Beyond the Liquid Tensor Experiment(xenaproject.wordpress.com)
xenaproject.wordpress.com
Beyond the Liquid Tensor Experiment
https://xenaproject.wordpress.com/2022/09/12/beyond-the-liquid-tensor-experiment/
10 comments
Heehee, my first thought reading the headline was Tony Levin's bald head...
Whoa, it’s a real paradigm shift! https://youtu.be/8tnAlPBWr0Q
I literally clicked on the main link thinking they'd release a new album or something :D
In case you didn't know, they did last year!
i'm not alone...
got a tiny bit excited for a couple seconds here
got a tiny bit excited for a couple seconds here
I’d say more like one step closer to Universal Mind.
But I all seriousness, I am happy to see that there are kindred spirits, who like Liquid Tension Experiment.
But I all seriousness, I am happy to see that there are kindred spirits, who like Liquid Tension Experiment.
Another fan, here. They're the spiritual successors to The Dregs.
The mathematicians referenced in this article are indeed fans of the band:
> The name of the challenge was the Liquid Tensor Experiment (LTE), and yes it was a homage to the Liquid Tension Experiment, a band which Scholze is fond of.
> The name of the challenge was the Liquid Tensor Experiment (LTE), and yes it was a homage to the Liquid Tension Experiment, a band which Scholze is fond of.
In case anyone didn't make it to the end:
> So where are we going and what is the point? [...] Scholze and Clausen are arguing that condensed abelian groups are an important new definition in mathematics. How can AI developers create a chatbot which helps graduate students to learn about the category of condensed abelian groups, until professional mathematicians have taught computers what a condensed abelian group is? I know that some people in the machine learning crowd would answer things like “use language models, train on the papers about condensed abelian groups” but I am not yet convinced that this alone will turn into a useful tool. Lean (and other theorem provers) offer a framework where one can create a library of useful theorems, examples, and counterexamples. We do not even have to type in all the proofs. If “we” (the mathematical community, for example PhD students in arithmetic geometry) start translating important statements, examples, counterexamples, references and so on, and somehow combine this with language models, we will at least make some kind of interesting thing, which I imagine will get more helpful to humans over time.
> So where are we going and what is the point? [...] Scholze and Clausen are arguing that condensed abelian groups are an important new definition in mathematics. How can AI developers create a chatbot which helps graduate students to learn about the category of condensed abelian groups, until professional mathematicians have taught computers what a condensed abelian group is? I know that some people in the machine learning crowd would answer things like “use language models, train on the papers about condensed abelian groups” but I am not yet convinced that this alone will turn into a useful tool. Lean (and other theorem provers) offer a framework where one can create a library of useful theorems, examples, and counterexamples. We do not even have to type in all the proofs. If “we” (the mathematical community, for example PhD students in arithmetic geometry) start translating important statements, examples, counterexamples, references and so on, and somehow combine this with language models, we will at least make some kind of interesting thing, which I imagine will get more helpful to humans over time.
[0]https://en.wikipedia.org/wiki/Liquid_Tension_Experiment