I can't believe it's been 8 years.
8 years later: A world Go champion's reflections on AlphaGo
71–80 of 88 posts
Re: 8 years later: A world Go champion's reflections on AlphaGo
#72Earlier quoted context omitted.
> “Driving” is solved. Driving with humans on the road - doing unpredictable human things - is far off still. Plus there's serious questions about liability with self driving cars which are still unresolved in most of the world - if the goal is to have vehicles operate themselves with no human supervision, who goes to jail when they kill someone? Despite all of the progress that's been made with AI it's mostly been i…
The answer almost certainly has to be the manufacturer. I'm sure not responsible if my properly maintained and used self-driving car kills someone. That said, it's a novel area that doesn't have a clear analog to other products today.
There's also the question of incident response, if a human driver "malfunctions" you take them out of service and the rest of the world keeps going, but if a self-driving model malfunctions there are potentially millions of vehicles running the same software ready to make exactly the same mistake until the issue is isolated and fixed. Should we ground the entire fleet of vehicles running that software until the issue is resolved and software re-certified, if the software is demonstrably dangerous? How much would that cost?
Re: 8 years later: A world Go champion's reflections on AlphaGo
#73Earlier quoted context omitted.
Complexity of the game has nothing to do with the number of legal positions. It's very easy to design a game with arbitrary number of positions which is very simple. While go might be more complex than chess using a more reasonable measure this argument was used to for arguing nonsense in scientific papers in the past (that some poker games are more complicated than chess because they have more possible states).
Are you sure that the complexity of a game has absolutely nothing to do with the number of legal positions? I mean I am open to hear the justification for this, but I was fairly certain that all measures of game complexity are a function of the number of legal positions. Now certainly there are other factors, namely the cost of computing the transition from one legal move to another legal move so a simple game might…
Take a game where you get to pick a single number between one and a billion. If you pick 10 you win. This has a billion states, but it's trivial. I can increase the bound above a billion, it doesn't matter.
State count gives an upper bound, though, to how complex a game can be, for sure.
Re: 8 years later: A world Go champion's reflections on AlphaGo
#74Earlier quoted context omitted.
Beyond rule complexity, there are at least 5 measures of game complexity in Combinatorial game theory [1]: State-space complexity Game tree size Decision complexity Game-tree complexity Computational complexity [1] https://en.wikipedia.org/wiki/Game_complexity
Those are all meaningful measures of complexity, but it's worth noting that all of them are a function of the number of legal positions (among other things as well).
That would be quite a strange function. For example, Tic-Tac-Toe has 26830 possible games on 5478 possible positions, while 2x2 Go has 386356909593 possible games on only 57 possible positions.
The major difference of course being that Go allows stones to be captured.
Re: 8 years later: A world Go champion's reflections on AlphaGo
#75> Go is a deeply complex strategic game — famously far more complicated than chess, with 1,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000 possible board configurations. The correct number of legal Go positions is over twice as much, or to…
All these digits are only making it more obfuscated. Using the order of magnitude it's 10^44 for chess versus 10^170 for Go. Thus, Go is 10^126 times more complex than chess. For reference, the estimated number of individual atoms in the universe is thought to be between mere 10^80 and 10^83.
> For reference, the estimated number of individual atoms in the universe is thought to be between mere 10^80 and 10^83.
Yes, but what are the estimated number of states of all these atoms?
Re: 8 years later: A world Go champion's reflections on AlphaGo
#76Earlier quoted context omitted.
Those are all meaningful measures of complexity, but it's worth noting that all of them are a function of the number of legal positions (among other things as well).
That would be quite a strange function. For example, Tic-Tac-Toe has 26830 possible games on 5478 possible positions, while 2x2 Go has 386356909593 possible games on only 57 possible positions. The major difference of course being that Go allows stones to be captured.
I don't see what's strange about that, different functions grow at different rates. That's like saying it's strange that exp(x) is a billion for x = 21 while sqrt(x) is merely ~4 when x = 21.
Go's complexity grows much faster than tic tac toe's complexity as a function of legal positions, but the complexity is still a function of the number of legal positions (among other things, as I pointed out).
