@marcelorinesi.bsky.social
Cognitive architecture designer and consultant. On https://rinesi.com there are links to my blog and newsletters (one for articles, "what's new on arXiv," etc, the other for original short-short SF).
"Lloyd's wasn't in London: where Lloyd's was, so was the city. Rhizomatic and infertile, neither deigned to spawn yet where everywhere embedded as the secret soul of the world."
#ShortFiction #ScienceFiction #AdversarialMetanoia
I guess the tl;dr is that I'm very surprised [I did not think this would work this well] and/but I see it as improved tooling and workflows [not "AI solves math"] and/but I believe the speed of math advances is an underestimated driver of long-term human progress so improvements matter and compound.
That' also part of my distaste with vibe coding, by the way: I believe I *win* something, cognitively, with the use of programming languages (or maths, or DSLs in the most general sense). Non-natural language isn't a liability, it's like half of the power.
Of course, none of this implies (or should imply) natural languages. The *point* of much of math is to trade natural languages for ones you can prove things with and about; in that sense I think non-mathematicians misunderstand what mathematicians want (and why/how mathematics works).
Longer-term I think/hope we're on our way to a speed-up on effective large-scale collaborative math (and large-scale (in the sense of annoyingly large and detailed definitions, proofs, etc) math in general).
Perhaps ironically, the thing I'm most happily surprised by isn't new maths but the increased speed of formalization projects; if it turns out Lean + an expanded lib + some sort of LLM interface can be practical for day-to-day use, that be an undeniable win.
"Lloyd's wasn't in London: where Lloyd's was, so was the city. Rhizomatic and infertile, neither deigned to spawn yet where everywhere embedded as the secret soul of the world."
#ShortFiction #ScienceFiction #AdversarialMetanoia
There's a problem in that "LLMs + Lean et al + new workflows is incredibly powerful" is misread/misquoted/twisted into "mathematicians are just asking ChatGPT to do mathematics for them," but the frustration of having to fight that nonsense is tempered by the awe-inspiring long-term possibilities2/2
I keep regular track of some of this work (and pick up what gets to the arXiv as it does) and even for somebody who has been into computer-assisted math since using QBasic to calculate the size of Borges' Library, I'm regularly astounded by what's going on. 1/
by @alexwilkins.bsky.social
This mirrors & clarifies the way larger & more powerful organizations trap themselves by the psychological and social investment of their members in untenable assumptions about the status quo.
Or: Ig's ability to call obvious BS BS is a [literal?] survival trait these days.
via @bruces.bsky.social
... and if you aren't doing that then what even? That but for a whole class {F} of experiments? (?) I love putting things in metric spaces as much as everybody else, but it feels like it's something you [happily] discover is possible, not something you can unilaterally decide to do. 2/2
Thinking about the [/handwaves] metaphysics of the whole thing gives me a useful headache, because if you are building d(x,y) to mean "similar results under experiment F" then you're just building a model for experiment F *and* assuming things like the triangle inequality... 1/
Current global politics not beating the half of it being about gender allegations.
via @kevinr.free-dissociation.com
"Lloyd's wasn't in London: where Lloyd's was, so was the city. Rhizomatic and infertile, neither deigned to spawn yet where everywhere embedded as the secret soul of the world."
#ShortFiction #ScienceFiction #AdversarialMetanoia
Not making any prediction myself --and granting the main goal of market prices is to predict market prices-- their information processing dynamics are entirely incompatible w/their social and political role. "Markets expect" is the older "ChatGPT says," not w/o a traceable genealogy between them.
PS: Yes, I know, the literal answer to what I asked is "Senator Cory Booker, you dummy.' But that's a stunt bill, not AFAIK in any way Democrat platform.
Frankly I suspect the only reason Centrist Ds are still in control of the party is that they are much better funded than progressives, a counterpoint in a minor key of the billionaire-fueled push of Conservatives to Junior although well-paid and often driving policy partners of the Fascist wing.7/7
I'm not saying that's the root cause or main driver of the whole Fascism thing, but in a locked-in bipartisan system like the US it *is* a fundamental bottleneck to political response. 6/
Anyway, the median polls I see (*not* polls about median voters) describe a core view closer to "kidnapping bad, pedophilia bad, rich people should pay taxes, I don't care about trans people either way" which the whole professional political class seems to view as Radical. 5/
If my entire professional and social life had been defined by the idea of professional D and R politicians are the opposite ends of rational, sane policy preferences, (1) a lot of current grassroots ideas would look radical to me, (2) I'd be utterly wrong. That framing is not even close to true. 4/
Why, I'm not sure. On one hand this is people who spend a lot of money and undeniable expertise on data gathering and so on. OTOH -if they are at all like the orgs I've worked with- personal framing can overdetermine data analysis, and DC is a very unique social environment. 3/
For all that [some? tactically?] Republicans/MAGA are fighting a Democrat party that's about five million times more radical than it is, I suspect Democrat leadership sees the progressive wing as more radical than it is (while "mainstream America's" social conservatism is oversold). 2/
It goes w/o saying that Kogan has eleventytimes the access/knowledge I do but from out here... who are those Democrats and are there in any way influential? The most I see is a bottom-up call for the rich to pay *some* taxes, which would actually help fund programs. 1/
via @jonathancohn.bsky.social
Don't know yet how well it works in my use cases but it's damn clever: "[...] avoids explicit likelihood modeling [...] we place priors directly on the causal estimands and update these [...] which yields generalized posteriors for causal effects." by Emil Javurek, @dennisfrauen.bsky.social et al
Interesting on its own but one thing I'd like to highlight -which I don't think is appreciated outside the mathematical community- is how much of the recent acceleration has been driven by the assembly of new organizational, technical (and longer-term, social) *infrastructures* not just raw tech.
TIL! I had *not* known about the popularity of VAAs in Europe (or that specific term for them); their look at instabilities is interesting - also thinking now about causal impact & alternative ways to leverage them.
by Giovanni Astante, @robysinatra.bsky.social, and @vedransekara.bsky.social
"Lloyd's wasn't in London: where Lloyd's was, so was the city. Rhizomatic and infertile, neither deigned to spawn yet where everywhere embedded as the secret soul of the world."
#ShortFiction #ScienceFiction #AdversarialMetanoia
"Lloyd's wasn't in London: where Lloyd's was, so was the city. Rhizomatic and infertile, neither deigned to spawn yet where everywhere embedded as the secret soul of the world."
#ShortFiction #ScienceFiction #AdversarialMetanoia
I don't know enough about Spanish politics to have an opinion on Sánchez either way, but "We are not alone — we are the first" is a great phrase.
*If you can* it's always better to appear to have more support than you do, of course, but if you can't this is a good way to make the best of it.
I don't fully agree with the abstract: observational data being discarded in trial design due to bias concerns is in my experience far less common than a "who cares about bias just throw in the data" attitude, but anything that makes leveraging existing data more efficient and robust is a win.