calf

37 ms·

691.
by calf 3y ago
Yeah it's an interesting divide. Two EECS professors actually teaching deep learning said in lecture nobody truly understands anything right now. Then you have many other scientists who call it a stochastic parrot, or super autocomplet
692.
by calf 3y ago
If only Michael Crichton were alive, I bet he'd have something penetrating to say about all the naive hype as well as the naive anti-hype/rationalizations.
693.
by calf 3y ago
I don't like the prevailing cynicism of dismissing the content of the message just because some bad/prejudiced/wrong factions are involved in it. It's not the way I learned to do research and answer questions. It's
694.
by calf 3y ago
If people have saved chats maybe we could all just re-ask the same queries, and see if there are any subtle differences? And then post them online for proof/comparison.
695.
by calf 3y ago
A point being moot is a reason to agree to disagree, if they can't agree on the premise. But they need to say that. If I were writing a letter, I would say it because that's just being sensible. This isn't about logical debat
696.
by calf 3y ago
Do you have any stories of how AI/DL has ignored foundational scientific problems? I do know that my old EECS professors who have pivoted toward AI and coming from adjacent/tangential research areas, are specifically interested in
697.
by calf 3y ago
Thanks for the link, I read it with interest but philosophically it is a flawed argument (IMHO). It's a nonsequitur or something, for the following two reasons. First, I'm inclined to think that longtermism is an invalid and harmf
698.
by calf 3y ago
The criterion is that p in B must halt before returning to a Program Counter value (what they define as repeating a Turing machine "state"). That's the length of the program text. So, I have no idea why they are talking about
699.
by calf 3y ago
Yes thanks I figured that out later; they define the "state" as what in CS we would call the lines of code or control flow blocks. What I'm curious about is with our usual "modern programmer's" understanding of
700.
by calf 3y ago
I got confused here: > Clearly, B is polynomial time decidable, since we need only run a program p for at most n steps, where n is the number of states in p, to determine whether or not it is in B. Couldn't program p have an exponen
701.
by calf 3y ago
Yeah if there indeed is some kind of theoretical reason that makes NP-complete problems hard to parallelize, that would be interesting to know about and understand. Maybe somebody has studied this problem. I think that unlike the story with
702.
by calf 3y ago
So I went down the rabbit hole a little and it turns out they did try all the stuff, it just hasn't panned out. Yet. It's one of those questions, theoretically it should work, but empirically it doesn't seem to work that comp
703.
by calf 3y ago
n is like a million variables so you would need 2^n gazillion cores. I'm no expert but my thesis advisor is cited in the OP article, so I'm just guessing but it's an interesting question. It's not for the lack of trying
704.
by calf 3y ago
Summarization is actually a less effective learning technique, according to Dunlosky and other literature in cognitive learning science. Dunlosky recommend other techniques namely "practice testing" and "distributed practice&
705.
by calf 3y ago
I don't have a rigorous argument but as I recall Rice's theorem says that computers cannot compute nontrivial properties about programs, in general. So we cannot have virus checkers, etc. (Well we actually do have imperfect virus
706.
by calf 3y ago
I wonder if the regulators didn't argue that Microsoft having Blizzard games would indeed pose problems for Apple hardware.
707.
by calf 3y ago
The relevance is because all (all known buildable aka algorithmic, and sufficiently powerful) models of computation are equivalent in terms of formal computability, so if you could violate/bypass the Godel or Turing theorems in neural
708.
by calf 3y ago
I spent my Sunday afternoon reading about learning strategies and this one was interesting. In STEM there's a heavy emphasis on doing problem sets for homework, but in addition to that, comprehensive textbooks will have several worked-
709.
Worked-Example Effect(en.wikipedia.org)
3 points by calf 3y ago|1 comments
710.
by calf 3y ago
I watched his first video in the series, but immediately I don't know what a loss function is. Can you recommend a book/free course that goes over the more basic concepts?
711.
by calf 3y ago
A tangent to the essay but I also really dislike the new term "alignment" and how it is used uncritically. Maybe I misunderstand the concept, but to me it is subtly oppressive and possessive. Does one characterize their children o
712.
by calf 3y ago
IMO This an incorrect and unrigorous understanding of what "internal model" means which is why there is a valid scientific debate about this issue.
713.
by calf 3y ago
I remember in 2010 a postdoc came to teach a course on model checking and the classroom was just packed with CS students. I never took it but it will be interesting to see what kind of synthesis between traditional logic and neural network
714.
by calf 3y ago
And Chomsky is in touch with other colleagues who agree with him, it's not as if his disagreement stems from being an old, isolated hermit. At the least you'd have to argue his colleagues are also mistaken.
715.
by calf 3y ago
Chomsky said that LLMs are statistical regurgitators which means LLMs can never actually reason and explain which language understanding requires. That they are a wrong model of computation by definition. It's an interesting position a
716.
Why You Should Temper Steaks Before Cooking(youtube.com)
1 points by calf 3y ago|1 comments
717.
by calf 3y ago
In restaurant cooking folklore, tempering, or leaving a raw steak in a warm kitchen for several hours, is necessary for a more evenly-cooked, pink interior; a refrigerator-cold steak takes longer to cook through, by which time the outer reg
718.
by calf 4y ago
I'm confused, people have said Notion is a relational database. Is that not the same as "trees of lists"? I thought a relational database like MySQL or whatever is formally more powerful than trees and list structures.
719.
by calf 4y ago
But if this were a college homework assignment you would be expected to explain briefly why n divides m, the comment did skip over it which can confuse people.
720.
by calf 4y ago
Floor functions trigger my fear instinct, but at least with this question I could just sit and visualize what the graph looks like, e.g. the basic sawtooth function from first year engineering.

More ›