Terence Tao

The https://mathstodon.xyz/ instance loads mathjax in the web interface which allows people to type latex math in their toots. I would really love it, if there was push from Mastodon, to get app developers to also load Mathjax or similar parsing library within the app itself.

As with any social network, you need to join, follow a few people, see what they say and who they quote, and then build your list of followees from there. I'm @[email protected][0][1] and I follow a few tech people. One is @cypherpunk on mastodon.social and then you can follow people quoted by them. Who would you like to follow? If you are on any Mastodon instance then you can follow anyone on any other instance. [0] https://mathstodon.xyz/@ColinTheMathmo [1] https://mathstodon.xyz/ [2] https://mastodon.social/@cypherpunk

A couple of months ago I wrote a small post [1] (in Spanish) about subsampled MRI image reconstruction using compressed sensing and how it relates to government funding issues. At the time Tao had written some posts [2] about how the IPAM, where Terrence Tao was working, was losing some funding because UCLA wouldn't follow some new federal government policies. [1] https://fintualist.com/chile/ciencia/los-efectos-de-las-polemicas-de-t… [2] https://mathstodon.xyz/@tao/114956840959338146

I find your ideas intriguing and would like to subscribe to your newsletter. Related and perhaps interesting: https://mathstodon.xyz/@mjd/114730157688607856

> For years, despite functional evidence and scientific hints accumulating, certain AI researchers continued to claim LLMs were stochastic parrots: probabilistic machines that would: 1. NOT have any representation about the meaning of the prompt. 2. NOT have any representation about what they were going to say. In 2025 finally almost everybody stopped saying so. It's interesting that Terrence Tao just released his own blog post stating that they're best viewed as stochastic generators. True he's not an AI researcher, but it does sound like he's using AI frequently with some success. "viewing the current generation of such tools primarily as a stochastic generator of sometimes clever - and often useful - thoughts and outputs may be a more productive perspective when trying to use them to solve difficult problems" [0]. [0] https://mathstodon.xyz/@tao/115722360006034040

You can add a ".rss" to mastodon profile account page to follow or read it through something that can process rss feeds for a "better" reading experience. E.g. https://mathstodon.xyz/@tao.rss

Specifically, https://mathstodon.xyz/ seems to be pretty popular among the math scientists I know, https://biologists.social/ for biologists, computer security folk seem to be using https://infosec.exchange/ Don't look at mastodon as a whole, look at specific popular servers and their niche.

That's because between the two approaches "explain me this thing" or "write code to demonstrate this thing" the LLMs are much more useful on the second path. I can ask it to calculate some third derivatives, or I can ask it to write Mathematica notebook to calculate the same derivatives, and the latter is generally correct and extremely useful as is - the former requires me to scrutinize each line of logic and calculation very carefully. It's like https://www.youtube.com/watch?v=zZr54G7ec7A where Prof. Tao uses claude to generate Lean4 proofs (which are then verifiable by machine). Great progress, very useful. While the LLM only approachs are still lacking utility for the top minds: https://mathstodon.xyz/@tao/113132502735585408

The story of Post-Quantum algorithms. Classic McEliece is getting adopted because it's solid, but not recommended. While lattice based cryptography get recommendations while having some weaknesses. New paper from Classic McEliece team has difficulty of getting past Joppe W. Bos (FrodoKEM , Kyber) and things get heated. For McEliece, when a new paper says one can see that your public key is a public key in only 2^529 operations, the reaction is: Wow, best-paper award! For Kyber, when a new paper says an attack recovers secret keys faster than breaking AES-128, the reaction is: Yeah, it's Monday. https://mathstodon.xyz/@[email protected]/114353096622329030

This convinced me to (re) create my account. Greg Egan is a fantastic author, and lives in Perth - and probably Sydney at some point. Read his books. He's also in the fediverse at https://mathstodon.xyz/@gregeganSF

Tao has an interesting followup on Mastodon: https://mathstodon.xyz/@tao/112617150031897750

