Building an assessment of artificial intelligence capabilities
Stuart W. Elliott
Abstract
This chapter synthesises the expert contributions of the report and offers perspectives for building a comprehensive assessment of artificial intelligence (AI) capabilities. It compares and contrasts the contributions of psychologists and computer scientists along two dimensions: whether they focus on human or AI taxonomies and tests, and whether they test isolated capabilities or more complex tasks. The chapter argues that a more complete assessment of AI must bring together the different approaches. It illustrates this argument with an example in the area of language. Finally, the chapter offers next steps towards a systematic assessment of AI capabilities, which will allow for drawing fine-grained implications for work and education.