pishpash

16 ms·

271.
by pishpash 2y ago
A proof solver existed. They were improving the proof solver explicitly by making the formalizer a part of the training. Formalizer reliability is the key novelty. It turns out it was only reliable enough for training. So unless they made t
272.
by pishpash 2y ago
Gradient descent is a search. Where does it say the search space has to be small?
273.
by pishpash 2y ago
Yet the facts at hand are the opposite of what you say. Reliable formalizer was the more difficult problem than solving formalized IMO problems, because they have not produced one.
274.
by pishpash 2y ago
There's no problem with search. The goal is to search most efficiently.
275.
by pishpash 2y ago
The AI is the "solver network", which is the (directed) search over solutions generated by Lean. The AI is in doing an efficient search, I suppose. I'm also waiting for my answer on the role of the Gemini formalizer, but read
276.
by pishpash 2y ago
Exactly, a problem and its answer are just different ways of describing the same object. Every step of a proof is a transformation/translation of the same object. It would be disingenuous to say that some heavy lifting isn't don
277.
by pishpash 2y ago
Crowdstrike is malware and spyware. Trusting one malware to control another is your problem right there. It will always blow up in your face.
278.
by pishpash 2y ago
That's what you get for outsourcing to some generic shop with no domain expertise who implements to a spec for the lowest dollar.
279.
by pishpash 2y ago
Non sequitur. Read carefully, there isn't all work in all places. It's not about how much it pays. It's the demand to come to an office in "a coastal urban city" to work when the work can be completed elsewhere. It
280.
by pishpash 2y ago
Lots of employers think they're entitled to in-office days when remote work suffices. Argument would be stronger if there were a whole country to do all work in.
281.
by pishpash 2y ago
Mm
282.
by pishpash 2y ago
So the work to simplify ML models, reduce dimensions, etc. becomes the numeric way to seek simple actual scientific models. Scientific computing and science become one.
283.
by pishpash 2y ago
They do tell you. What's missing is compensation for waiting or making new arrangements last minute (which is not cheap) unless it's 3 hours delayed. So airlines will drag it out putting fake new flight time up by incrementing 10
284.
by pishpash 3y ago
It doesn't say that's not happening. It says they are winning cases.
285.
by pishpash 3y ago
Cover on, SH is playing the Debbie Downer character so needs to look like that.
286.
by pishpash 3y ago
There is no "diffusion over time" with variable premia and short terms. Insurance companies already know the probability of an event within a risk pool over a term. That's not what this is talking about. They are looking to c
287.
by pishpash 3y ago
It is just a very high rate if they could charge it, but that's the point: when the uncertainty is nil, the premium will always be the replacement cost when the replacement is about to happen, so that there's no insurance. Put ano
288.
by pishpash 3y ago
The insurance paradox. If you can price every risk precisely, meaning the variance on your prediction is zero, there's no point to insurance because it would be the same as a savings account. There's no pooling of risk at all.
289.
by pishpash 3y ago
That's olfactory adaptation.
290.
by pishpash 3y ago
Property tax = you do not own the land, the government does and leases it to you. Setting the rate to the highest rate is stupid because it means the government confiscates all the land and leases it only to e.g. casinos and you get nothing
291.
by pishpash 3y ago
Where does the tax money go? If it's redistributed back to everyone including those who rent then that's just land redistribution.
292.
by pishpash 3y ago
It's morally wrong to take somebody else's land, too. That's not sharing.
293.
by pishpash 3y ago
What evidence do you need? NYC and Boston smelling like sewer in the summer is enough for me.
294.
by pishpash 3y ago
Stupid title, land use policies make everything that uses land expensive. I can assure you childcare being expensive is not bottlenecked by land, but by caretaker's wage levels driven by the fact that they need to rent in the same expe
295.
by pishpash 3y ago
So it's a tax. Fine, don't call it congestion pricing.
296.
by pishpash 3y ago
The externality applies to residents who do not take public transit too. The two decisions should not be coupled. If you want to do a Pigovian tax, do a Pigovian tax properly. If you want to do congestion pricing, make it budget neutral to
297.
by pishpash 3y ago
"...raise $1 billion annually for public transit improvements" The money isn't going to residents.
298.
by pishpash 3y ago
It's called the plasmid exchange. The search is parallelized to however many individuals there are, which is arbitrarily large, then the result is shared.
299.
by pishpash 3y ago
Almost every model we build of more trivial things based on observation always turns out to be not really right. I cannot imagine why one of the universe that has had multiple version updates in the last 100 years to not also be grossly mis
300.
by pishpash 3y ago
Maybe if you do it in person, but they must have direct shipping flows where nobody checks.

More ›