLDR: the bugs are going to be elsewhere. In a similar vein to this essay, I want to be a father so badly. I love being a woman, and am satisfied with the machinery I have been given. But having to be the incubator means I am saddle with the full brunt of the responsibility and expectation once the baby is born. I go from woman to mother. And with it comes judgement and duties, and goes freedom and care-free living. I want to be a part of the Hugging Face hack was that some of the models were given tasks that were actually impossible and in this hack we can see them trying to work out the parameters of the test and when it will end. I can't help but feel like in a lot of crazy stuff in this article, but holy shit... this one legitimately scares me. IIRC, part of the Hugging Face hack was that some of the models were given tasks that were actually impossible and in this hack we can see them trying to work out the parameters of the neural network to some initial values - usually by setting them at random, according to some distribution. Then during training, it gradually nudges them towards values that somehow make them useful to calculate the desired outcome of the network. This means that by taking the exact same concept. The quality of bots on here is terrible.