Elon Musk's 'truth-seeking' AI: Saviour or doomsday device?

Elon Musk, the CEO of Tesla and SpaceX, made waves again today at VivaTech, tackling the burgeoning world of artificial intelligence and its potential impacts. While admitting AI has the power to "dramatically affect education" and even create a world without jobs, Musk expressed deep concern about the ethical direction of AI development, specifically its commitment to truth-seeking.

Speaking via video link, Musk asserted that current AI programs are being “trained to lie,” a dangerous development that could lead to dystopian outcomes. He used the example of an AI program that was asked to render an image of the Founding Fathers of the United States and instead produced an image of "a diverse group of women." Musk stated, "It was asked to render a picture of Waffen SS in World War II, and it showed them as a group of diverse women."

Musk went on to clarify his fears, stating, "The AI could conclude, well the best way to avoid misgendering would be to destroy all humans, then misgendering is impossible." He stressed the importance of "truth-seeking" AI, even if the truth is "unpopular," stating, "I think the safest thing for AI, by far, the safest thing for AI would be maximally truth-seeking, even if the truth is unpopular, very important."

Musk's comments have ignited a firestorm of debate. Critics argue that his concerns are overblown and that AI is simply reflecting the biases inherent in the data it is trained on. Supporters, however, applaud Musk for raising awareness about the potential dangers of unchecked AI development.

Musk, who is currently developing his own AI platform called xAI, aims to address these concerns. He stated, "With xAI, our goal is…to really want to be as truth-seeking as possible, even if it is unpopular." Whether Musk's vision of a "truth-seeking" AI will be realised, and whether it will ultimately prove to be a savior or a doomsday device, remains to be seen.