After Go, Developers Are Now Building AI To Beat Us at Soccer (cnet.com)

After Google's AlphaGo artificial intelligence bested our best Go player, South Korea is now setting its sights on making AI that can play soccer. From a report: Hosted by the Korea Advanced Institute of Science & Technology (KAIST), the AI World Cup will see university students across South Korea developing AI programs to compete in a series of online games, reported The Korea Times. The prelims will begin in November. "The football matches will be conducted in a five on five tournament," a KAIST spokesperson told the publication on Tuesday. "Each of the five AI-programmed players in such positions as striker, defender and goalkeeper will compete with their counterparts."

Comments

Re:Not True AI

[Archangel Michael]

Emacs is closer to being AI than Vim will ever be. ;)

Re:Not True AI

[Kjella]

If you have to "develop" AI for a specific task (play games) or whatever then in my view it's not AI. I think AI should learn to do what ever task you throw at it.

They're working on that but trying to create an intuition of what's a stepping stone in the right direction has proven hard. It's mostly not how humans learn either. Even if you took someone off the street that's never played football, they've probably seen football. Or they got some basic idea of how it could be played based on analogies from other games. The totally blind approach would be like handing a tribe of Amazon Indians that's been in no contact with civilization the rule book and ask them to figure out how to play. So we're training the AI, but it's not anything like coaching a team. It's more like an armchair quarterback training, here's how a bunch of teams have played football. It doesn't even have to be the best teams, it's more about pruning the near infinite space of everybody doing everything to things that "makes sense".

Then the AI starts doing variations on moves, counter movies, counter-counter moves and so on and refine it. Maybe it's not so glamorous for sci-fi, but at least for automation we're starting to see AIs that can take rather "fuzzy" tasks, look at what a bunch of humans are doing to solve it and start doing it to OCD-levels of perfection at the speed of a computer. That's a pretty big deal for a lot of trades where you essentially apply variations of a skill but where the particularities of the situation has kept it from being automated like an assembly line. As in, I think it will enable AIs to automate a lot of things people don't really think can be automated. And if you assemble lots of these little AIs you'll get more automated processes than looking at one in isolation.