4 ms·
Garry Kasparov: Learning to Love Intelligent Machines
- pmoriarty 10y agoPaywalled.
- wslh 10y agoIt is time HN to have an arrangement with these paywalled sites ;-)
- mckoss 10y agoOr just ban the links from HN. Paywall articles have no business showing up on a site meant to engage in community discussion. There are plenty of other things for us to share.
- rory096 10y agohttps://t.co/qtzhpiPofm https://t.co/qtzhpiPofm
- asdfologist 10y agoWhoa how does that work?
- taesis 10y agoThey 'closed' the Google referer header loophole by switching to Facebook and Twitter headers. That is a shortened Twitter link, thus you see the full article.
- petters 10y agoI'm glad to see Kasparov's views on AI are now reasonable. Eight years or so ago he had the opinion that poker AI would always be impossible (because of bluffing, I think).
- vanderZwan 10y agoI think you're mixing up two different things though: one is the question about whether AI is good or bad for humanity, the other about what the limits of its capabilities are
- petters 10y agoYou may be right. It's good to see some reasonable views at least.
- gspetr 10y agoProfessional Go players also thought Go was impenetrable to AI because of a vastly larger search space than chess. They too were humbled by AlphaGo last year.
- chubot 10y agoWow! Deep Blue was 20 years ago. Time flies. If anything, it feels like people thought faster progress on AI would have been made over the last two decades. Deep learning is huge now, and will no doubt get bigger, but right now it feels limited to a few application areas and a few companies. It seems like the expansion of the internet into mobile computing and our social lives were really the big story about tech in the last two decades. Google started in 1998; iPhone was released in 2007. Kasparov makes a good point: humans and computers are good at different things. I expect that to remain true for awhile. We are seeing superhuman performance on tasks that are relatively well-defined. I think there will be a long tail of judgement and common sense that will lag behind the wins in data-driven recognition. And that is why human-computer systems will beat computer-only systems in almost all interesting applications.