Greg Egan is fairly active on Mastodon. You can follow along at https://mathstodon.xyz/@gregeganSF

Mathstodon ( https://mathstodon.xyz/about ) integrates this library. It forks/extends Mastodon to support inline LaTeX in Mastodon posts and equations, which get rendered by MathJax on the client side. - "This is important, since several math-based instances exist (such as https://mathstodon.xyz ,https://types.pl) and produce math-based posts. However, when these posts federate to other instances they cannot be rendered, since other instances do not have MathJAX installed. Thus, a more portable version based on open standards is necessary." https://github.com/mastodon/mastodon/issues/19806

> Distorting the shapes of the tile edges a tiny bit is enough to prevent the unwanted extra tilings, and produce a tile which rules out all but single-handedness non-periodic tilings of the plane by its shape alone. > The distortion can be anything you like: the precise shapes don’t matter, as long as they match. My favourite way to do this is to make each of the original edges the hypotenuse of a 1-by-2 right-angled triangle. This results in a 27-sided shape with only two different edge lengths and lots of rectangular 'tabs' and 'slots' to help fit the tiles together. Image: https://mathstodon.xyz/@OscarCunningham/110488184680656303

I use social.coop, and it seems to work well. They have some kind of participatory administration too that I don't know very well. My recommendation is to find an instance with some kind of governance plan, and hopefully a stable stream of donations. Volunteer instances indeed could go at any time. And of course, if you can, donate to one or a few instances out of good citizenship. I think it's also worth mentioning there are other services built around ActivityPub as well (and federation). I found friendica pretty good (as a Mastodon/facebook alternative), although I don't use it. Instances I can recommend to HN fellows are probably https://floss.social/about, https://chaos.social/explore and https://mathstodon.xyz/about. But don't worry too much, you can move later without too much issue (though moving posts themselves is not possible).

https://fosstodon.org/ https://mathstodon.xyz/ https://sigmoid.social

As a followup, this proof has now been declared as fatally flawed by John Carlos Baez, with some additional analysis by Noam Zeilberger. The argument is beyond my personal mathematical knowledge, but I would consider this a pretty definitive verdict. [1]: https://mathstodon.xyz/@noamzoam/109567981846531700

It was cool seeing https://mathstodon.xyz/ on the HN front page recently. I really like the idea of different organizations running their own instances. You don't need an overbearing, exhaustive and centralized system of verification for journalists when something like WAPO or NY Times can run an instance for their employees, much like how email works.

The issue in this case is not a canonical URL but rather that https://mathstodon.xyz/web/@[email protected]/109046752062665710 redirects to https://retr0.id/notice/ANqlvFZD1eIqjmRiFc. Our software has followed the redirect. This seems correct? I wonder if something has changed since you posted this?

https://mathstodon.xyz/ Is geared towards math topics. That made the rounds here earlier this year. https://news.ycombinator.com/item?id=14564624

I joined a few days ago, it's been great so far :) https://mathstodon.xyz/@j2kun

I’m in the skeptical camp. Whatever theory that will eventually emerge will not be as solid as: 1. Theory of pattern recognition (as developed in 80s and 90s) 2. Theory of thermodynamics 3. Theory of gravity 4. Theory of electromagnetism 5. Theory of relativity Etc. because of two reasons: 1. While half of deep learning is how humans construct the architecture of networks, the more important half relies on data. This data is a hodgepodge of scraped internet data (text and videos), books, user interactions etc., which really has no coherent structure 2. To extract meaningful insights from this much data, it takes models of enormous size like 10B+. The thing about random systems (in the mathematical sense) is that it takes “something” of order of magnitude bigger size to “understand” it, unless there is some concentration of measure type mathematical niceties (as in thermodynamics), which I don’t think is there in these models and data. This is the same reason I don’t think humans will ever be able to “understand” human consciousness. It will take something of an order of magnitude bigger than our own brains to do that. Here is Terence Tao explaining this concentration stuff in another context: https://mathstodon.xyz/@tao/113873092369347147 I would love to be proven wrong though.

