|
On March 11 2016 12:24 trulojucreathrma.com wrote:
Well, Deep Blue had a human playing.
Btw, brute force doesn't mean it is stupid. As long as you randomly iterate, I think you can call it brute forcing.
Most of the time, "brute force" does mean stupid. Bruteforcing is what you do when you revert to exploring every possibility available until you reach your stop condition, abandonning anything but the most basic algorithm and relying on raw computing power to find a solution. In a game, it would be a tree search of possible moves without cutting any branches for example.
If you tag as "brute force" any iterative analysis of future lines based on candidate moves, humans are also "brute forcing".
|
I'd say brute force is when you don't use heuristics. I.e. when you try to bruteforce a pin code and just enter all 10.000 possibilities of 4-number codes until you find the one.
As soon as heuristics come into play, it's no longer brute forcing imo, but it's up for debate. You could argue the algorithm trying out all the possible moves, and picking the best one according to some heuristics, is similar to trying all pin codes until it works.
|
On March 10 2016 05:11 {CC}StealthBlue wrote:
I would imagine the real development will be when AlphaGo knows the Human has lost before the player has made his/her move or counter.
I wouldn't be so sure. The algorithm (to me understanding) is mostly "pattern finding". Being able to determine definitively the outcome of a game given certain circumstances would likely require an extremely expensive depth first search on the remaining game states given the current one. By that I mean, the AI could one day be able to say "99% of victory" at a certain point but could likely never determine the inevitable outcome of a game from a nontrivial starting state.
To all of you crying "brute-force" :
A brute force solution would look like this: given the current board state, determine all possible next states, then all the next states for those, etc. until you have computed every possible end state, then use that to inform you as to what next state to choose.
Obviously alphago cannot do that, because when the game is not near the end go just has too many possible permutations to compute.
|
AlphaGo "knew" game 2 was won as in the late game it did some suboptimal moves just to settle the center. In go this kind of plays are often made by humans when they know they win for sure. You simplify the game to not let a chance to your opponent to pull out a tricky sequence that could reverse the result.
|
So you can only brute force when you move through all of possibility space in a totally arbitrary manner, ignoring all information about where in the possibility space the solution is most likely to be found?
I disagree. Monte Carlo randomly picks something. That's not thinking. That's relying in sheer calculation power in being able to evaluate so many positions. You force a solution through sheer calculation power.
It is like a human Chess/Go player deciding what move to make by having a hundred (trillion) million people play all his candidate moves and then playing the move that wins most often.
Brute force isn't a technical term. I can use it just fine.
Now many in game AI you have no real smart algorithms, because there are no definite ways to measure a position. You have to evaluate it. But in physical sciences, you can measure and Monte Carlo is brute force.
And in the mind of a lay person, it also is.
|
On March 11 2016 21:08 trulojucreathrma.com wrote:
So you can only brute force when you move through all of possibility space in a totally arbitrary manner, ignoring all information about where in the possibility space the solution is most likely to be found?
Yes, that is the definition of brute force. Search by exhaustion. No reasoning.
You can disagree with the semantics if you like, most people won't.
|
That's stupid to say. Also, many people don't agree. Go google "Monte Carlo" " brute force".
If only exhaustive methods are brute force, then that word loses 99.999 of it's meaning. When do you ever do an exhaustive search?
|
It is used all the time in cryptography and security. If we're gonna throw google terms around, have a look at "brute force attack"
|
They don't use it. That's why our encryption is safe.
|
Dude, it is used all the time, and what you call "safe" encryption is never safe for long periods.
Some years ago the 'DES' algorithm was a standard for symmetric encryption. Then it got brute-forced.
So Triple-DES and AES were introduced. With current computing power, they cannot be brute-forced. Yet. NIST estimates Triple-DES will be brute-forced by 2030. And of course who knows if the NSA has a supercomputer that can do it already.
Encryption is never safe.
|
DES was developed in 1970's.
|
Laurens, Truloblablah is obviously trolling you.
|
Yes, and it was considered "secure" until 1998 or so. Just like you think our current encryption is safe. Wait some years and computational power has increased to the point where our current encryption can be brute-forced, and the cycle continues.
Hence my point that brute-force is used all the time. And calling AlphaGo brute force is an insult to the team behind it.
|
On March 11 2016 21:50 Furikawari wrote:
Laurens, Truloblablah is obviously trolling you.
Oh.
|
My name was randomly generated. What you see is your own bias.
So DES was declared no longer secure before you were born. Not sure why you bring it up.
I can say that the electrical telegraph is no longer used. But then you bring up that it is, by hobbyists. It's disingenuous to bring that up considering the nature of the debate we were having.
|
I was born in 91, and in the context of our discussion it was very clear why i brought it up.
But I do believe you are trolling now, so I'll just stop responding.
|
Really? PM me when you publish your first MC paper.
User was warned for this post
|
when is the 3rd game to be played?
EDIT:
ahh its tomorrow
The matches will be held at the Four Seasons Hotel, Seoul, South Korea, starting at 1pm local time
(4am GMT; day before 11pm ET, 8pm PT) on March 9th, 10th, 12th, 13th and 15th.
|
On March 11 2016 21:50 Furikawari wrote:
Laurens, Truloblablah is obviously trolling you.
They were having a rather mature conversation until you showed up.
It's interesting how you escalated them to the point where one of them it acutally thinking he's being trolled, lol.
|
On March 11 2016 22:03 Laurens wrote:
I was born in 91, and in the context of our discussion it was very clear why i brought it up.
But I do believe you are trolling now, so I'll just stop responding.
Ahuh ahum, I am sorry for being so mature. I don't really care about your future publications on Monte Carlo anymore
No! You are a troll!! omgz
I guess you are allowed to call people trolls, but calling a Monte Carlo algorithm 'brute force' is a deep deeply cutting demeaning insult.
I like Monte Carlo. MCMC's are about the only real comp sci algorithms we use in our lab.
Google is obviously using something much more advanced. But we are a simple lab where no one even has a comp sci degree.
|
|
|
|
|
|
EPT
