Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

A Eureka machine that thinks like nature and explores what AI cannot

iisc.ac.in

Post reply on HN
Print view View original

41–49 of 49 posts

Re: A Eureka machine that thinks like nature and explores what AI cannot

#41

Post by unmole »

ngriffiths wrote: post #39
stared wrote: post #35

The whole title is a buzzword cluster, until proven otherwise. Which tasks, in particular, does it do better? Not as in "it could do them better", but actually there are benchmarks. If they are, they are buried beneath marketing; if not - well, we have our answer. What is "thinks like nature"? Spin systems, are no more (or less) nature than transistors. That said, I am all for exploring various systems for computatio…

Yeah, I mean it's obviously meant to be a marketing pitch but it's not a very good one. > The hardest computational problems are not waiting for faster chips – they are waiting for machines that compute in a fundamentally different way. Surely they don't actually believe that, right? Like you say the benefits must be limited to specific shapes of problems (not all of "the hardest" ones), and the whole history of comp…

You're describing The Hardware Lottery: https://arxiv.org/abs/2009.06489

Re: A Eureka machine that thinks like nature and explores what AI cannot

#42

Post by ninjagoo »

Seems like the key elements in this are the use of a neuromorphic autoencoder (instead of a 'regular' one), plus Fowler–Nordheim annealing dynamics and Ising energy minimization so that the system is not just passively settling, it's being taken through a controlled search process designed to avoid premature trapping and scale to higher-order combinatorial optimization problems. [1]

A 'regular' autoencoder is a neural network trained to compress data and then reconstruct it.

A neuromorphic autoencoder is instead implemented using brain-inspired computing elements like spiking neurons, event-driven updates, local interactions, sometimes specialized hardware. In this paper, looks like the autoencoder is being used as a structured energy-minimizing circuit for an Ising optimization problem. The architecture manipulates Ising clauses rather than only pairwise spin interactions.

Ordinary artificial neurons compute matrix ops such as y=f(Wx+b), while this uses artificial neurons that accumulate input, which emit a spike when they cross a threshold, like biological neurons (event driven neural dynamics).

[1] https://www.nature.com/articles/s41467-026-71937-4

Re: A Eureka machine that thinks like nature and explores what AI cannot

#43

Post by otikik »

stared wrote: post #35

The whole title is a buzzword cluster, until proven otherwise. Which tasks, in particular, does it do better? Not as in "it could do them better", but actually there are benchmarks. If they are, they are buried beneath marketing; if not - well, we have our answer. What is "thinks like nature"? Spin systems, are no more (or less) nature than transistors. That said, I am all for exploring various systems for computatio…

Yeah, I was reminded of the Rockwell Retro Encabulator a little bit.

https://www.youtube.com/watch?v=RXJKdh1KZ0w

Re: A Eureka machine that thinks like nature and explores what AI cannot

#44

Post by adrianwaj »

K0balt wrote: post #38

I wonder how this compares to thermal wells? https://extropic.ai They seem to work in a similar way, sampling from chaotic datasets to find the lowest energy state. Is one fundamentally more scalable? More efficient?

Computers that don't require cooling - so cool - literally!

Re: A Eureka machine that thinks like nature and explores what AI cannot

#45

Post by thesz »

dave1010uk wrote: post #16

> Explores what AI cannot In other words, gradient descent isn't good at combinatorial optimisation. I'm sure the research is better but the hype in the blog post leaves a bad taste. There must be a version of Rich Sutton’s Bitter Lesson that applies to alternative computing like this, along with all the other exciting specialised hardware we've seen come and go over the years, like expert systems, optical computing,…

  > gradient descent isn't good at combinatorial optimisation.
If you convolve your problem with sufficiently wide Gaussian, you can use gradient descent. The approach is called Natural Evolution Strategies [1].

[1] https://en.wikipedia.org/wiki/Natural_evolution_strategy#Nat...

It requires O(N^4) evaluations to compute Fisher Information Matrix for N-dimensional parameterization of the problem in original formulation. But there are closed form solutions and more economical representations of covariance matrix (LoRA, hehe).

Re: A Eureka machine that thinks like nature and explores what AI cannot

#46

Post by ngriffiths »

unmole wrote: post #41
ngriffiths wrote: post #39

Earlier quoted context omitted.

Yeah, I mean it's obviously meant to be a marketing pitch but it's not a very good one. > The hardest computational problems are not waiting for faster chips – they are waiting for machines that compute in a fundamentally different way. Surely they don't actually believe that, right? Like you say the benefits must be limited to specific shapes of problems (not all of "the hardest" ones), and the whole history of comp…

You're describing The Hardware Lottery: https://arxiv.org/abs/2009.06489

That's interesting, thanks. I only read the abstract so far but was immediately reminded of this recent HN submission[1] and the whole thing that certain ideas go together, and so they are adopted together, but the resulting bundle of ideas might be poorly suited to certain problems.

[1]: https://news.ycombinator.com/item?id=48237163

Re: A Eureka machine that thinks like nature and explores what AI cannot

#48

Post by CamperBob2 »

bitwize wrote: post #29
dave1010uk wrote: post #16

> Explores what AI cannot In other words, gradient descent isn't good at combinatorial optimisation. I'm sure the research is better but the hype in the blog post leaves a bad taste. There must be a version of Rich Sutton’s Bitter Lesson that applies to alternative computing like this, along with all the other exciting specialised hardware we've seen come and go over the years, like expert systems, optical computing,…

> General purpose commodity silicon with rapidly evolving software generally beats specialised hardware. All of the Amiga people are sighing right now, as they recall how their beautiful, elegant system synergistically designed with custom chips was outpaced by CPU/memory brute force in the early 90s.

Likewise, the lesson Seymour Cray had to learn a couple of times because it didn't quite sink in the first time: "Never go up against CMOS when your company is on the line."

Re: A Eureka machine that thinks like nature and explores what AI cannot

#49

Post by tmzt »

anthk wrote: post #23
dave1010uk wrote: post #16

> Explores what AI cannot In other words, gradient descent isn't good at combinatorial optimisation. I'm sure the research is better but the hype in the blog post leaves a bad taste. There must be a version of Rich Sutton’s Bitter Lesson that applies to alternative computing like this, along with all the other exciting specialised hardware we've seen come and go over the years, like expert systems, optical computing,…

In hardware Prolog/Kanren/expert systems? That would be possible with libre microcode for Intel, and not this spyware corporate shithole we are living it. We would be able to switch microcode at boot and set one for security, another one for C performance, others for Lisp performance and so on.

Sounds a bit like Transmeta Crusoe.
Post reply on HN