The Globe and Mail reports in its Saturday, Sept. 19, edition that a key worry among the artificial intelligence cognoscenti is what will happen as AI models apply their intelligence to improving themselves. The Globe's Ian McGugan writes that this process, called RSI, or recursive self-improvement, is a loop in which AIs make themselves smarter, then use their increased intelligence to make themselves even smarter, and so on, over and over. For now, a bit of human judgment is still required to make sure the endless self-improvement process doesn't go off the rails. But researchers say that within a couple of years the process is likely to be fully automatic. At that point, the only thing restricting the abilities of AI will be access to electricity and computing resources. Godlike AIs could accelerate technological progress and revolutionize medicine, ushering in a golden era. However, they could also disrupt economies, lead to widespread job loss, concentrate wealth among a few major players and result in a permanent loss of security, as a superintelligent AI could hack any system. If there's a 5- to 10-per-cent chance that AI could cause catastrophic loss of life, as many experts suggest, we must regulate it.