Hacker News: tonii141

New comment by tonii141 in "Yann LeCun to depart Meta and launch AI startup focused on 'world models'"

tonii141 — Thu, 13 Nov 2025 08:25:29 +0000

I just see a lot of people who’ve put money in the LLM basket and get scared by any reasonable comment about why LLMs aren’t almighty AGIs and may never be. Or maybe they are just dumb, idk.

New comment by tonii141 in "Yann LeCun to depart Meta and launch AI startup focused on 'world models'"

tonii141 — Wed, 12 Nov 2025 12:45:04 +0000

I’m not demanding anything, I’m pointing out that performance tends to degrade as context scales, which follows from current LLM architectures as autoregressive models.

In that sense, Yann was right.

New comment by tonii141 in "Yann LeCun to depart Meta and launch AI startup focused on 'world models'"

tonii141 — Wed, 12 Nov 2025 10:16:27 +0000

a) That "no-tools" win depends on prompt orchestration which can still be categorized as tooling.

b) Next-token training doesn’t magically grant inner long-horizon planners..

c) Long context ≠ robust at any length. Degradation with scale remains.

Not moving goalposts, just keeping terms precise.

New comment by tonii141 in "Yann LeCun to depart Meta and launch AI startup focused on 'world models'"

tonii141 — Wed, 12 Nov 2025 09:35:01 +0000

a) Still true: vanilla LLMs can’t do math, they pattern-match unless you bolt on tools.

b) Still true: next-token prediction isn’t planning.

c) Still true: error accumulation is mitigated, not eliminated. Long-context quality still relies on retrieval, checks, and verifiers.

Yann’s claims were about LLMs as LLMs. With tooling, you can work around limits, but the core point stands.

Are AI Agents Compromised by Design?

tonii141 — Wed, 15 Oct 2025 12:19:09 +0000

Article URL: https://www.computer.org/csdl/magazine/sp/5555/01/11194053/2aB2Rf5nZ0k

Comments URL: https://news.ycombinator.com/item?id=45591255

Points: 3

# Comments: 1

New comment by tonii141 in "Ask HN: Why doesn't AI use dynamical neurons instead of static activations?"

tonii141 — Mon, 13 Oct 2025 12:24:50 +0000

Not true. AI has been around far longer than modern LLMs and has performed well in non-generative areas, often with orders of magnitude fewer parameters.

New comment by tonii141 in "Ask HN: Why doesn't AI use dynamical neurons instead of static activations?"

tonii141 — Mon, 13 Oct 2025 10:57:04 +0000

What do you mean by "they add instability"?

New comment by tonii141 in "Ask HN: Why doesn't AI use dynamical neurons instead of static activations?"

tonii141 — Mon, 13 Oct 2025 10:20:47 +0000

I agree, but maybe there is no need for billions of neurons to be simulated right away. Artificial neural networks were pretty small at the time.

Ask HN: Why doesn't AI use dynamical neurons instead of static activations?

tonii141 — Mon, 13 Oct 2025 09:51:07 +0000

Why doesn’t AI use dynamical neurons instead of static activations even when it is known that biological neurons are dynamical systems?

Comments URL: https://news.ycombinator.com/item?id=45566581

Points: 1

# Comments: 7

Fusion Power Company CFS Raises $863M More from Google, Nvidia, and Many Others

tonii141 — Tue, 02 Sep 2025 13:38:49 +0000

Article URL: https://techcrunch.com/2025/08/28/nvidia-google-and-bill-gates-help-commonwealth-fusion-systems-raise-863m/

Comments URL: https://news.ycombinator.com/item?id=45102963

Points: 3

# Comments: 0

New comment by tonii141 in "Researchers Are Already Leaving Meta's New Superintelligence Lab"

tonii141 — Wed, 27 Aug 2025 08:55:41 +0000

https://archive.is/lYTN8

How to build a world model? Introduction to Laplace Neuron

tonii141 — Wed, 20 Aug 2025 07:32:26 +0000

Article URL: https://abibulic.github.io/blog/laplace_neuron.html

Comments URL: https://news.ycombinator.com/item?id=44959576

Points: 4

# Comments: 0

How to build a world model? Introduction to Laplace Neuron

tonii141 — Tue, 19 Aug 2025 14:04:47 +0000

Article URL: https://abibulic.github.io/blog/laplace_neuron.html

Comments URL: https://news.ycombinator.com/item?id=44951704

Points: 1

# Comments: 0

New comment by tonii141 in "Autonomous drone defeats human champions in racing first"

tonii141 — Fri, 06 Jun 2025 12:16:19 +0000

I have been involved in research focused on ML control for some time, and believe me, I would love to see an AI model capable of controlling arbitrary systems at different operating points or in different environments. However, it is simply not feasible yet. This AI drone is no different, especially because reinforcement learning was used to train the model, which is generally not practical for real-world systems due to disturbance variables and the continuous need for adaptability.

New comment by tonii141 in "Autonomous drone defeats human champions in racing first"

tonii141 — Thu, 05 Jun 2025 08:05:15 +0000

Let's not forget that this works solely for this particular racing setup. If you change a single gate, the AI they are using would not be able to adapt. Still fascinating, though.

New comment by tonii141 in "Bagel: Open-source unified multimodal model"

tonii141 — Mon, 26 May 2025 07:34:20 +0000

If the model uses FP16 precision and has 7 billion active parameters, it would require approximately 14 GB of VRAM. I didn't read the paper.

New comment by tonii141 in "I got fooled by AI-for-science hype–here's what it taught me"

tonii141 — Tue, 20 May 2025 11:18:58 +0000

AI is used in scene understanding for those applications, but there is no neural network that is steering the wheel.

New comment by tonii141 in "I got fooled by AI-for-science hype–here's what it taught me"

tonii141 — Tue, 20 May 2025 08:14:56 +0000

"they just totally failed to live up to expectation"

Because the expectation was too high. If you are aiming for precision, neural networks might not be the best solution for you. That is why generative AI works so well, it doesn’t need to be extremely precise. On the other hand you don't see people use neural networks in system control for cricital processes.

New comment by tonii141 in "AI in my plasma physics research didn’t go the way I expected"

tonii141 — Tue, 20 May 2025 07:46:38 +0000

This article addresses the misconception that arises when someone lacks a clear understanding of the underlying mathematics of neural networks and mistakenly believes they are a magical solution capable of solving every problem. While neural networks are powerful tools, using them effectively requires knowledge and experience to determine when they are appropriate and when alternative approaches are better suited.

New comment by tonii141 in "Chain of Thought empowers transformers to solve inherently serial problems"

tonii141 — Tue, 17 Sep 2024 10:27:30 +0000

Random generator of tokens can also solve any problem if you give it enough time and memory.