Check your alignment
On Alexander Hamilton, AI-generated homework assignments, and the need for good alignment before we do too much unintentional damage
In 1781, Alexander Hamilton gave us a beautiful line that seems to have perfectly anticipated our technology-saturated world: "Nothing is more common than for men to pass from the abuse of a good thing, to the disuse of it." We find it easy to believe the worst about new technologies because it is easy to imagine how we might abuse them in our own self-interest. But those cases are often oversold.
■ Nobody has much difficulty in imagining how generative artificial intelligence (AI) could be used to make it easier to cheat in high school and college classrooms. A technology made for the purpose of imitating human language has pretty obvious utility for doing things like writing essays.
■ The good news is that artificial intelligence doesn't appear to be increasing rates of cheating. (The bad news is that cheating was already widely self-reported long before AI came into the picture.)
■ But there are prospective dangers ahead that will undoubtedly lurk in the shadows of AI use, which is why the issue of AI alignment is so important. Requiring technology to serve human interests requires developing a lot of rules and definitions around hard questions like the classic, "What does it mean to be human?"
■ It would be a cruel irony if, while we are in the phase of "abuse of a good thing", we were to err on the side of ignorance in our approach to AI alignment, simply because too many people proved too impatient for their own good and failed to study enough of the humanities to become good technologists down the road.