Despite whatever scenarios that can be speculated on in our culture’s works of fiction regarding artificial intelligence, the intelligence that emerges from this field of research may very well be utterly alien to what we might expect. It could be beneficent, malevolent, calculatingly cold, or it could even see compassion as the pinnacle of logic. Two recent developments in the field of AI research illustrate two completely diametric paths that a future artificial mind could take.
On March 23, Microsoft launched an advanced chatbot called "Tay" onto it’s own Twitter account. Tay, based on their otherwise successful Xiaoice chatbot program, was designed to mimic the linguistic patterns of a teenage American girl, and to learn from, and adapt and respond to human chatters. Unfortunately, Tay’s programmers didn’t take into account the influence of internet trolls on the otherwise naive chatbot: within hours, Tay was spouting racist, pornographic, and other offensive tweets, having absorbed too many bad habits from her less civilized chat companions. Microsoft started deleting some of Tay’s more offensive tweets, and ultimately had to resort to taking the chatbot offline, a mere 16 hours after launch.
But a more positive story comes from Google’s Deep Mind Challenge Match, illustrating the potential for AI to not only develop a deep sense of creativity, but to also prompt a similar creativity in the humans it interacts with. The Deep Challenge Match, held from March 8-15, was a tournament of the ancient Chinese game of Go between Google’s AlphaGo AI program and human Go champion Lee Sedol. Go, a deceptively simple strategy game played with small black and white stones on a 19-by-19 playing grid, can involve deeply creative strategies that are developed by it’s players.
Previous AI versus human tournaments would involve chess matches, however computers like IBM’s Deep Blue would play by simply comparing the state of the pieces on the board against a large database of moves, and then execute the set of maneuvers most likely to win. Go, on the other hand, is a strategic game, as opposed to chess’s tactical focus, meaning that it requires actual creativity in the planning of it’s moves — applying statistically-favored tactics aren’t as effective a tactic for the computer as it is in chess. AlphaGo won the tournament, astounding observers with it’s level of creativity, executing aggressive strategies that would have otherwise would not have even been considered by a human opponent: in some strategies, AlphaGo goaded Lee into well-planned traps, and in others, wisely retreated to fend off attacks from it’s human opponent.
By the fourth game in the tournament — the only match that Lee won, out of the five rounds — Lee was forced to respond to AlphaGo’s unorthodox strategies by resorting to executing a wedge move, something that is considered to be a bad move in Go, as it prevents the player from gaining territory. This fourth match illustrated one amazing aspect of AlphaGo’s AI: it brought such an unorthodox creativity to the board, that it forced it’s human Go master opponent to approach the game from an entirely different angle.