Research: AI

BACKGROUND:

AI is fast, efficient, and in many ways even lives up to the second half of its name: intelligent. Some experts predict that not too many years hence, AI will reach the holy grail of general AI; that is, it will no longer be limited to specific tasks, such as protein folding or text translation, but will be able to do pretty much anything human intelligence can do, a hypothetical moment often referred to as “the singularity.” At that point, the possibilities of AI would be limitless, with the potential to help humans in ways we haven’t even imagined yet, says Max Tegmark, physicist and machine-learning researcher at MIT. But the disadvantages of AI are being increasingly documented as well. AI can be horribly biased, as when one AI system predicted that black defendants were far more likely to be repeat offenders than they actually were, or when it astonished its creators by labeling photos of Black people as “gorillas.” AI’s mistakes can be deadly. A self-driving car hit and killed a pedestrian in Tucson in 2018. But perhaps most disturbingly, in many cases, AI is a black box — we don’t know what’s going on inside. — The Pros and Cons of Artificial Intelligence (Discover Magazine, September 2, 2022)