How long until human-level AI? Results from an expert assessment

摘要

The development of human-level Al has been a core goal of the AI field since its inception, though at present it occupies only a fraction of the field's efforts. To help understand the viability of this goal, this article presents an assessment of expert opinions regarding human-level AI research conducted at AGI-09, a conference for this AI specialty. We found that various experts strongly disagree with each other on certain matters, such as timing and ordering of key milestones. However, we did find that most experts expect human-level AI to be reached within upcoming decades, and all experts give at least some chance that some milestones will be reached within this time. Furthermore, a majority of experts surveyed favor an integrative approach to human-level AI rather than an approach centered on a particular technique. Finally, experts are skeptical about the impact of massive research funding, especially if it is concentrated in relatively few approaches. These results suggest that the possibility of achieving human-level AI in the near term should be given serious consideration.