taliesinb

I'd love to see a version of the article marked up with information about which parts of these examples were fraudulent, and how, all with Matt Levine-style snark.

but http://tmfassociates.com/blog/2019/12/12/reality-and-hype-in... less so

On rewatching it I'm certainly not as impressed. Maybe I was just enjoying watching some anti-hype for once. Currently reading https://caseyhandmer.wordpress.com/2019/11/02/starlink-is-a-... which seems more optimistic but http://tmfassociates.com/blog/2019/12/12/reality-and-hype-in... less so

The video doesn't make that assumption. It has relatively straightforward calculations about the size of the market, the existing competitors, the number of satellites needed to support a customer base of whatever size given SpaceX's own numbers. And the economics don't make much if any sense. The only objection I had is that low latency internet is more useful than just to gamers. E.g. voip and video calls, even ordinary browsing are unbearable with high latency connections.

Can you elaborate on what is wrong with it?

I’ve been fairly skeptical about Starlink since watching Common Sense Skeptic’s video, which I recommend: https://youtu.be/2vuMzGhc1cg

Why do you think that? (I happen to agree)

I've been exploring hypergraphs a lot over the last 12 months. They are an incredibly rich way to look at a lot of existing (and new) mathematics, so much so that I think hypergraphs will have a good chance to become a signature mathematical structure over the next 100 years.

Unfortunately people don't yet have a good notion (or notation) for what a hypergraph even is! The wikipedia page is a disaster area, for example. Partly this is due to a lack of careful, general groundwork on how to think about hypergraphs. I'm confident this will get done though -- and aim to play a role in that. My personal view is that you need you really need to adopt an ADT-type viewpoint to do this, but I'll have to leave it at that for now.

Funny this showing up here -- fractons are one of the first things I came across in pursuing my strange belief that fundamental particles are manifestations of defects in some "space-time crystal". But where Seiberg says “Quantum field theory is a very delicate structure, so we would like to change the rules as little as possible,” -- I have the exact opposite view: we'll probably need to tear down the vast majority of mathematical structures that are the foundation of QFT and QM, and build from scratch on a discrete combinatorial foundation. Luckily doing that is turning out to be a lot of fun and involve some interesting new mathematical ideas. But more often old mathematical ideas, like curvature, holonomy, and gauge symmetry, but resurrected in new forms.

Just thought the same thing. Funny that they spent probably an entire minute worth of skit on it.

I've been learning representation theory (among other topics) from his Youtube lectures! They're a real treasure trove. And he's got a dry sense of humor that I quite like.

I can also recommend this interview with him by Curt Jaimungal, which is filled with his insights and opinions about his practice of mathematics, and his life as a mathematician: https://youtu.be/xu15ZbxxnUQ

I'm certainly not claiming to be a "real physicist", whatever that is, but I'm involved in the project (I actually named and continue to work on the "token-event graphs" that Max Piskunov developed for his local multi-way systems), and I don't think the project has convincingly shown yet how the basic ingredients of Hilbert spaces, complex amplitudes, operator algebras, density matrices, etc. are supposed to work. Which to me is an extremely important first step. I think we'll fix that, though.

The attention ecosystems of entertainment, science, computing seem to favor a model people who are good at self-promoting a particular "brand" of work that really represents the labor of a much larger group of people. This simplification is useful for a couple of reasons: it nurtures the ego of the brand-manager, which almost has to be outsize to play this role. It is a simple enough story for the public to understand, and efficiently creates a large "catchment area" for people to become interested in the brand. And the brand managers can sometimes be good coordinators.

Funny coincidence that two days ago I rewatched James Cameron's 1989 movie The Abyss, in which perflourocarbon breathing features as a major plot device. Great movie!

Who is the author of that quote?

Notability (and also concepts) has turned into a big part of my process of "live thinking".

A lot of what you said has resonated deeply with me. Is there any way I can get in touch to discuss further?

While the continual one-upmanship of ever more intricate biologically plausible learning rules is interesting to observe (and I played around at one point with a variant of the original feedback alignment), I think OP's alternative view is more plausible.

