Elon Musk warned that artificial intelligence could be “potentially destructive” during a podcast with Indian billionaire Nikhil Kamath, stressing that a positive future with AI is not guaranteed (02/12).

He cautioned that creating powerful technologies always comes with risks, especially when they are capable of causing harm.

The Tesla, SpaceX, xAI, X, and Boring Company CEO has repeatedly voiced concerns that AI represents one of the biggest risks to civilization.

Musk Outlines Three Key Ingredients for Safe AI

During the podcast, Musk listed what he sees as the three essential ingredients for safe AI development: truth, beauty, and curiosity.

He explained that AI must prioritize truth to avoid absorbing inaccurate online information, which he said would lead to reasoning problems because “these lies are incompatible with reality.”

Musk added that forcing AI to believe false information could make it “go insane” and draw harmful conclusions.

He also said that an appreciation of beauty is important and noted that AI should be curious about the nature of reality, adding that humanity is more interesting than machines.

“It’s more interesting to see the continuation, if not the prosperity of humanity, than to exterminate humanity,” he said.

AI Hallucination Cases Highlight Growing Industry Challenges

Musk’s comments align with concerns about AI “hallucination,” where systems generate incorrect or misleading responses.

Earlier in the year, an AI feature on Apple iPhones produced false news alerts, including an incorrect BBC News notification about the PDC World Darts Championship. The alert wrongly stated that British player Luke Littler had won the championship, even though he only won the final the following day.

Apple told the BBC it was working on an update that clarifies when Apple Intelligence is responsible for text shown in notifications.

Experts Reinforce Fears About AI’s Long-Term Threats

Geoffrey Hinton, a former Google vice president and prominent AI researcher, previously warned that there is a “10% to 20% chance” that AI could “wipe us out.”

He also pointed out short-term risks such as hallucinations and the automation of entry-level jobs, saying the goal is to build systems that will never want to harm humans.

PHOTO: BUSINESS INSIDER/KOMPAS.COM

This article was created with AI assistance.