If machines can beat us at games, does it make them more ?

[Leto]

Isn't it bizarre that the article claims their computer searched for a week and it didn't reach a depth of 20 ply on that problem? To me that sounds like something you'd expect from a computer from the 60's. Any computer today should have no problem reaching 20 ply in seconds. A recent dev version of Stockfish finds the mate in less than a second.

I believe the article author meant reaching 20 ply by brute-force. That is, compute all the possible moves.

He then compared the difficulty (> one week computation) with human intuition.

That was why I shared the article, because I think the author (who was a PHD in artificial intelligence) didn't understand what he was writing, something that is supposed to be his research area.

Other troubling lines I found are "Most chess programs operate via brute-force search, which means they look through as many future positions as they can before before making a choice." and "Most computer chess programs won’t see the winning strategy. Instead, they will move the white king to the centre of the board which is the common strategy when there are only a few pieces on the board."

Correct me if I'm wrong but most chess programs don't use brute force, even Deep Blue wasn't brute force. Selective search techniques have been used for many decades. Seems to me the opposite of his statement is true, most computer chess programs today would find the winning strategy rather quickly.

I agree, I think the author really doesn't know what he's talking about when it comes to computer chess.

[Lyudmil Tsvetkov]

even without extensive search, but only some goodish eval, an engine should easily see in a couple of plies that white wins, as objectively the a4 square, where the king should head is at least of the same value as the central e3 square.

I would actually say that centralisation in the eg is much less important than reaching advanced ranks, many engines probably still follow the simple centralisation rule, because some engine started it 20-30 years ago, it proved to be some worth, and later few people cared to improve on it.

with sufficiently refined eval, a single ply is more than sufficient to pick up the best move.

you can even do it simply based on move ordering, with approximate scores, this will still find the best move, though will not return precise eval.

[syzygy]

I guess that if we keep going on the way we are, that computers would get the upper hand and take over.

We are no closer to actual machine intelligence today than in the 1950s... We can only calculate a lot faster and process huge amounts of data.

Is calculating even faster and processing even larger amounts of data going to be sufficient for machines to take over the world? I doubt it.

On the other hand, I do not know in what way human intelligence differs from mere calculation and data processing.

[syzygy]

Correct me if I'm wrong but most chess programs don't use brute force, even Deep Blue wasn't brute force. Selective search techniques have been used for many decades.

Define brute force...

Deep Blue was full width plus selective extensions. If we equate brute force with Shannon's Type A Strategy, then Deep Blue was brute force.

Of course Deep Blue did use a transposition table, so it too would have solved the test position in a fraction of a second.

Even modern chess engines are mostly intended to be brute force in the sense that they (eventually) enumerate all possible solution candidates. They are just a lot better at looking at more promising lines earlier than the engines of the past.

[nionita]

Beside this, we do not really understand what processes in the human brain are responsible for example for the spatial "orientation" which makes possible to see "immediately" that the white king can go through A4 to victory.

Maybe that processes are a lot more "brute force" than searching a few millions nodes, but we do not feel it like this, as this looks "natural" for us.

We and the machines are leaving in different function families, in respect to what our "hardware" can do natively, and this makes solving "easy" tasks from the other world very difficult.

[duncan]

At some point in the next two decades, the computers will surpass us in full on brain horsepower.

if this link is true then in at least in the field of lung cancer diagnosis it has surpassed humankind.

http://www.wired.co.uk/article/ibm-wats ... cal-doctor

According to Sloan-Kettering, only around 20 percent of the knowledge that human doctors use when diagnosing patients and deciding on treatments relies on trial-based evidence. It would take at least 160 hours of reading a week just to keep up with new medical knowledge as it's published, let alone consider its relevance or apply it practically. Watson's ability to absorb this information faster than any human should, in theory, fix a flaw in the current healthcare model. Wellpoint's Samuel Nessbaum has claimed that, in tests, Watson's successful diagnosis rate for lung cancer is 90 percent, compared to 50 percent for human doctors.

I guess that if we keep going on the way we are, that computers would get the upper hand and take over.

how would the computer take over if you can just unplug it

The military seems to have overlooked Asimov's robot rules and are already building fully automated killing systems that can decide for themselves about targets.

I think that even if such systems exist, no decision so far has been taken to use them.
https://www.washingtonpost.com/national ... 52eb63d8aa

Autonomy is really the Achilles’ heel of robotics,” said Johann Borenstein, head of the Mobile Robotics Lab at the University of Michigan. “There is a lot of work being done, and still we haven’t gotten to a point where the smallest amount of autonomy is being used in the military field. All robots in the military are remote-controlled. How does that sit with the fact that autonomy has been worked on at universities and companies for well over 20 years?”

[hgm]

A good static evaluation can indeed also recognize this position as an easy win, and guide the play towards the win. They hardly come any easier than a clean 6th-rank protected passer ahead.

But of course one could counter this argument by saying that such a static eval would indeed be intelligent, while fixed-depth search without TT is stupid.

[Dann Corbit]

I guess that if we keep going on the way we are, that computers would get the upper hand and take over.

how would the computer take over if you can just unplug it

How can you unplug it, if it shoots you first?
https://www.nytimes.com/2016/10/26/us/p ... .html?_r=0
https://en.wikipedia.org/wiki/Lethal_autonomous_weapon
http://www.dailymail.co.uk/sciencetech/ ... orism.html
http://yournewswire.com/russia-to-deplo ... -in-syria/

[RJN]

A 100-Drone Swarm, Dropped from Jets, Plans Its Own Moves

https://www.technologyreview.com/s/6033 ... own-moves/


PS: for some fun, watch the end:

https://www.youtube.com/watch?v=njos57IJf-0

[Dann Corbit]

Asimo's older brother:
http://www.maxim.com/news/atlas-google-robot-2016-2