Slashdot Mirror
Alpha Go Takes the Match, 3-0 (i-programmer.info)
mikejuk writes: Google's AlphaGo has won the Deep Mind Challenge, by winning the third match in a row of five against the 18-time world champion Lee Se-dol. AlphaGo is now the number three Go player in the world and this is an event that will be remembered for a long time. Most AI experts thought that it would take decades to achieve but now we know that we have been on the right track since the 1980s or earlier. AlphaGo makes use of nothing dramatically new — it learned to play Go using a deep neural network and reinforcement learning, both developments on classical AI techniques. We know now that we don't need any big new breakthroughs to get to true AI. The results of the final two games are going to be interesting but as far as AI is concerned the match really is all over.
Every time a computer beats a human at a "smart" game, we hear the same thing. And every time, when all is said and done, all we have is a program that can play a game well (and maybe a really aggressive marketing campaign to sell consulting services, Watson).
Look, we barely understand what intelligence is let alone know what it means to have a computer replicate it. We can have computers perform tasks that we ascribe to smart people and call it intelligence, but that's about it right now.
And, with deep learning and neural nets, we haven't gained any real insights into intelligence. We just have a black box mathematical function that can play a game.
Although in a sense it is "nothing new" (neural networks and monte carlo statistical techniques), the combination is one of the most convincing demonstrations of short-circuiting huge branching factors to arrive at what a human player would call "intuition".
Chess has a branching factor of about 35. This is small enough that if you prune the most dismal lines, you can brute force the rest to many ply and arrive at a very good result, but this is not a well generalizable technique.
Go has a branching factor of about 250. This cannot be brute forced, even with aggressive pruning. The result of a NN evaluator function plus Monte Carlo has been astonishing: it was not predicted for computer Go to reach this strength for decades yet, but here we are.
The implications of this combination of techniques to other kinds of problems requiring "intuition" will be interesting to watch.
Is this AI software written in Rust?
I believe it is written in C++ and Lua, because that is what the authors used in previous projects. Most of the computing is done on GPUs, which is most likely done with CUDA, because that is what they used in the past, but they could use OpenCL.
They've already stated that their next goal is to do it again without the human database, but rather through iteration. And the big final advancement *this time* was made by it playing a ton of games with itself. Studying the human games was not enough to get to this level.
And humans do the same thing. They spend their lives studying the important games that came before. So the point is it did it pretty much the same way humans do. And it has already played a move that no strong human has ever played (Game 2, move 37). At first it (not surprisingly) appeared to be blunder, until its strength became clear. Humans will now learn from the computer and their level of play will rise. It happened in chess and checkers, and in a very big way in backgammon. Any strong human backgammon player today would trounce the World Campion of 20 years ago.
Until this week you could have made a similar statement about Go.
We know now that we don't need any big new breakthroughs to get to true AI.
Grossly exaggerated claim. The following article worth reading on this subject by no one else than two of authorities in the field, one did work on the backgammon game in the 90s and the other one on the IBM Deep Blue program that win over the world chess champion Garry Kasparov in 1997. http://www.ibm.com/blogs/think... In particular:
"However, research in such “clean” game domains didn’t really address most real-life tasks that have a “messy” nature. By “messy,” we mean that, unlike board games, it may be infeasible to write down an exact specification of what happens when actions are taken, or indeed what exactly is the objective of the task. Real-world tasks typically pose additional challenges, such as ambiguous, hidden or missing data, and “non-stationarity,” meaning that the task can change unexpectedly over time. Moreover, they generally require human-level cognitive faculties, such as fluency in natural languages, common-sense reasoning and knowledge understanding, and developing a theory of the motives and thought processes of other humans."
Achille Talon
Hop!
Life could be a really complex search algorithm we just can't begin to comprehend. Life could be a non-linear approximation of 42...
Not much has changed in AI, the fundamentals on these new systems are still the same as before. The difference is we have more computing power (and CS work hours) to put into problems we thought were more amazingly complex than they were. It's likely that Go is actually as difficult as we thought it was so our progress on this search problem is not a result of a huge leap but merely a realization by some of us that the problem wasn't as big as we thought. It is akin to rating password security simply by the number of combinations and ignoring randomization: your 8 letter password might seem complex enough but in actuality it is weak because you picked a word from the dictionary. The prediction of your decision (randomness) aspect is a critical aspect.
