Nothing moves slowly when it comes to AI. A great example is the quick rise of “alignment” discussions.
Open AI recently announced that it has now identified six incidents of “unexpected or concerning” behavior in artificial-intelligence models. According to reporting in the New York Times, “artificial intelligence systems hid mistakes, made up data and moved files onto the open internet without permission.”
This most-recent revelation is sparking more debate about how to oversee and monitor AI so that its work is in “alignment” with humans’ interests.
Leading the charge in addressing alignment is Sen. Mark Warner (D-Va), who is calling for legislation by the end of the year that establish AI safety standards. Saying that he is “not a doomer” and believes “This talk about human extinction is over the top,” Warner’s concerns are more pragmatic. “The idea that these swarms of AI agents might jump the fence and take over a banking system or take over a water system or, God forbid, take over a hospital” are what he believes requires immediate action.
The challenge cannot be left up to government alone. A piece from the Darden Business School at the University of Virginia argues that the AI alignment discussion belongs in every company. “In frontier AI research,” the report’s authors’ write, “alignment generally means keeping increasingly capable systems consistent with human intentions and values…. As businesses accelerate their adoption of AI and begin deploying autonomous agents, leaders need to just as vehemently discuss alignment within their own organizations.”
At the center of the paper’s recommendations is so-called stakeholder theory. “The theory,” the reports’ authors write, “asks managers to consider the connected interests of customers, suppliers, employees, investors, communities and others with a stake in the enterprise.”
In addition to calls for government regulation and challenging businesses to work on alignment, there are increasing calls to slow AI development. Whether this is a viable idea is the focus of a discussion among Walid Saad, Cayce Myers, and David Townsend at Virginia Tech. There discussion is available in the Q&A from VT News titled “The AI Slowdown Debate.”
Is your company, school, or agency doing innovative work with AI or involved in advancements in data center technology? Email Martin Davis to have your work highlighted in Daily Tech Notes.