See peer reply re: yes, your self-chosen benchmark has been reached. Generally, I've learned to warn myself off of a take when I start writing emotionally charged stuff like [1]. Without any prompting (who mentioned apps? and why would you without checking?), also, when reading minds, and assigning weak arguments, now and in my imagination of the future. [2] At the very least, [2] is a signal to let the keyboard have a rest, and ideally my mind. Bailey: > "If [there were] new LLMs...consistently solving Erdos problems at rapidly increasing rates then they'd be showing...that" Motte: > "I can['t] pop into ChatGPT and pop out Erdos proofs regularly" No less than Terence Tao, a month ago, pointing out your bailey was newly happening with the latest generation: https://mathstodon.xyz/@tao/115788262274999408. Not sure how you only saw one Erdos problem. [1] "I'll wait with bated breath for the millions of amazing apps which couldn't be coded before to start showing up" [2] "...or, more likely, be told in 6 months how these 2 benchmarks weren't the ones that should matter either"

Tao has a comment relevant to that question: "I doubt that anything resembling genuine "artificial general intelligence" is within reach of current #AI tools. However, I think a weaker, but still quite valuable, type of "artificial general cleverness" is becoming a reality in various ways. By "general cleverness", I mean the ability to solve broad classes of complex problems via somewhat ad hoc means. These means may be stochastic or the result of brute force computation; they may be ungrounded or fallible; and they may be either uninterpretable, or traceable back to similar tricks found in an AI's training data. So they would not qualify as the result of any true "intelligence". And yet, they can have a non-trivial success rate at achieving an increasingly wide spectrum of tasks, particularly when coupled with stringent verification procedures to filter out incorrect or unpromising approaches, at scales beyond what individual humans could achieve. This results in the somewhat unintuitive combination of a technology that can be very useful and impressive, while simultaneously being fundamentally unsatisfying and disappointing - somewhat akin to how one's awe at an amazingly clever magic trick can dissipate (or transform to technical respect) once one learns how the trick was performed. But perhaps this can be resolved by the realization that while cleverness and intelligence are somewhat correlated traits for humans, they are much more decoupled for AI tools (which…

>I'm not expecting novel scientific theories today. What I am expecting are signs and hints of such genius. Like I said, what exactly would you be expecting to see with the capabilities that exist today ? It's not a gotcha, it's a genuine question. >And, I'm sorry, I don't buy the excuse that these tools are not "yet" as capable as the best and brightest humans. There's nothing to buy or not buy. They simply aren't. They are unable to do a lot of the things these people do. You can't slot an LLM in place of most knowledge workers and expect everything to be fine and dandy. There's no ambiguity on that. >They contain the sum of human knowledge, far more than any individual human in history. It's not really the total sum of human knowledge but let's set that aside. Yeah so ? Einstein, Newton, Von Newman. None of these guys were privy to some super secret knowledge their contemporaries weren't so it's obviously not simply a matter of more knowledge. >Are they not intelligent, capable of thinking and reasoning? Yeah they are. And so are humans. So were the peers of all those guys. So why are only a few able to see the next step ? It's not just about knowledge, and intelligence lives in degrees/is a gradient. >If all this is true, surely we should be seeing incredible results produced by this technology. If not by itself, then surely by "amplifying" the work of the best and brightest humans. Yeah and that exists. Terence Tao has shared a lot of his (and his peers)…

(Terence Tao is his name.) Yes, he takes a rather measured view on AI, but I think for myself, not in terms of what X great person thinks. He is smarter than I am (you probably are not even aware how amazing he is, frankly, and I only say that to convey my immense admiration), and more successful by a million times, and is a millionaire with a tenure track job, and he's basically a fields medalist among fields medalists. The effect of AI on his life is very little compared to the effect of AI on mine. I am always impressed by Terence Tao, but there's basically no life lesson the average mathematician can glean from him. He is truly astounding (to be fair, there are a few other astonishing people in mathematics). The truth is that with a few more innovations, even Terence Tao will have little to add to an AI's problem solving ability. I will personally enjoy having mathematics explained to me by the AI, but it will be in relative poverty and material insecurity caused by the AI. A recent AI data point occurred this last weekend with many coming together to answer MathOverflow post, because Terence Tao answered it with some tedious parts done by AI. https://mathstodon.xyz/@tao/115325229364313459 I think his tone is optimistic, but falsely so.

