"At the same time, China is also the world's leading producer of electric cars..."

Kind of interesting for a professionally branded company to use "..." like that

I think its worth emphasizing that his argument isn't completely against generative ai, but rather its environment. Although I don't see why it would be impossible for something like an LLM to learn some sort of self-play within its context window

I don't completely disagree but its worth noting how new a lot of the empirical evidence in favour of LLMs are, so its not impossible to be a tad ignorant of the present

No

> It uses 384 routed experts (top-8) with hybrid attention (full-attention + sliding-window 128 at 6:1 ratio) over 70 layers (1 dense + 69 MoE)

https://recipes.vllm.ai/XiaomiMiMo/MiMo-V2.5-Pro

> But I am a bit reasurred that at least my job won't be fully replaced with AI :)

I honestly can't comment with certainty that training from videos alone and whatever tokenization scheme they're using will ever get perfect dynamics.

However it is worth noting that transformers can do a pretty good job at learning dynamics with the right pipeline (not video): https://arxiv.org/pdf/2605.15305 https://arxiv.org/pdf/2605.09196

My point here being that representationally, it might be possible to learn good dynamics without a radically different approach/arch. There are already models that extract 3D tracking points from videos, so they could possibly be leveraged for learning dynamics (which on its own gives precedent for end-to-end approaches also possibly working).

Thanks for the additional reading. I've often thought about LLMs and their ability to represent the physical world with its laws. And always concluded it is not really possible to do so with "just" text tokens and their relations in a latent space. It looks to me there are different approaches being taken to tackle this:

* You could instruct your LLM to interact with a simulator to run experiments and infer behaviour

* You could edit the transformer model and inject spatially relevant data rather than text as is done in above paper

* You could change the architecture to be more condusive for representating a world state. I.e., LeCun's JEPA world model.

* You could further enhance some of the above by using a differentiable physics engine (eg. NVIDIA Newton) to calculate losses directly.

But at the end of the day if a model has any hope to always produce realistic physics, it HAS to learn the laws of nature in some form or other. It looks to me that the next big leap could be achieved by combining the last two approaches.

P.S.: I like discussing such topics. If anyone knows a forum or discord with like-minded people, please let me know :)

> P.S.: I like discussing such topics. If anyone knows a forum or discord with like-minded people, please let me know :)

Unironically twitter (and only use the "Following" tab as opposed to the "For You")

Make an account that only follows university affiliated researchers with less than 1000 followers. In my experience discord servers get suffocated by beginners and crackpots because conversations don't naturally self-organize into their own threads.

Thanks, I'll try using the "Following" tab. I have a lurker account but never really used it because I only ever saw crap in "For You".

FYI it’s not an Arab country

Interesting comment from him:

"

SPIEGEL: So you don't consider Collins to be a true scientist?

Venter: Let's just say he's a government administrator.

"

https://www.science.org/content/blog-post/craig-venter-venti...

They factorize the distribution in which they are trained on which is essentially generalization

https://arxiv.org/abs/2602.02385

Yeah I wouldn't be surprised if journalist are getting high on their own supply of resentment and fear mongering

Maybe a nitpicky HN comment, but why are we lumping the term automation with very recent grievances about certain kinds of automation

The article literally draws that distinction in the first paragraph.

It does?

" Software brain is powerful stuff. It’s a way of thinking that basically created our modern world. Marc Andreessen, the literal embodiment of software brain, called it in 2011 when he wrote the piece “Why software is eating the world” as an op-ed in The Wall Street Journal. But software thinking has been turbocharged by AI in a way that I think helps explain the enormous gap between how excited the tech industry is about the technology and how regular people are growing to dislike it more and more over time. "

I mean, even going back, people had all kinds of problems with all kinds of automations, e.g. Luddites and the subsequent starving in the streets.

I mean, I would think the opposite it the truth.

Other than a few masochist CEOs, most people don't like having to work for a living to ensure they don't starve and are homeless. It's just in the current paradigm it's what we have do to. And because we have to do it, people get really nervous when rich people attempt to replace human work with automation. Not because we won't have to work, but because we will have to starve.

People not wanting their jobs be automated is different from not yearning for automation as a principle. Most people want or (at least don't mind) elevators, tap water, dishwashers, traffic lights, electrical fuses, sliding doors, etc. Its a very general term

People want their bills and chores eliminated. Show them tech that does that and you'll be every working person's favorite human being. They'll be naming their kids after you.

They wouldn't mind their jobs being eliminated, except for that whole bills thing. Eliminate their jobs without eliminating their bills and they'll hate you.