gradys

13 ms·

241.
by gradys 11y ago
So in concrete terms, does this mean that we show the network an image, choose one layer's output vector, and then back-propagate gradients to the image such that the direction of that vector stays the same, but the magnitude increases
242.
by gradys 11y ago
Does anyone have a good sense of what exactly they mean here: >Instead of exactly prescribing which feature we want the network to amplify, we can also let the network make that decision. In this case we simply feed the network an arbitr
243.
by gradys 11y ago
I'd be curious to hear more about this. You're saying that it's idiomatic in some languages to say something that translates to "dad's mother-in-law" to refer to one's maternal grandmother? What is the sem
244.
by gradys 11y ago
The difference being that credit ratings are used to evaluate how creditworthy you are by looking at tangible data that is (mostly) directly related to your use of credit and actually predictive of your credit worthiness. This system gives
245.
by gradys 12y ago
Your first sentences needs some more support. We know that some of it can be explained by statistical modelling, and we don't yet know how to explain the rest of it. Why does that make you sure that it doesn't run on statistics?
246.
by gradys 12y ago
Dan Dennett, the author of the piece that sparked this one, is an obvious intellectual ancestor to these ideas. I can't recommend his book "Intuition Pumps and other Tools for Thinking" enough. In it, he lays out his entire m
247.
by gradys 12y ago
Is there reason to believe that the same key is used on all machines?
248.
by gradys 12y ago
Numerical constants alone don't fully specify the universe. They don't account for the laws that relate numerical constants to each other. What's missing is a sort of structural constant -- a minimal set of laws that describe
249.
by gradys 12y ago
The 3D modelling is already done!
250.
by gradys 12y ago
My understanding (from an undergraduate US law class) is that you actually own the column defined by your property lines up into the heavens (and down into the Earth). You implicitly grant an easement (a formalized privilege allowing someon
251.
by gradys 12y ago
If homomorphic encryption gets off the ground, could an AI hide itself in homomorphically encrypted cloud compute capacity, earning Bitcoin to fund its survival?
252.
by gradys 12y ago
I encourage you to explore some other philosophers' thoughts on AGI. Dan Dennett in particular is an AI-optimistic philosopher of mind who directly addresses Dreyfus' AI-skepticism, quite effectively in my opinion.
253.
by gradys 12y ago
As an FYI, donations to the EFF are being matched right now: https://supporters.eff.org/donate/power-up-2014
254.
by gradys 12y ago
>They don’t tell you that a lot of programming skill is about developing a knack for asking the right questions on Google and knowing which code is best to copy-paste" This is a misleading simplification. I'm sure I'm prea
255.
by gradys 12y ago
Well, you already mentioned Dawkins, but to anyone else reading this, I can't recommend The Selfish Gene enough. It dramatically changed the way I understand nature, and with the chapter on memes, the way I understand culture. I hear T
256.
by gradys 12y ago
We have come to a lot of the same conclusions, particularly w.r.t. evolution being a metaphysical phenomenon and its being deeply interwoven with (or perhaps analogous to) intelligence. Do you have any favorite authors or texts in this area
257.
by gradys 12y ago
But notice that that's only integration between Google services. A 3rd-party API is a whole different game.
258.
by gradys 13y ago
It looks like it says there are enough cases where libraries crave two multiplication operators, elementwise multiplication and matrix multiplication, that it makes sense to add an operator to the language.
259.
by gradys 13y ago
I would say that Bing already is a viable alternative. Having used it for several months now, there aren't too many queries that it doesn't handle well, and usually when I check Google's results for those queries, they are eq
260.
by gradys 13y ago
Less energy, fewer calories. My understanding is that it's not continuousness or discreteness that determines whether "less" or "fewer" is appropriate, but whether the noun the word describes is a "count"
261.
by gradys 13y ago
Given n>40,000, statistical analysis is one approach.
262.
by gradys 13y ago
I thought it would be more plausible and probably more interesting to do this in maybe 40 questions. To do this in 33, as several others have pointed out, would require 33 questions that each almost perfectly bisect the population and are a
263.
by gradys 13y ago
Stack ranking as decided by peer review, no less.
264.
by gradys 13y ago
>... in a manner reasonably calculated to convey that such use is approved, endorsed, or authorized by the National Security Agency. Hmm.
265.
New Zealand bans software patents(arstechnica.com)
1 points by gradys 13y ago|0 comments
266.
by gradys 13y ago
I do agree with your point from the perspective that it will turn people off and marginalize the argument that bad things really are happening. At the same time, a place where 10% of communications are read by the government is not 90% free
267.
by gradys 13y ago
> Since the bitcoin ledger itself is public record anybody could analyze the addresses associated with the fund and see the inflows and outflows[1]. Even if new addresses are generated for each transaction at some point the fund manager
268.
by gradys 13y ago
Are you of the opinion that all libertarians think this way? None of the ones I know would fail to understand this.
269.
by gradys 13y ago
He probably would have been convicted anyway, but Supreme Court rulings set precedent for future cases.
270.
by gradys 13y ago
IANAL, but wouldn't this incentivize the police to just question suspects before they are under arrest?

More ›