Artificial intelligence and common sense

Abstract

rtificial intelligence (AI) seems the latest frontier of science and engineering, a race for space in a wild territory and land of opportunity where civilisation is suspended, new heroes arise (Steven Jobs, Elon Musk, Mark Zuckerberg and others), investments flow, and new rules of the game are forged.The enthusiasm for algorithms might be medieval, but it materialised as AI since the 1950s in several hype cycles of public attention and generous capital investments.Modern machine intelligence emerged in WWII, and was formalized in machines which are no longer distinguishable from human performance as certified by the Turing or Winograd benchmark tests.Thus, since 2015 we are in the third spring of AI, after a surge during the 1980s, and the initial take-off in the 1950s.The quest to emulate human common sense is part of this history, and this ambition stills poses a challenge with few accomplishments.The engineering endeavours have bifurcated into AI for specific purposes, so-called expert systems on the one hand, and into Generalpurpose Artificial Intelligence (GPAI) on the other.No such GPAI system yet exists beyond imagination and discourse.Progress is coming from tools that search for particular tasks in which humans are competent, but AI matches or exceeds human expertise by being faster or more accurate, as for example in diagnosing medical X-ray images.This efficiency also shows in board games like chess or go, where computers spectacularly beat the world's best players.However, tools remain domain specific with some leeway for functional drift.A chess computer is hopeless at go; and a machine that reads pulmonary X-Rays does badly at diagnosing heart issues.Or most recently

Ähnliche Arbeiten