To Test a Powerful Computer, Play an Ancient Game (1997) | Hacker News Reader

To Test a Powerful Computer, Play an Ancient Game (1997)

nytimes.com

The sole quote responsible for the reworded title is:

''It may be a hundred years before a computer beats humans at Go -- maybe even longer,'' said Dr. Piet Hut, an astrophysicist at the Institute for Advanced Study in Princeton, N.J., and a fan of the game. ''If a reasonably intelligent person learned to play Go, in a few months he could beat all existing computer programs. You don't have to be a Kasparov.''

There is also: "The offer [of a price for beating go champs with a computer] expires in the year 2000. Go programmers are hoping it will be extended for another century or two."

Edit: I think the reworded title reflected very well the spirit of that article as a whole, before it was reverted.

Right—it's against the HN rules to rewrite a title to single out just one detail. That's a form (actually the leading form) of editorializing.

Submitters: if you want to point out what you think is important about an article, the place to do that is in the thread, on a level field with other participants.

Sorry for that, it was unashamedly lifted from another and more generalist site, title and all. Won't happen again.

No worries!

When it's not a news article, it seems perfectly fair that someone should be able to submit a particular paragraph for conversation, rather than being forced to generically link 'the original source'.

There's no necessity to have a level playing field when there is no playing field. There's generally no one else submitting an article if it's 20 years old.

Actually, given the inspiration source where I got this article it was very likely someone else would post. And someone did: https://news.ycombinator.com/item?id=11324885

Coincidentally with the same title, as that was deliberately crafted to attract attention that very same statement (given the fact that it didn't stand the test of time).

The level playing field I'm talking about isn't between multiple submitters, but between the submitter and other users. On HN, submitting a story confers no special rights to frame the discussion for everyone else. Readers here can make up their own minds about what's important in a story. For that it's important (critical, actually) to have accurate, neutral titles that avoid spin.

> The level playing field I'm talking about isn't between multiple submitters, but between the submitter and other users. On HN, submitting a story confers no special rights to frame the discussion for everyone else.

You can't avoid framing. The act of submitting an article frames a discussion around its contents. If you came up with the submission idea by yourself, you are unavoidably privileged in deciding what the discussion is about.

If there is a collection of 200-word pages on a site, and I submit one of them, nobody will accuse me of editorializing. But if I try to submit a 200-word segment of a large page, it's a problem. Something's off there. The url structure of a site should not be the determining factor in whether editorializing is happening.

> Readers here can make up their own minds about what's important in a story. For that it's important (critical, actually) to have accurate, neutral titles that avoid spin.

In many cases, including this one, the story is the important facts. The story here is not "computer played go in 1997". The story is the prediction. Avoiding spin is great, but a title that doesn't mention the story is a bad title.

There's a really simple cure for this - write your own piece pointing out and commenting on what you think is the interesting bit and submit that, with the title of your choosing. Then you absolutely get to frame, comment, editorialize. You just don't get to do it in the submission titles.

Ah but that doesn't work, because it's not the 'original source', and people will replace your link entirely.

No they won't, if you write something that is of interest as the sites rules ask you. This touches on the more fundamental problem is that a tiny snippet from an old article is likely not of great intellectual interest as the comments in this thread show.

If you're just focusing on a particular fact reported elsewhere, you're not writing something of interest.

>This touches on the more fundamental problem is that a tiny snippet from an old article is likely not of great intellectual interest

If there had been slightly more written about the justification it probably would have been very interesting to dissect in the comments, despite being a mere snippet of an old article.

Predictions seem to be consistently wrong, or at least random. We don't have flying cars. My mother was told in the 30's that there would be movies in the home in 80's. She really hoped her eyesight would be good enough to see the TV at that age.

I think if anyone would actually be good at predictions they could somehow take advantage of that. Maybe in the stock market?

Think about weather forecasts. If you want to know if it'll rain on this day next year, you can look at historical data, and make a guess. If you want to know if it'll rain tomorrow, the forecast is really good.