Great set of observations, and indeed it's worth remembering that the specific details of assistance and setup make a difference of several orders of magnitude. And ha, he edited the last post in the thread to add this comment: > Related to this, I will not be commenting on any self-reported AI competition performance results for which the methodology was not disclosed in advance of the competition. (3/3) (This wasn't there when I first read the thread yesterday 18 hours ago; it was edited in 15 hours ago i.e. 3 hours later.) It's one of the things to admire about Terence Tao: he's always insightful even when he comments about stuff outside mathematics, while always having the mathematician's discipline of not drawing confident conclusions when data is missing. I was reminded of this because of a recent thread where some HN commenter expected him to make predictions about the future (https://news.ycombinator.com/item?id=44356367). Also reminded of Sherlock Holmes (from A Scandal in Bohemia): > “This is indeed a mystery,” I remarked. “What do you imagine that it means?” > “I have no data yet. It is a capital mistake to theorize before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts.” Edit: BTW, seeing some other commentary (here and elsewhere) about these posts is very disappointing — even when Tao explicitly says he's not commenting about any specific claim (like that of OpenAI), many people seem to be eager to…

(flailing here, to get Tomte in on this, don't mind me :) supportive case studies abound in open source-- Valve saving desktop Linux most obviously.. (Possible "the-house-that-Jack-built" counterexample is, ofc, Linus & his kernel-- it even screws with the Microsoft latent framing! Srsly, what's the monystory there?) More productively: riffing on Jobs' "if you care about sw you must work on hw", here's another "pecuny-agnostic" approach If you care about turning academics self-sufficient, you must help the self-sufficient pursue academics without disappearing into the usual failure modes (quackery, irrelevance, Bryan Johnson, Sam Altman, PG, Peter Thiel?!?! etc) The organizational issues are much the same, the relative stakes are .. interesting[0]. See the fraught debate following Terry Tao's very recent tacit endorsement of BigTech* https://mathstodon.xyz/@tao/114586574834318736 *I gather that Tao not keeping those livecoding tubes on tap may indicate that he's seriously considering the issues raised therein [0] .. than in the case of sports[1]: the mass appeal of μάθημα, to be (insanely?) charitable, is debatable, but the power potentiality, esp. when integrated over time[3], isn't[4]. (Sorry if Grothendieck/Teddy was channeled, not too inadvertently..) [1] blissfully ignoring the uh not insignificant cases of Olympic horsemen turning into field marshals [3] The Altman interval, of elite historiographers (or history-makers): exactly one thousand years [4]…

I know that this will probably be down-voted to death but I don't like these hyperbolic takes. I know that Sabine did use this title for click-bate purposes that she is now mostly doing YouTube videos (she had horrible experience that unfortunately not rare in scientific community [1]). I understand that the field of particle physics which is the corner stone in fundamental physics is not showing the great advances that it used to have a couple of decades ago. But I think people really don't understand that the field is still advancing and although these advances are less catchy to be reported in mainstream (and don't get traction if posted on HN) it is not dead or dying. There is a reason why we had a particle data group updating the PDG [2] each two years (you can order physical copies for free but please don't do if you don't need one). People were writing about that since after the big discovery of Higgs boson (that was 12 years ago). We still have a lot of measurement and puzzles that is less about unification theory that people usually would talk about. Theory people are coming up with all different ideas even if some are not testable now but that job of theorist is mainly come up with ideas and help bridge the gap later. I would suggest everyone interested in this topic to read the electroweak current chapter of the book called "How experiments End" [3] to understand a historical example to how we approached the standard model when it was first proposed. Most of the…