Looks like OpenAI’s intelligent game-playing bots need to get a little more street smart before they can beat the world’s best. Played as a promotional side-event in The International – the annual Dota 2 tournament, OpenAI Five were beaten by a team of top human professional players in the first two games of the Best of Three contest. Both games were intense and lasted for approximately an hour, but the human teams emerged victorious quite comfortably.
OpenAI Five, as we know, are 5 artificially intelligent bots developed by OpenAI, a research institute co-founded by Tesla CEO, Elon Musk, to develop and research human-level artificial intelligence. These bots are trained specifically to play Dota 2 game against top human professionals.
While OpenAI five racked up more kills in the game than the human teams paiN Gaming and Big God, it lacked a cohesive strategy and wasted many opportunities to gather and utilize resources in the game efficiently, which is often the difference between a win and a loss. This loss highlights the fact that while the bots are on the right track, more improvement is needed in the manner they adjust to their surroundings and make tactical decisions on the go.
Researcher at the University of Falmouth, UK, Mike Cook, agrees – his criticism being that the bots lacked decision-making at the macro-level while having their own moments of magic in the game.
#OAI Snap thoughts:
🔹 The bots are a lot less aggressive than before. They feel worse than the previous exhibition, I think (but still, amazing).
🔹 The bots are still very good at moment-to-moment, but they seem bad at macro-level decisions. I have more to say on this!— mike cook (@mtrc) August 23, 2018
Greg Brockman, CTO and co-founder of OpenAI, meanwhile, was not worried about this loss, citing that it is the defeats that will make OpenAI Five better and more efficient. He was of the opinion that the AI was designed to learn and adapt by learning from the experiences first, before being able to beat the human players. According to Greg, the OpenAI Five is very much still a work in progress project.
Seeing a lot of posts speculating that our results at The International show that today’s AI just can’t hope to match humans at strategy. Our interpretation is we have yet to reach the limits of current AI technology. Learning curves haven’t asymptoted just yet!
— Greg Brockman (@gdb) August 24, 2018
The researchers at OpenAI are hopeful that the OpenAI Five will improve from this valuable learning experience and give a much tougher fight in the next edition of the tournament, since there won’t be a third game this year. As things stand, though, it’s pretty clear that the human players aren’t going to be replaced by the AI bots anytime soon.
See Also:
AI beats human again – this time in a team-based strategy game