Thank you for your service brave memer

I know the author will probably never read this, but in the off chance they or someone else working on open source accessibility reads it:

Thank you. So much ❤️ Your hard work and dedication keeps me going when times get rough.

And thanks for the rant. Nobody should have to suffer in silence, and that includes you. So any time you want to rant at us leeches about working in open source, please don’t hesitate to put us in our place.

I hope that we find a way to make the internet a more positive and celebratory place for people like you who do hard work the rest of us don’t have the time or energy for.

Just fyi, console can be more usable than gui for many disabled people. Text-to-Speech software relies on there being text, and mouses require fine motor movements.

I’d encourage you to research more about this space and learn more.

As it is, the statement “Markov chains are still the basis of inference” doesn’t make sense, because markov chains are a separate thing. You might be thinking of Markov decision processes, which is used in training RL agents, but that’s also unrelated because these models are not RL agents, they’re supervised learning agents. And even if they were RL agents, the MDP describes the training environment, not the model itself, so it’s not really used for inference.

I mean this just as an invitation to learn more, and not pushback for raising concerns. Many in the research community would be more than happy to welcome you into it. The world needs more people who are skeptical of AI doing research in this field.

I see a lot of misunderstandings in the comments 🫤

This is a pretty important finding for researchers, and it’s not obvious by any means. This finding is not showing a problem with LLMs’ abilities in general. The issue they discovered is specifically for so-called “reasoning models” that iterate on their answer before replying. It might indicate that the training process is not sufficient for true reasoning.

Most reasoning models are not incentivized to think correctly, and are only rewarded based on their final answer. This research might indicate that’s a flaw that needs to be corrected before models can actually reason.

Beautiful! I’ll definitely give this a go

python -m http.server is still my media server of choice. It’s never let me down.

On any site with unverified signups (all of them) you can’t.

If you want to talk to real people, you’d have to use a platform that has in-person ID verification. Like a pub, or a park.

Good luck finding a bot free place on your phone. It’d have to involve zero-sum proofs and biometrics. And even then you can’t really be sure that person isn’t using a bot to write without full root access to their system and a live webcam feed.

deleted by creator

I want clear evidence that the LLM will tell me if it doesn’t know something, and will never hallucinate or make something up.

Every step of any deductive process needs to be citable and traceable.

I mostly agree, but “never” is too high a bar IMO. It’s way, way higher than the bar even for humans. Maybe like 0.1% or something would be reasonable?

Even Einstein misremembered things sometimes.

Eh, certain parts of LA are safe. But LA is actually pretty conservative in other areas, due to a large religious population, and a lot of first-gen immigrants.

I have to ask: would this story be so popular if they didn’t mention that the four people that did this were Chinese?

Racism doesn’t disappear just because the article doesn’t say the quiet part out loud. We all know the thought process that led to this article’s virality.

Let’s do better, Lemmy. We all have an opportunity to make the world a more tolerant and empathetic place through what we post and upvote.

It only seems to make a difference when the rich ones complain.

I read the article, and it’s way less bad than the title made it sound. They just set company chats to disappear after some number of days and told employees to not “comment before you have all the facts.” This has been the policy of every company I’ve worked at, including university IT and Amazon.

The title made it sound like they were deleting specifically chats related to open court cases, which is like level 10 ultra-illegal.

Link to the actual paper: https://www.nature.com/articles/s41586-025-08839-w

The repro and verification will take time. Months or even years. Don’t trust anyone who says it’s definitely real or definitely bunk. Time will tell.