Re: 8 years later: A world Go champion's reflections on AlphaGo
#77Google's documentary on AlphaGo https://www.youtube.com/watch?v=WXuK6gekU1Y Truly a must watch! (just look at the video comments to be convinced)
+1
What stuck with me is Lee Sedol's strong emotional reaction, leading him to leave professional Go playing.
It's understandable he didn't expect AlphaGo to be that strong. Or that (for him) losing to a machine took the 'soul' out of the game.
But come on... I've been cornered by Pac-Man ghosts many times. That doesn't make Pac-Man less fun to play.
Nor does losing to the crude 'AI' steering those ghosts. Instead, you play, aim for a high score, see how long you can survive, how many levels you can complete, or how many fruits & ghosts you can eat in a game.
And (if you care) compare how those 'metrics' stack up against other players.
If a machine with superhuman Go-playing ability isn't fun or challenging, then stick to human opponents.
Of course it's his views and choices, and I respect that. But other than providing extremely challenging opponent, I don't see how human-beating machine would take the fun out of a game. Rather the opposite: new tactics, new insights, a raised upper bound for a Go player's strength (human or otherwise), etc.
Re: 8 years later: A world Go champion's reflections on AlphaGo
#78Earlier quoted context omitted.
All these digits are only making it more obfuscated. Using the order of magnitude it's 10^44 for chess versus 10^170 for Go. Thus, Go is 10^126 times more complex than chess. For reference, the estimated number of individual atoms in the universe is thought to be between mere 10^80 and 10^83.
> For reference, the estimated number of individual atoms in the universe is thought to be between mere 10^80 and 10^83. Yes, but what are the estimated number of states of all these atoms?
is it even countable?
Re: 8 years later: A world Go champion's reflections on AlphaGo
#79Earlier quoted context omitted.
“Driving” is solved. Driving with humans on the road - doing unpredictable human things - is far off still. My guess is industrial and home robotics will solve a lot of the “doing things around humans” problems in the next ten years. Why the hell people decided to automate giant death machines before perfecting small things never made sense to me.
Fwiw, I work in home robotics, but have no experience in self driving. My halfway-naive belief is that self-driving is easier than getting useful home robots —in fact I feel it’s not even a close comparison. Some reasons: - The home is a very unstructured environment, whereas roads have at least _some structure_, and perhaps ~70% of the most useful roads even have clear lane markings and other signs. - People already…
>to do something useful in the home, imo you need to solve navigation _and_ complicated manipulation problems. For self driving, you only need to solve the navigation problem.
Right. It can be challenging to figure out how fast, what lane, should I brake, etc. in many cities. But there are really only a few things the car can control. And its objectives are pretty simple: Obey the law, don't hit anything (and avoid being hit), and get to point B.
By contrast, think of all the different types of manipulation you need to clean up around the house and the 100 judgements you make you decide what needs to be cleaned--which will vary by person.
Re: 8 years later: A world Go champion's reflections on AlphaGo
#80Earlier quoted context omitted.
Yes this is my point - the home is a hard place to operate in but less potential for lethal outcomes. If we can solve home robotics I think cars would be easier. Also, a robot that replaces a housekeeper would have a huge market. I’d pay a handsome sum to have perfectly cleaned kitchen and bathrooms every day when I wake up.
For clarity, I’ll call out the areas where I think we disagree: > “the home … [has] less potential for lethal outcomes.” I don’t think this is true. Roads already have systems in place to make them safer, and people are aware of the dangers. This isn’t the case at home, and useful home robots certainly have the ability to cause serious injuries/deaths > “If we can solve home robotics I think cars would be easier” I a…
My main thing with road safety is the presence of giant dangerous SUV which one has no control over. At least I can control what is or isn’t in my home, on the roads some asshole driving their Cybertruck at 40 mph over the limit will annihilate my hatchback. Point taken regardless, but I still worry more about cars than anything in my home.
Otherwise I have a small child in the house, so I’d be grateful for 1 percentile capability at the moment. ;-)
Thanks for your thoughts tho, I think we can agree future seems interesting at the least.