Can an AI manager ever play like a human?

Good article in The Economist about a competition called the 2KBotPrize where contestants attempt to programme a computer 'bot' (for games like Quake, Call of Duty, etc) and "trick human judges into thinking they are playing against other people". The winner is the person whose bot tricks the most judges.

It's a twist on the original Turing Test, where Alan Turing said that if a machine could make someone think that they were having a typed conversation with another human, it could be said to have 'artificial intelligence'.

It made me wonder whether, in some future version of the game, the AI managers in FM could ever pass the test (or whether that should actually be the aim). I've only occassionally played network games or FML, so playing human managers isn't something I've done too much of.

Anyway, here's the article:

IF A computer could fool a person into thinking that he were interacting with another person rather than a machine, then it could be classified as having artificial intelligence. That, at least, was the test proposed in 1950 by Alan Turing, a British mathematician. Turing envisaged a typed exchange between machine and person, so that a genuine conversation could happen without the much harder problem of voice emulation having to be addressed.

More recently, the abilities of computers to play games such as chess, go and bridge has been regarded as a form of artificial intelligence. But the latest effort to use machines to emulate the way people interact with one another focuses neither on natural languages nor traditional board and card games. Rather, it concentrates on that icon of modernity, the shoot-’em-up computer game.

At a symposium on computational intelligence and games organised in Milan this week by America’s Institute of Electrical and Electronics Engineers, researchers are taking part in a competition called the 2K BotPrize. The aim is to trick human judges into thinking they are playing against other people in such a game. The judges will be pitted against both human players and “bots” over the course of several battles, with the winner or winners being any bot that convinces at least four of the five judges involved that they are fighting a human combatant. Last year, when the 2K BotPrize event was held for the first time, only one bot fooled any judges at all as to its true identity—and even then only two of them fell for it.

Computers can, of course, be programmed to shoot as quickly and accurately as you like. To err, however, is human, so too much accuracy does tend to give the game away. According to Chris Pelling, a student at the Australian National University in Canberra who was one of last year’s finalists and will compete again this year, a successeful bot must be smart enough to navigate the three-dimensional environment of the game, avoid obstacles, recognise the enemy, choose appropriate weapons and engage its quarry. But it must also have enough flaws to make it appear human. As Jeremy Cothran, a software developer from Columbia, South Carolina, who is another veteran of last year’s competition, puts it, “it is kind of like artificial stupidity”.

Mr Pelling says that one of the biggest challenges lies in programming the bots to account for sneaky tactics from the judges. It is relatively easy to manipulate the game and do unnatural things in order to elicit behavioural flaws in a badly programmed bot. And if a judge observes even a single instance of unnatural behaviour the game is, as it were, over.

Even if a bot does eventually fool the judges, though, would it really mark a significant advance in artificial intelligence? One of their number, David Fogel, the chairman of Natural Selection, a software development firm based in San Diego, California, thinks not. As he observes, once Garry Kasparov had been defeated by Deep Blue, the significance of beating the best human chess player suddenly seemed less important.

Someday, yes. But that'll still be a long time.

Interesting article RT--!

Like one of the quotes from the story mentions, this changes the Turing Test from mimicking the intelligence and flexibility of a man in conversation to copying the individual's tics and failures in a virtual world.

AI managers 'playing intelligence' is directly connected with the fun factor in the game imo. If the AI managers are too good (and I don't mean unfairly advantaged through large transfer budgets etc.) a lot of players will get disheartened and will play less FM. If AI managers are 'too stupid' and easy to play against the same thing will happened.

The AI needs to pose a challenge to a player without making the game impossible. This is really no easy concept or idea to implement into FM as human players of differing skill levels will require different levels of opponent AI for them to fully enjoy the game.

So what do we need? Dynamic AI!

Atm though, I am sufficiently challenged by the AI in FM, the main thing that hinders my enjoyment of the game is that I have the "load & reload (if you don't win)" syndrome... So with FM10 I am going to try and take the results as they come, without reloading.

Routines to prevent humans to exploit the game have already been developed and used since a while, otherwise the game would be exploitable with a winning mixture of sets.

This is enough to give the AI enough ways to prevent the human to use its brain to exploit but only to try to exploit

That's the " addictive " factor of the FM series