tonii141

8 ms·

1.
by tonii141 10mo ago
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.
2.
by tonii141 10mo ago
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.
3.
by tonii141 10mo ago
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 wit
4.
by tonii141 10mo ago
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 o
5.
Are AI Agents Compromised by Design?(computer.org)
3 points by tonii141 11mo ago|1 comments
6.
by tonii141 11mo ago
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.
7.
by tonii141 11mo ago
What do you mean by "they add instability"?
8.
by tonii141 11mo ago
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.
9.
Ask HN: Why doesn't AI use dynamical neurons instead of static activations?
1 points by tonii141 11mo ago|7 comments
10.
Fusion Power Company CFS Raises $863M More from Google, Nvidia, and Many Others(techcrunch.com)
3 points by tonii141 1y ago|0 comments
11.
by tonii141 1y ago
https://archive.is/lYTN8
12.
How to build a world model? Introduction to Laplace Neuron(abibulic.github.io)
4 points by tonii141 1y ago|0 comments
13.
How to build a world model? Introduction to Laplace Neuron(abibulic.github.io)
1 points by tonii141 1y ago|0 comments
14.
by tonii141 1y ago
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
15.
by tonii141 1y ago
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.
16.
by tonii141 1y ago
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.
17.
by tonii141 1y ago
AI is used in scene understanding for those applications, but there is no neural network that is steering the wheel.
18.
by tonii141 1y ago
"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’
19.
by tonii141 1y ago
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 neura
20.
by tonii141 2y ago
Random generator of tokens can also solve any problem if you give it enough time and memory.