curuinor

38 ms·

421.
by curuinor 10y ago
... if you can deal with the 250-2000ms display lag sure and you have to use a remote windowing dealie
422.
by curuinor 10y ago
Causal statement in the headline. You might think only correlational statement on paper, and this is somewhat true but it is a controlled ecologicalish study. no random assignment. bad temporal proximity of data gathering on personality. me
423.
by curuinor 10y ago
I tend to believe that there are fairly deep similarities in the US's situation and Korea's situation four years ago.
424.
by curuinor 10y ago
word2vec is the latest in a long line of connectionist claims that what we call grammar does not exist. Instead of thinking of a set of deterministic rules, you think of a dynamical system which flays off language as it toddles on.
425.
by curuinor 10y ago
Player's associations aren't companies, they're unions. They do union things in the unioniest way and do a pretty alright job.
426.
by curuinor 10y ago
Get short positions individually
427.
by curuinor 10y ago
It would prolly mack on the jobs board if you actually posted it first of the month. You should prolly ask dang and see if he's cool with it
428.
by curuinor 10y ago
We should have this monthly, like the jobs postings. Think of it as startup founding postings
429.
by curuinor 10y ago
Any and every machine learning algorithm can be used for compression. Most ConvNet shit cannot be run in true real time in any way shape or form even with GPU: latencies are on the order of 500ms-1sec, so noticeable even if you have more da
430.
by curuinor 10y ago
Quite hilarious, since Reddit was one of YC's first investments
431.
by curuinor 10y ago
It is definitely the case that a fair few of those 700k sloc are devoted to poking at and wibbling with Google infra specific things. Which is of basically negative value if you're some rando using it. There are also a fair few non-inf
432.
by curuinor 10y ago
I mean, there was a nice 15 or so year period between the "new" post-Minsky-and-Papert Perceptron book connectionism and the current "new new" connectionism where neural nets were a definite backwater. Most of the PIs do
433.
by curuinor 10y ago
Nobody really cares about this stuff anymore except weirdoes like me so you might actually be best off asking your prof. There's also a lot of resurrection that goes on with weird ideas: FastFood cites a paper that cites a paper that c
434.
by curuinor 10y ago
Other things off the top of my head: Smolensky's hamonium (rbm) the LeNet papers Rumelhart's BPTT Werbos's thesis Jaeger's ESN nature paper Forget gate paper, whose author I forget (ironically)
435.
by curuinor 10y ago
One more objectionable fact about TF: There's a separate better documentation repo for Google insiders only. That's not kosher.
436.
by curuinor 10y ago
Well, if there's any overlooked nuggets of wisdom, Hinton would probably still know, and Sejnowski, and Smolensky, and McClelland, and all those other folks still doing DL stuff, as they have been doing continuously since Reagan was pr
437.
by curuinor 10y ago
Chainer is about 50k sloc of very idiomatic Python and a little dealie for emulating numpy in GPU and a bit of Cython. TensorFlow is about 700k sloc of C++, Python, some Golang for some godawful reason, with an enormous annoying build deali
438.
by curuinor 10y ago
Use dumber ML first, try some random forests. Not even because they're even that much better or worse, just because DL requires an enormous amount of knowledge and fiddliness but what you prolly want is for the bulk of the actual work
439.
by curuinor 10y ago
The PDP papers are in a bunch of volumes, in addition to the handbook. Chapter 3 of the handbook, I recall, has a little bit from Hinton's original BM work which is of interest to the cognitive scientist, as well as a pretty obstruse b
440.
by curuinor 10y ago
No PDP book? It's old and weird but interesting and has a lot of original ideas, notwithstanding the actual original backprop being from before then. Nor the original backprop stuff?
441.
by curuinor 10y ago
Both, actually. And of mechanistic statements. For example, if you have this sort of multiplicative central limit theorem with no lower bound, where you imagine independent variables being multiplied together - the attractor of that dynamic
442.
by curuinor 10y ago
Mean of lognormal distribution: e ^ (parameter + (other parameter ^ 2) / 2) Mean of pareto (power law) distribution, where parameter < 1: infinity
443.
by curuinor 10y ago
This is Richardson 1948 shit, so Richardson had the good excuse of doing all the work while people did their computations by hand. This numbnuts is not only like 60 years late, but does not have that excuse. Lot of people who don't lik
444.
by curuinor 10y ago
Clauset Shalizi Newman 2007 has not-nice things to say about the classic physicist's idiot trick of fitting power law distributions by drawing a straight line on a log-log graph: it's got huge bias. https://arxiv.org&#x
445.
by curuinor 10y ago
I had the interesting honor of contacting Edward Snowden so Perry and Ken Taylor could get a radio interview with him once for their philosophy radio show. He does cool stuff
446.
by curuinor 10y ago
There's a lotta good citations in SEP entry: https://plato.stanford.edu/entries/connectionism/
447.
by curuinor 10y ago
It's not so much resentment inasmuch all symbolic AI is evaporating in the shadow of the new new connectionist wave (the "new" connectionist wave being the PDP stuff). Why not, for example, see it from Hinton's philosoph
448.
by curuinor 10y ago
I mean, the source is, I had dinner with the guy for a school dealio and he said he didn't really like computers and had grad students to do all the programming. GEB is full of mathematical content, but it isn't hung up on compute
449.
by curuinor 10y ago
Algolia, Swiftype and Constructor.io (disclaimer: I worked at the latter) all do the learning for that stuff, at a price.
450.
by curuinor 10y ago
I heard a story from an AI professor a few years ago, who had an interesting heuristic for picking grad student applications to trash: if it mentions GEB, off it goes into the circular file. Fun book. Hofstadter actually doesn't like c

More ›