What I'm trying to say is that the massive branching factor of Go in terms decision trees is like having long password length (like 56 letters long vs chess being 8) but the INTUITION of good decisions is akin to approaching a true random unpredictable password and it is NOT as difficult as we thought it was! Therefore, human professional intuition is not as miraculous or brilliant as previously thought. That is what this "leap" in Go is demonstrating. This is development humbling and/or frightful and should make one ponder our human hubris.
Democracy Now! - uncensored, anti-establishment news
But, this isn't 'AI', it's just another 'expert system'.
No. Alpha-Go is pretty much the opposite of an "expert system". Expert systems encode expert human knowledge in a series of explicit rules and if-then tables. Alpha-Go is based on neural nets and self-learning. There is no list of explicit rules.
We know now that we don't need any big new breakthroughs to get to true AI.
Err, no. Just... no.
systemd is Roko's Basilisk.
There's a well known phenomenon where every time some AI research produces a successful result, someone comes along and says "That's not true A.I" "It's just a computer program that has to be told what to do."
(This is the "no true Scotsman" argument.)
So let's see the list of such "non-AI" technologies:
- Natural-Language translation (getting pretty usable now)
- Speech recognition combined with ability to answer fairly arbitrary questions quite well on average.
(talking to Google via my Android phone)
- Self-driving cars (getting pretty close - will be better drivers than people on average pretty soon)
- Chess
- Jeopardy
- Go
- Detection of suspicious speech and patterns of communication (no doubt used by NSA on most Internet and phone traffic)
- Recognition of particular writer from all writers on Internet by analysis of their writing style
- Person identification by face picture recognition
- Object type and locaton type recognition from pictures
- Walking, box-stacking robot "Atlas 2"
Just algorithms.
Does it actually matter what you personally choose to call this kind of technology? It is what it is, and it's advancing quickly.
"It's not true AI" sounds like the desperate retreat cry of a person in a very defensive stance, afraid of losing a sense of human uniqueness.
Where are we going and why are we in a handbasket?
"We know now that we don't need any big new breakthroughs to get to true AI."
This is so wrong that it's hard to know where to start.
Just cruising through this digital world at 33 1/3 rpm...
So far, every post I have read that makes the same claim as yours lacks a critical piece: a clear description of what would qualify as AI.
Often, when I state that question, I get a long, rambling, disorganized list of random things humans do, and no indicator that making a computer do them would yet qualify as true AI. That is why I keep emphasizing the word "clear." Make it clear or you are just being religious.
So, exactly where are those goal-posts?
Despite the name "neural network", there is nothing "very similar" between the way AlphaGo works and brains work.
That seems correct. AlphaGo is playing go at a level beyond that of humans. The take home point seems to be that brains aren't really competitive and are probably a dead-end technology.
Game 2, move 37 was amazing. The commentator had already pointed out the issues with the two stones trapped to the lower/middle left and the loose group to the lower right. This move linked everything together in a light way. This was an "ear-reddening move."
To me, "true AI" is something that can decide to do something other than that for which it was constructed
Many people can't even decide to stop eating.
For a professional player, it's fairly easy to remember an entire game. For them, it's a story with many familiar patterns and memorable surprises.
How would this same AI model (i.e. not a retrained model) do in Chess?
Why can't it be retrained? A human Go player with no experience in chess wouldn't know the first move, but after observing some games to deduce the rules, then playing enough games to practice, they'd probably be competent.
Guess what; that's exactly how this system was trained. In fact, an earlier model taught itself to play dozens of old Atari games in the same way.
Could it hold a basic conversation? Identify a picture of a cat?
Actually, Google's context-aware voice recognition & response system is largely driven by a similar layered neutral network, as is it's visual search that can indeed identify a picture of a cat. And with a robot arm attached I wouldn't bet against the burger flipping either.
Strong AI it isn't, but neither is it a classical pre-programmed computer. "Intuitive" computing is the best description I've seen of modern neural networks - it weighs a large number of factors gained from experience and makes a decision.
Why would anyone engrave "Elbereth"?