Toward a New Science of Common Sense

Common sense has always been of interest in AI, but has rarely taken center stage. Despite its mention in one of John McCarthy's earliest papers and years of work by dedicated researchers, arguably no AI system with a serious amount of general common sense has ever emerged. Why is that? What's missing? Examples of AI systems' failures of common sense abound, and they point to AI's frequent focus on expertise as the cause. Those attempting to break the resulting brittleness barrier, even in the context of modern deep learning, have tended to invest their energy in large numbers of small bits of commonsense knowledge. While important, all the commonsense knowledge fragments in the world don't add up to a system that actually demonstrates common sense in a human-like way. We advocate examining common sense from a broader perspective than in the past. Common sense should be considered in the context of a full cognitive system with history, goals, desires, and drives, not just in isolated circumscribed examples. A fresh look is needed: common sense is worthy of its own dedicated scientific exploration. The Common Sense Gap The modern-era data-intensive machine learning juggernaut continues to roll on, with a wide array of extraordinary results and significant commercial impact. But an increasing number of articles and books (for instance Pavlus 2020; Marcus and Davis 2019) point out that even the best of current AI falls short of the robust, general intelligence envisioned by the field's founders. Blunders made by generally powerful systems have been recounted, such as shocking misidentifications of objects by otherwise accurate image recognition programs (Szegedy et al. 2014; Mitchell 2019) . Surprising gaffes of seemingly remarkable systems like GPT (Vincent 2020) have been revealed as both humorous and disturbing (Marcus and Davis 2020). Self-driving cars make terrifying unexplainable mistakes (Hogan 2021). Several authors (see, for example Levesque 2017; Marcus and Davis 2019; Mitchell 2019; Toews 2020) have made the case that AI is still missing something critical to avoiding these mistakes, and they identify the missing ingredient as what we