Garry Kasparov: Open-Ended AI
00:00:00.000 | People who just look at my matches with Deep Blue 20 years ago or recent games are played
00:00:06.480 | by newer machines in Go or in popular video game Dota.
00:00:12.280 | That means a very important point that in all these cases, all these games are different.
00:00:18.160 | Go, for instance, is much more complex than chess.
00:00:19.920 | It's more abstract, more strategic, so it's more difficult to accumulate sophisticated
00:00:25.440 | knowledge as we did in chess.
00:00:28.040 | But all these games, they represent closed systems, which means we humans fill the machine
00:00:34.480 | with a target, with rules.
00:00:39.080 | There's no automatic transfer of the knowledge that machines could accumulate in closed systems
00:00:44.520 | to open-ended systems.
00:00:46.840 | Machines can do many things, and they will learn even more things in the future, but
00:00:51.560 | I don't see any chance in the foreseeable future for machines to ask the right questions.
00:00:56.720 | Now, they can ask questions, but they don't know what questions are relevant.