Fwiw I am involved in an ongoing project that is investigating a biologically plausible model for generating connectomes (as neuroscientists like to call them). The connectome-generator happens (coincidentally) to be a neural network. But exactly as the OP points out, this "neural network" need not actually represent a biological brain -- in our case it's actually a hypernetwork representing the process of gene expression, which in turn generates the biological network. Backprop is then applied to this hypernetwork as a (more efficient) proxy for evolution. In the most extreme case there need not be any learning at all at the level of an individual organism. You can see this as the ultimate end-point of so-called Baldwinian evolution, which is the hypothesized process whereby more and more of the statistics of a task are "pulled back" into genetically encoded priors over time.

But for me the more interesting question is how to approach the information flow from tasks (or 'fitness') to brains to genes on successively longer time scales. Can that be done with information theory, or perhaps with some generalization of it? I also think it is a rich and interesting challenge to parameterize learning rules in such a way that evolution (or even random search) can efficiently find good ones for rapid learning of specific kinds of task. My gut feeling is that biological intelligence has many components that are ultimately discrete computations, and we'll discover that those are reachable by random search if we can just get the substrate right, and in fact this is how evolution has often done it -- shades of Gould and Eldredge's "punctuated equilibrium".

(if anyone is interested in discussing any of these things feel free to drop me an email)

So personally I'm quite excited about the whole "multiway" point of view on non-determinism. The causal graphs that Stephen presents here are one way of summarizing how local events interact with one another across the space of possible histories, but you can go further. I'm working on a generalization of this idea, which I'm calling a token event graph, which "factors" causally distinct subsystems of a rewrite system (such as an NDTM, a hypergraph rewriting system, a string rewriting system, etc.) in a clean and precise way. While it is still early days, there is already an intriguing amount of rich mathematical structure to describe and formalize. I hope to publish a blogpost about this soon.

You might like https://twitter.com/xbresson/status/1271124929505714176?s=12

How do you know it’s inkscape? Also, is inkscape the best for this kind of thing? I’ve grown kind of fond of Affinity

Is anyone aware of a resource that collects examples of what complex numbers “encode” in various settings in which they naturally appear?

For example, complex numbers in the eigendecomposition of a real-valued square matrix encode permutations among dimensions in a particular eigenspace via the roots of unity — which amount to rotations. So they’re a bookkeeping device that allows the eigendecomposition of an automorphism of a real vector space to actually work in all cases.

From people with more experience than me: does JAX have a good story about how to manage state cleanly, where it needs to occur?

Haha I was going to mention Reich Remixed! I spent a lot of time programming to that music when I was a teen. Definitely fun!

My favorite track was probably Nobukazu Takemura’s remix of Proverb https://youtu.be/-ipSM8bwo4M

Reich is brilliant! “Come out” is a very affecting piece, and the history behind it is also quite interesting.

Also, judging from the name, OP is probably also a fan of Arvo Part. There is a moving “video etude” for the piece Healing of Arinushka that I would recommend if you’re feeling contemplative https://www.youtube.com/watch?v=H7exvD2nJG8

Out of curiosity, are any of those projects public?

If video is more your thing, Stephen just wrapped up a long, interesting livestream about combinators, and the life of Moses: https://www.youtube.com/watch?v=PG2G5xSz0NQ

That’s cool.

What do you make of the algebra of the dihedrons? https://youtu.be/lqH4BLHGsFw . It’s a “sister algebra” of the quaternions.

Yeah. There is this system wide bug on iOS where the animations actually BLOCK user input / state changes. You can see this if you type in your pass code fast enough on the lock screen, such that you are interrupting the animated fade in of the dot for the previous digit. You’ll often lose a digit and have to start again from scratch. Infuriating! It’s the most fundamental relationship to betray, the relationship between touch and input. It registered my touch, but decided to ignore it because it wasn’t done with its pointless animation that some manager thought would look neat. Same thing happened with calculator and they fixed that one case, whereas it’s clearly a fundamental issue with UIKit.

Thanks, I do appreciate you extracting this. I did follow one or two of these originally. But I'm not going to buy an ebook on Amazon to figure out if he has anything useful to say.

Hey Adele! Thanks for the resource. Here's something a bit wilder I'm still trying to wrap my head around:

https://www.sciencedirect.com/science/article/abs/pii/S09252...