6 ms·

> Something remarkable has happened these past two years. For 45 years I’ve devoted myself to building a taller and taller tower of science and technology—which

> Something remarkable has happened these past two years. For 45 years I’ve devoted myself to building a taller and taller tower of science and technology—which along the way has delivered many outputs of which I’m quite proud. But starting in 2020 with the unexpected breakthroughs of our Wolfram Physics Project we’ve jumped to a whole new level. And suddenly—yes, building on our multi-decade tower—it seems as if we’ve found a new paradigm that’s incredibly powerful, and that’s going to let us tackle an almost absurd range of longstanding questions in all sorts of areas of science. An author who sets a piece up like that had better deliver the goods - big time. The author does not. You might expect that "new paradigm" of his to have rocked the scientific world. It hasn't. The big idea is: "Multicomputation." And despite voluminous writing on the topic it's nothing more than noodling cellular automata in ways that have had no effect whatsoever on chemistry, physics, or biology.

narrator 5y ago
I'd say that gradient descent and deep learning have made obsolete whatever cellular automata stuff Wolfram has come up with. Neural nets can approximate any function. Cellular automata has a much more limited range that only includes the generation of some interesting pictures and not much else of interest, IMHO. To be charitable, if he had some way to mathematically slide between and combine networks of cellular automata, he might have something, but it would probably just end up reinventing deep learning and neural nets on a slightly different set of primitives.
- Stupulous 5y ago
  Cellular automata can be Turing complete, so they cover at least the entire range of a neural net, and more if it's not recurrent.
kspacewalk2 5y ago
>An author who sets a piece up like that had better deliver the goods - big time. The author does not. That's been my impression of Stephen Wolfram and what he does in general. He has some accomplishments but his ego is comically large by comparison with those. Very hard to take him seriously.
- jacklizhui 5y ago
  Very hard? You miswrote. You meant impossible.
daneelsan 5y ago
> it's nothing more than noodling cellular automata in ways that have had no effect whatsoever on chemistry, physics, or biology. I mean, you could at least read a bit about what his Multicomputation stuff is really studying. CA are not being noodled there.
- copperx 5y ago
  For those of us that are not going to read Wolfram's papers, what is multicomputation?
  - convexfunction 5y ago
    Seconded -- I don't have much interest in reading the guy's own writing at this point, but I'll gladly read a concise summary that leaves out all the boasting since I assume there might be some actual "content" buried in there.
  - wildmanx 5y ago
    In short, Wolfram uses this term to describe distributed local state updates of a global state space. In his particular model, the global state space is a hypergraph, and the state updates are replacements of some local subgraphs with other subgraphs. This happens in parallel in all sorts of places, which is a not-too-surprising generalization of cellular automata. And that's it. It's not very deep, and nobody outside his sphere of influence uses this term for that purpose. The thing with Stephen Wolfram is that he invents all those terms, uses them as if they are standard terminology in the field (of physics or computer science) while freely mixing them with _actual_ standard terminology. That goes for "branchial graph", "rurial space", "principle of computational equivalence" and also "multicomputation". He is just diving deeper and deeper into his own buzzwordial space.
narratorr 5y ago
stuntkite 5y ago
That's not fully true that multicomputation and Wolfram's research with cellular automata and combinators is nothing more than noodling. I think there are some very interesting applications coming soon and the decentralized VPN product NKN/NKS/nConnect is doing something with Wolfram's research in application that I think has wider applications than their current model. This discussion between the CEO of NKN and Wolfram is pretty good. [0] I am not saying it is the most important thing happening in "multicomputation" but I think there's a lot that this approach can lead to when we start talking about using clusters of computers, GPUs, TPUs, cell phones, edge devices, etc in a secure fashion. I think there's something to it, but I am also an idiot and could be very wrong. [0] https://medium.com/nknetwork/stephen-wolfram-creator-of-nks-tries-to-understand-nkn-b4018d747a3 https://medium.com/nknetwork/stephen-wolfram-creator-of-nks-...

narrator 5y ago

I'd say that gradient descent and deep learning have made obsolete whatever cellular automata stuff Wolfram has come up with. Neural nets can approximate any function. Cellular automata has a much more limited range that only includes the generation of some interesting pictures and not much else of interest, IMHO. To be charitable, if he had some way to mathematically slide between and combine networks of cellular automata, he might have something, but it would probably just end up reinventing deep learning and neural nets on a slightly different set of primitives.

Stupulous 5y ago
Cellular automata can be Turing complete, so they cover at least the entire range of a neural net, and more if it's not recurrent.

kspacewalk2 5y ago

>An author who sets a piece up like that had better deliver the goods - big time. The author does not. That's been my impression of Stephen Wolfram and what he does in general. He has some accomplishments but his ego is comically large by comparison with those. Very hard to take him seriously.

jacklizhui 5y ago
Very hard? You miswrote. You meant impossible.

daneelsan 5y ago

> it's nothing more than noodling cellular automata in ways that have had no effect whatsoever on chemistry, physics, or biology. I mean, you could at least read a bit about what his Multicomputation stuff is really studying. CA are not being noodled there.

copperx 5y ago
For those of us that are not going to read Wolfram's papers, what is multicomputation?
- convexfunction 5y ago
  Seconded -- I don't have much interest in reading the guy's own writing at this point, but I'll gladly read a concise summary that leaves out all the boasting since I assume there might be some actual "content" buried in there.
- wildmanx 5y ago
  In short, Wolfram uses this term to describe distributed local state updates of a global state space. In his particular model, the global state space is a hypergraph, and the state updates are replacements of some local subgraphs with other subgraphs. This happens in parallel in all sorts of places, which is a not-too-surprising generalization of cellular automata. And that's it. It's not very deep, and nobody outside his sphere of influence uses this term for that purpose. The thing with Stephen Wolfram is that he invents all those terms, uses them as if they are standard terminology in the field (of physics or computer science) while freely mixing them with _actual_ standard terminology. That goes for "branchial graph", "rurial space", "principle of computational equivalence" and also "multicomputation". He is just diving deeper and deeper into his own buzzwordial space.

narratorr 5y ago

stuntkite 5y ago

That's not fully true that multicomputation and Wolfram's research with cellular automata and combinators is nothing more than noodling. I think there are some very interesting applications coming soon and the decentralized VPN product NKN/NKS/nConnect is doing something with Wolfram's research in application that I think has wider applications than their current model. This discussion between the CEO of NKN and Wolfram is pretty good. [0] I am not saying it is the most important thing happening in "multicomputation" but I think there's a lot that this approach can lead to when we start talking about using clusters of computers, GPUs, TPUs, cell phones, edge devices, etc in a secure fashion. I think there's something to it, but I am also an idiot and could be very wrong. [0] https://medium.com/nknetwork/stephen-wolfram-creator-of-nks-tries-to-understand-nkn-b4018d747a3 https://medium.com/nknetwork/stephen-wolfram-creator-of-nks-...