Technically, predictions are just made up, forecasts have some model behind them. but the notion of near vs far is helpful. Back then, there really wasn't any hope of beating a human at go. We were in a position of not knowing what we don't know. As we get better at it, the upper and lower bounds can be reevaluated.

A nice comparison is self driving cars. 20 years ago, the answer would be a long time, without really complicated road infrastructure. Like magnets embedded in highways. But once the DARPA grand challenge fell, we were able to revise the time down dramatically. Alphabet is saying 30 years for perfect in all conditions. But as we understand and revise the problem, we can get pretty good highway drivers right now. They'll steadily improve handling more and more adverse situations. so the predictions can be revised.

Far flung predictions are cultural aspirations. As we understand, we improve or abandon that prediction.

> We don't have flying cars.

There are flying cars - they are called "helicopters": http://xkcd.com/1623/

Not sure if Tezhi Luzhanqi or L'attaque (1910) qualify as ancient games, but let me play a heuristically tuned AlphaStratego mesosphere and I'll beat it 5-0, now and for the next 5 years at least! Jack6/Shark bridge AI doesn't consistently beat the top human pairs yet. Then there are Connect6 and Emergo.

At the same time there were predictions that computers would have more processing power than humans within 1 or 2 decades.

Those two things obviously don't go together.

Seems consistent to me. Just because you have hardware doesn't mean you know how to use it well. Consider everyone accidentally writing an O(n^2) algorithm on their multi-gigahertz/gigabyte/terabyte (CPU/RAM/HDD) machines. At the time, there was no reasonable algorithm which even on the top supercomputer could hope to beat the world champion. Then MCTS came along a few years later, and at that point maybe you could've scraped together enough hardware to beat him; then another few years, and AlphaGo... We are in the same position now for AI: the bottleneck is not the hardware but knowing what to run. (Some technical examples of this: model compression, binary networks, and residual networks, show that what we can run far outstrips what we're able to train.)

It's not clear we have enough hardware. I've been trying to get a handle on the processing power needed for human level AI. If you take the naive approach and assume it's just a matter of getting models with the same number of parameters as the brain has synapses, our biggest models are about a million times too small. If you assume Moore's law will remain dead, then it may be a long while. Then again if modern deep learning is amenable to some sort of ASIC it could happen much quicker than one would expect, depending on how much investment is put into such things. And who knows how algorithmic advances will change things. If we ever get quantum computers, they're likely to be very useful in the design of conventional computers and materials technology in general, so another age of exponential advancement isn't out of the question.

It's very, very hard to be certain about these things.

As I've come to understand it when implementing deep learning machine learning the hardware requirement for running a trained neural network model efficiently is much lower than the one required to train it. I've also read somewhere that if an instance of AlphaGo run on a modern powerful desktop machine would compete with the cluster used in Lee See-dol match which consisted of something close to 2000 of CPUs and more than 100 gpus then the ratio in wins would be only be 3:1 in favor of the cluster. Much less than if you would compare it to the ratio of computational power so that gives you hint of how much of the intelligence in it is about previous knowledge gained vs brute computing power. To me that also implies that a lot of the innovation done at DeepMind is more about improving the speed of which a neural net learns and aquires knowledge and less about the running time of applying the model. And that's probably an important distinction in the discussion of this technology.

> If you take the naive approach and assume it's just a matter of getting models with the same number of parameters as the brain has synapses, our biggest models are about a million times too small.

And if you had that hardware, what would you run on it, exactly? There is no off the shelf algorithm which we could plop on that futuristic GPU and say, yes, this will definitely yield a general human-level intelligence. That is my point. We may or may not have the necessary hardware, but we sure don't have the necessary software right now.

I agree. But I often here the claim that we have enough hardware and now it's just a software problem. I don't think that's at all clear.

Maybe it would have been more insightful to ask Game engine engineers and AI researchers than an astrophysicist.

In 1997 the most powerful computers were perhaps being used by astrophysicists?

AI researchers are not any better at predictions. There was a time they thought true AI was just a matter of enough hardware, so they extrapolated hardware growth speed.

Obviously they were wrong: For true AI we need new software too - all the hardware in the world would not accomplish it.

Reply on news.ycombinator.com