currymj

42 ms·

481.
by currymj 6y ago
"Addiction By Design" by Natasha Dow Schull is a really great book about the slot machine industry. The author draws a distinction between two types of gambling addiction. One kind of addicted gambler is what you would typically t
482.
by currymj 6y ago
do you have any thoughts on what causes these systems to stick to trains? to me, as a user of public transit, they just seem somehow more pleasant, in a way I can't really make sense of. I strongly dislike riding in buses and try to av
483.
by currymj 6y ago
interesting -- i wasn't aware they were sticking that closely to the canon of the source material. it seems like in the promotional trailers released so far, they're definitely leaving anything non-futuristic-looking in the backgr
484.
by currymj 6y ago
sure, that's a fair point. Neuromancer trilogy has European family offices, it has an American media/tech company Sense/Net, Zeiss and Braun are mentioned. but the books really do feel like Japan is the dominant economic powe
485.
by currymj 6y ago
The book "Void Star" is an interesting one. It has a William Gibson style plot very similar to Neuromancer but instead of 1980s aesthetic choices, it has 2017 aesthetic choices. So instead of ninjas you get MMA jiu jitsu guys, US
486.
by currymj 6y ago
if you’re interested in causal inference, offline RL in particular is actually a problem of causal inference. unlike online RL, where you can essentially run experiments by acting on the environment and seeing what happens, in offline RL yo
487.
by currymj 6y ago
Yes, this exists. it is called the "Natural Number Game" and interactively teaches you to use the Lean theorem prover to prove simple theorems about addition, multiplication, and so on. It runs in the browser. https://w
488.
by currymj 6y ago
it’s not about physical experience, though. you just described zero, negative numbers, and 10^10^10... what’s weird is that almost all numbers are inaccessible, even in our imaginations.
489.
by currymj 6y ago
the paper lays this out: it’s just really weird that essentially all real numbers cannot be named, described, or known to us. (or computed by any Turing machine, if you want to avoid paradoxes of natural language.) if you randomly sample a
490.
by currymj 6y ago
the field is getting pretty crowded. there are probably better things to do at this point for pure career advancement and money, for the amount of effort you would have to spend to get viable professional-level skills. i haven't touche
491.
by currymj 6y ago
here's a cool example from somebody at Microsoft: https://medium.com/@ahelwer/checking-firewall-equivalence-wi... it's not a pure SAT problem, but Z3 works (in a way I don't understand) by reducing const
492.
by currymj 6y ago
many of those sampling heavy American music styles (hip-hop, much electronic music) were very directly and strongly influenced by Jamaican dub which grew out of reggae. I think the argument is reasonable that we don't get any of it, at
493.
by currymj 6y ago
that's exactly it -- they take an existing form for the ansatz (or the general idea of it, at least), and make it more flexible by replacing pieces with neural networks that have many more parameters, while maintaining constraints requ
494.
by currymj 6y ago
https://arxiv.org/abs/1909.08423 and https://arxiv.org/abs/1909.02487 are some examples I've been looking at recently.
495.
by currymj 6y ago
this is very cool. I was thinking specifically of this and related approaches https://arxiv.org/abs/1909.08423 where they search for the ground state by iteratively using an MCMC sampler and doing SGD. The innovation
496.
by currymj 6y ago
There are lots of things that I think are comparable to machine learning in the sense that they combine applied math and heavy computation and are very practically important, like simulating chemical reactions, solving operations research p
497.
by currymj 6y ago
Some applications in computational physics involve solving a "variational" problem, where you have some parameterized function and try to numerically find the parameters that minimize energy or error. This does not necessarily inv
498.
by currymj 6y ago
I would be willing to bet that refusing to wear a mask is strongly correlated with spending a lot of time reading about news and politics online.
499.
by currymj 6y ago
it's interesting that he says that proving a mathematical theorem would be a bigger milestone. from my limited perspective, that seems less surprising than getting language models of GPT-3 quality. There are already theorem proving env
500.
by currymj 6y ago
there is some truth in this, frustratingly. at the same time, if you go to this year's ICML papers and ctrl-F "policy", there are several RL papers that come up with a new variant on policy gradient and validate it using only
501.
by currymj 6y ago
well, you don't actually have to have your models fight their models. there's lots of things to try that aren't the SOTA rat race. like, if you want to show relative improvement of some new variation of an RL algorithm, this
502.
by currymj 6y ago
there are many papers that seem like obvious crap that get into supposedly prestigious conferences (NeurIPS, ICML) every year. I think many people might reasonably believe that a paper in the supposed best venue in the field will be worth r
503.
by currymj 6y ago
wait, does Vanguard S&P 500 or SPY actually not own stock in all 500 companies? that is very surprising!
504.
by currymj 6y ago
in some limited cases it seems like maybe we can be pretty confident that retail investors "caused" a price move. even if professionals magnify it, the core thing would not happen without retail investors. Like the Hertz thing, or
505.
by currymj 6y ago
concretely, how much does the sentiment of retail investors actually matter to prices -- in general for the whole market, and for specific stocks that dumb retail investors get excited about (Tesla, Hertz, or whatever)? I have no idea what
506.
by currymj 6y ago
The theory of causal inference tells you how to do this -- what is the nature of the relationship between correlation and causation? When can we draw conclusions from natural experiments? There is actual rigorous theory that has very clear
507.
by currymj 6y ago
consider it from an alternative perspective — sending a 14 year old to grad school to get their PhD by 20 is putting them in a 60-hour-a-week high pressure job that causes mental breakdowns in many adults. i don’t doubt that academic accele
508.
by currymj 6y ago
the focus of the project seems not to be on differential privacy as used by private companies but by scientists. in many cases you can't (US Census, medical records) or wouldn't want to (public health research) stop collecting the
509.
by currymj 6y ago
thanks for posting the article, which is fascinating and presents a complex and interesting picture of how archaeologists and anthropologists are reacting to advances in genetics -- much more interesting than your summary suggests.
510.
by currymj 6y ago
i don't know, I think the new criticism didn't deny that books were written by specific authors, the authors had intentions in their minds, wrote in a historical context, and so on -- they just felt it was not interesting or fruit

More ›