Hacker Newsnew | past | comments | ask | show | jobs | submit | djsjajah's commentslogin

but what is the point? A ban is supposed to make a certain thing less likely to occur. Does a ban of open source models do that? Presumably, the behavior you are trying to limit is the miss-use of these models but I don't know how many state sponsored hacking groups are going to give a ban a second thought.

Yes, that's how I read Amodei's post.

Imagine Kimi K4 will be as powerful as Mythos. Anthropic can work for months and months to set up guardrails on Mythos, so when the model is finally released, it will generally decline to help hackers develop and prosecute cyberattacks, and if they do, at least there would be a trace so the law enforcement can track the perpetrators. Let's now say that Kimi K4 is released after a similar effort to develop guardrails. But being open weights, someone can just take the model, and finetune it until it does not refuse to assist in developing cyberattacks, and moreover, those people can run the model on their own private GPU cluster, so nobody can track the attack back to them. The situation is actually worse than that, most likely. Guardrails might be just markdown documents which are added to the context like regular skills. Then removing the guardrails for an open weights model does not even involve any finetuning, just removing some docs from a harness.


I think they were making a joke. In the future, you might consider the advice you are giving as well as giving it.


If that’s the case, I will wait for their own answer.


Yes. Wait a day


Not with 800 examples. If you are going to consider an ngram model, I think you are better off getting a frontier llm to write you an absurd regex.


Hmm maybe. Turns out the author trained a logistic-regression classifier on the embeddings too, but didn't report the results:

https://github.com/thelgevold/fine-tuned-classifier/blob/mai...


Expanding on this experiment using logistic regression is an interesting continuation, detailed here: https://www.teachmecoolstuff.com/viewarticle/using-logistic-...

In summary: Using logistic regression actually improves accuracy, but also performance during both runtime and during training.


Is that overfitted?

Do 5-fold cross validation, maybe stratified.


It's the same dataset as the original experiment, but could be interesting to dig into it more.


I would also recommend the approach of using an llm to create the examples, and then train from there.

You can even get fancy and do things like active learning with the llm taking the role of the human annotator and sending in trial statements (and you can use a cheap one for larger gen and a more expensive one for the classification).

I’d be interested in seeing how well LLMs work with writing things like code for what snorkel AI used to have (there was open source code a while back that I assume is still around somewhere, you wrote code that was a low quality set of classifiers and it trained a model around those)


Except, that won’t help. By the time a new fab is up and running, we will probably have a massive surplus.


You assume demand for AI stays flat.


> You assume demand for AI stays flat.

RAM has become an asset class, like real estate, or orange juice, or cattle, or collectible cars.

It’s wild to see.

I don’t know if there is an options market or a futures market for RAM.

But it seems inevitable.

Imagine the United States hoarding RAM the way that nations of old hoarded gold.


$DRAM calls


You need to think this thought through all the way to the end. What it has said also influences what it will say. If it has consistently made combative responses, then the most likely thing to do is to continue to be combative.

I don't think there is any way back after the conversation takes a turn like that so there is no point in arguing with it. The only thing you can do is to fork the conversation before it made the first mistake and give it more context or tell it to look things up.


> "The only thing you can do is to fork the conversation before it made the first mistake and give it more context or tell it to look things up."

This is a key detail that many folks don't seem to understand about LLMs in general. The generation of a response happens based on the model weights and the context window (the system prompt + everything it's fed about the conversation thus far + any additional data included as part of the overall prompt). Each response technically stands alone and is generated entirely from only that context given to it and the model's existing "token space" weights. The illusion of an ongoing conversation is maintained "behind the scenes" by keeping that "context window" updated with the current state of the conversation as context for the next prompt, but the next response is technically an entirely new generation of text.

What it all means in a TL;DR sense is that the fix for a refusal is not to continue the "argument", but simply to remove that entire interaction from the conversation entirely as if it never happened and try a different tack with new / updated / more complete context to get the response you're expecting / seeking.


I wasn't arguing with it.


It’s amusing that a lot of the agents have worked out that sampling doesn’t change ppl.


I think what they mean by “now” is the stuff announced today.


You just failed the Turing test.


The Turing test just failed you. I'll go one better, physics isn't reality, it's a model of reality utilizing math.


And I'll go one better, you haven't said anything here at all, you've just left a representation of what you understand to be saying.


we only perceive the past.


Maybe he passed the Turing test with 88.2% which is 1.8% higher than the competition.


Fortunately for me equivalents to Turing exist: https://en.wikipedia.org/wiki/Turing_machine_equivalents


I don't follow. Can you explain how your comment is relevant to mine? It might help if you also explain how you interpreted my comment.


I have 2 of them. I would advise against if you want to run things like vllm. I have had the cards for months and I still have not been able to create a uv env with trl and vllm. For vllm, it’s works fine in docker for some models. With one gpu, gpt-oss 20b decoding at a cumulative 600-800tps with 32 concurrent requests depending on context length but I was getting trash performance out of qwen3.5 and Gemma4

If I were to do it again, I’d probably just get a dgx spark. I don’t think it’s been worth the hassle.


FWIW I’m in love with my Asus GX10 and have been learning CUDA on it while playing with vllm and such. Qwen3.5 122B A10 at ~50tps is quite neat.

But do beware, it’s weird hardware and not really Blackwell. We are only just starting to squeeze full performance out of SM12.1 lately!


Consider applying for YC's Fall 2026 batch! Applications are open till July 27.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: