I would love to hear from the humans how good they thought the AI play was.
To be able to complete closely with top pros seems a victory despite the defeat to me. It's surely only a matter of time and I doubt a lot of time at that.
OpenAI’s bots were knocked out of The International – the Dota 2 computer game's annual Olympics – on Thursday after they lost to human pros 2-0 in a best-of-three contest. Dota 2 is a hugely popular online battle strategy game (think Command and Conquer meets Tolkien) played by several hundred thousand people worldwide. The …
If I were to compare the bots to humans they appear to have good short term strategy, but if you remove their mechanical superiority they would lose to your average regular players.
Despite 180 years a day of learning they did not seem to have worked out the full extent of how the abilities work even on an individual level, let alone how they might synergise.
They exhibit the development of short term strategies, such as individual sacrifice which is interesting. However, perhaps unsurprisingly they did not show long term strategy. Even humans have difficulty connecting the cause and effect of there is a long time delay between the two, and trying to achieve that in bots does seem to be an interesting problem. I think that if they are capable of connecting appropriately long term events then they could produce interesting new strategies and compete with to humans.
For whatever reason associative memory has not been integrated with deep learning yet. The result is current AI only has the shortest of short term memory. Which is fine for Go because the current pattern on the board is all that matters. It is not fine when the game requires temporal strategy.
Anyway I don't think the situation for humans is so rosy. Dawkins says the watchmaker is blind. There are suggestions that evolution need not be so impaired:
Well, that is only a hobby fun viewpoint. Up to you if you buy into it or not.
Biting the hand that feeds IT © 1998–2021