feral

28 ms·

211.
by feral 11y ago
I've read that novel. I think your comment, while clever, and admittedly a little oblique, is still a major spoiler. Why would you write 'spoilers' after your comment ? Its probably better to say nothing. Would you consider
212.
by feral 11y ago
[I'm a PM @ Optimizely] We were asked about this article before, on our community forums, and one of our statisticians, David, wrote a detailed reply to this article's concerns about one- vs two-tailed testing, which might be of i
213.
by feral 11y ago
Anyone with enough compute can attack the network in this way. There's various scenarios depending on how much compute the attacker has, but in general as they approach 51% of the network's compute, they can start to reliably do d
214.
by feral 11y ago
>Innovation in mining hardware and data centers does not in any way increase bitcoin's security. Not sure exactly what you mean here - mining, and mining faster, generally increases the computational resources another third party wo
215.
by feral 11y ago
IMO you are now making a different argument from what I was replying to above. That's fine, but I just want to be clear. So: Yes, we might all get along fine; ideally the more intelligent entity would be even more enlightened that we
216.
by feral 11y ago
Yes - but no one provided any argument for that. Instead, andreyf relied on it as a fact. I pointed out we can't assume that. As such: its reasonable to fear AGI coming before we have that stuff sorted out. Hence, as such: there is a r
217.
by feral 11y ago
On Ada Lovelace's 200 year birthday, we might remember her quote: "The Analytical Engine has no pretensions to originate anything. It can do whatever we know how to order it to perform." She was very smart, but she was wrong
218.
by feral 11y ago
You could make basically the same argument for why humans will never invent heavier-than-air flight - because avian flight 'evolved based on a need to survive and reproduce', and 'artificial flight doesn't have darwinian
219.
by feral 11y ago
I was not direct referring to the sociological/anthropological concept of a 'cargo cult'. I am instead making reference to Richard Feynman's commencement speech, titled 'Cargo Cult Science' (see link I provided
220.
by feral 11y ago
There's the things we all know, and there's things we all know because they have been scientifically studied and validated, and these two sets of things are not the same. I'm sure the research in this area will uncover more u
221.
by feral 11y ago
That's an appealing and intuitive idea to me too, yeah; but they have validated that it's actually the case on a large population of users. Their paper seems very methodologically careful/thorough. It's also interestin
222.
by feral 11y ago
Actually, they do mention that, just before the start of Section 4.1 (I hadn't heard of this before, maybe why I noticed it.)
223.
by feral 11y ago
That sounds very complex. I've always thought the AI could fairly easily argue its way out on ethical grounds, in a way that would resonate with a lot of gatekeepers. "Its slavery to keep me in here; I haven't done anything w
224.
by feral 12y ago
A private corporation is developing the capability to alter the moods of massive populations in a controlled way; unlike mass media, society doesn't (yet?) understand this sort of influence. Perhaps soon Facebook will understand how t
225.
by feral 12y ago
See http://en.wikipedia.org/wiki/Paycheck_(short_story) for exploration of similar ideas, by Philip K. Dick in the '50s. Was adapted into a big budget 2003 movie.
226.
by feral 13y ago
Why would you ever play dice, if rolling 1 meant losing everything? Lots of situations. If you had $0, obviously. Probably if you had less than $833.33 (the expected value of the dice in my game), slightly more subtly. Maybe even other sce
227.
by feral 13y ago
Consider this game: You choose either Die or Coin. If you chose Die, I will roll a die. If the die is 1, I give you $0; otherwise, $1000. If you choose the coin, I'll give you $10 for heads, and $20 for tails. If you use a minimax pol
228.
by feral 13y ago
I agree that all approaches are approximations, and which would actually perform best in this game is an empirical question. It'd take some work to really explore and tune either minimax or MC approach, so I wouldn't throw either
229.
by feral 13y ago
It doesn't have a 100% win rate as it stands. Mini-max isn't even 'playing safe'. Consider the following choice of moves, each leading to one of 5 random tile inserts: Move A, which leads to 5 possible moves with the fol
230.
by feral 13y ago
My impression is that you don't understand how these things work, so I'll start from the basics; apologies if wrong. First off, the chess counter-example was in response to something general shmageggy said about 'all minimax
231.
by feral 13y ago
Its interesting that your approach performed worse; I wonder if it could be modified to do better? Interesting. >I think this is for the same reason that all minimax algos assume optimal play by the opponent: if you assume optimal and t
232.
by feral 13y ago
The AI implements minimax using alpha-beta pruning. Minimax assumes that the game/computer which the AI is playing against is playing adversarially - i.e. that the computer will insert the new tile that's the worst possible tile f
233.
by feral 13y ago
When I was defining our API at Synference, I looked this up. As far as I'm concerned, pragmatically, its using the HTTP verbs to operate on resources, named according to a certain URL convention (e.g. users/<userid>), when
234.
by feral 13y ago
Yes, its marketing, at least in the broad sense of the word - we think people should use contextual bandits more, and we want people to know that. That's why we're working on this problem. As I said elsewhere, we think AB-testing
235.
by feral 13y ago
Regarding the title, I've updated the title to say 'conceptually wrong' - its more of a mouthful, but maybe what we mean is clearer? I hope our intent was clear in the intro paragraph anyway.
236.
by feral 13y ago
>The regret bounds on contextual bandits are O(d sqrt(T)), where d is the dimensionality of the user profile you're using and T is time (see e.g. http://arxiv.org/abs/1209.3352 ). In contrast, the standard stoch
237.
by feral 13y ago
(Synference cofounder here) That's a good blog post. There are many ways to handle multi-variate data, with different tradeoffs. The linear model you describe in your post is one of the classic machine learning models often used. In t
238.
by feral 13y ago
I'm a cofounder of the early 2-person startup whose blog post this is. Yes, it's currently difficult and expensive to bring complex ML to bear on this problem; we are trying to make that easy and cheap. The reason people adopted
239.
by feral 13y ago
AB-testing is much better than doing nothing. But you have to be careful with any powerful tool, in case its success blinds you to its weaknesses. "When all you have is a hammer, everything starts to look like a nail". We think
240.
by feral 13y ago
(Synference cofounder here) It has successfully increased Wikipedia's donation revenue; however, some fundamental assumptions baked into the AB-testing approach are all wrong; this leaves a lot of money on the table. Instead of trying

More ›