觀點人工智慧

Robots must up their game to solve real problems

No, your computer is not about to take your job. But you should think twice before challenging it to a board game — even one it has not played before. Seeing humanity’s top games-players falling to the machines is nothing new. Google’s DeepMind went one better this week: its general-purpose game-playing algorithm taught itself from scratch to become a champion at three different games in a matter of hours, out-doing the best purpose-built artificial intelligence systems.

At first blush, this makes it sound like the great gulf between machine and human intelligence may soon be crossed. The creation of a generalised intelligence, as adept as the human mind at addressing whatever problem is thrown at it, has been the dream of AI since its inception.

Oren Etzioni, head of the Allen Institute for Artificial Intelligence, was among the experts to applaud DeepMind’s adaptable gaming brain, calling it an “impressive technical achievement”. Referring to the limited domain it inhabits, however, he echoed the words of Hamlet: “There are more things in heaven and earth than are dreamt of in DeepMind’s philosophy.” To put it another way: the Google subsidiary has made a name for itself by beating humans at board games, but it is important to keep things in perspective.

您已閱讀28%(1274字),剩餘72%(3240字)包含更多重要資訊,訂閱以繼續探索完整內容,並享受更多專屬服務。
版權聲明:本文版權歸FT中文網所有,未經允許任何單位或個人不得轉載,複製或以任何其他方式使用本文全部或部分,侵權必究。
設置字型大小×
最小
較小
默認
較大
最大
分享×