What should we learn from past AI forecasts?

Abstract

Early research on artificial intelligence (AI) was marred by overly optimistic predictions about its potential impact, leading to periodic cycles of hype followed by disappointment. An analysis of historical AI forecasts, expert commentary, and relevant literature reveals several contributing factors. The initial enthusiasm stemmed from notable early successes in specific AI domains, coupled with an underestimation of the challenges involved in scaling these successes to more general tasks. Additionally, limited understanding of human cognition and the complexity of natural language hindered progress. Over time, as AI researchers gained more experience, a more realistic assessment of the field’s capabilities emerged, leading to a tempering of expectations. – AI-generated abstract.