Since the inception of the Turing Test, benchmarks for AI success have been closely tied to human capabilities. This has fueled a competitive race where ‘state-of-the-art’ achievement is defined by AI systems that outperform humans. Whether its defeating world champions in the game of GO or surpassing human performance in language tasks, these milestones are often hailed as significant triumphs in the field. However, this framework has its drawbacks. It has given rise to growing concerns about job displacement and frequently overshadows other urgent issues like algorithmic bias, misinformation, and other ethical concerns.

As AI increasingly becomes an integral part of our daily lives, it is imperative to question and redefine these traditional, narrow metrics. How can we shift the focus from evaluating AI based on its ability to replace human skills, to a paradigm where AI augments human abilities and fosters new forms of innovation? This talk aims to unpack the loaded history behind our current metrics for evaluating AI, with the goal of shifting the narrative toward a more optimistic future in which AI serves to enhance, rather than replace, human capabilities.