An Adviser to Elon Musk’s xAI Has a Way to Make AI More Like Donald Trump

A researcher affiliated with Elon Musk’s startup xAI has discovered a brand new strategy to each measure and manipulate entrenched preferences and values expressed by artificial intelligence fashions—together with their political beliefs.

The work was led by Dan Hendrycks, director of the nonprofit Center for AI Safety and an adviser to xAI. He means that the method might be used to make well-liked AI fashions higher mirror the desire of the citizens. “Possibly sooner or later, [a model] might be aligned to the precise consumer,” Hendrycks informed WIRED. However within the meantime, he says, an excellent default could be utilizing election outcomes to steer the views of AI fashions. He’s not saying a mannequin ought to essentially be “Trump all the best way,” however he argues after the final election maybe it needs to be biased towards Trump barely, “as a result of he gained the favored vote.”

xAI issued a new AI risk framework on February 10 stating that Hendrycks’ utility engineering method might be used to evaluate Grok.

Hendrycks led a group from the Heart for AI Security, UC Berkeley, and the College of Pennsylvania that analyzed AI fashions utilizing a way borrowed from economics to measure shoppers’ preferences for various items. By testing fashions throughout a variety of hypothetical eventualities, the researchers have been capable of calculate what’s often known as a utility operate, a measure of the satisfaction that folks derive from an excellent or service. This allowed them to measure the preferences expressed by totally different AI fashions. The researchers decided that they have been usually constant quite than haphazard, and confirmed that these preferences grow to be extra ingrained as fashions get bigger and extra highly effective.

Some research studies have discovered that AI instruments reminiscent of ChatGPT are biased in the direction of views expressed by pro-environmental, left-leaning, and libertarian ideologies. In February 2024, Google confronted criticism from Musk and others after its Gemini device was discovered to be predisposed to generate pictures that critics branded as “woke,” reminiscent of Black vikings and Nazis.

The method developed by Hendrycks and his collaborators gives a brand new strategy to decide how AI fashions’ views could differ from its customers. Ultimately, some specialists hypothesize, this type of divergence may grow to be probably harmful for very intelligent and succesful fashions. The researchers present of their research, as an example, that sure fashions constantly worth the existence of AI above that of sure nonhuman animals. The researchers say additionally they discovered that fashions appear to worth some folks over others, elevating its personal moral questions.

Some researchers, Hendrycks included, consider that present strategies for aligning fashions, reminiscent of manipulating and blocking their outputs, might not be adequate if undesirable objectives lurk underneath the floor throughout the mannequin itself. “We’re gonna need to confront this,” Hendrycks says. “You possibly can’t faux it’s not there.”

Dylan Hadfield-Menell, a professor at MIT who researches strategies for aligning AI with human values, says Hendrycks’ paper suggests a promising course for AI analysis. “They discover some fascinating outcomes,” he says. “The primary one which stands out is that because the mannequin scale will increase, utility representations get extra full and coherent.”