Barry Diller warns AGI unpredictability requires guardrails despite trusting Sam Altman

Billionaire media mogul Barry Diller has publicly defended the character of OpenAI CEO Sam Altman, asserting that the executive is sincere and possesses good values. The comments were made during the Wall Street Journal's "Future of Everything" conference, where Diller sought to counter recent reporting that has accused Altman of manipulative and deceptive behaviour. Despite his personal confidence in Altman's integrity, Diller argued that relying on the trust of individual leaders is an inadequate strategy for managing the risks associated with artificial intelligence.

Diller emphasised that the primary concern regarding the development of Artificial General Intelligence, or AGI, is the unpredictable nature of the technology and its potential to outperform humans on any task. He stated that the consequences of AGI are currently unknown, even to the creators who are in the process of building it. According to Diller, this great unknown means that personal trust is irrelevant, as the outcomes may surprise even the most dedicated stewards of the industry.

The media executive, who serves as chairman of IAC and Expedia Group and co-founded Fox Broadcasting, noted that the industry is moving towards AGI quicker and quicker. While he believes most AI leaders are good stewards, he warned that the rapid pace of progress means humanity is approaching a critical juncture where the technology could change almost everything. Diller suggested that whether huge investments succeed or fail is secondary to the fundamental reality that the technology is advancing regardless of financial backing.

Central to Diller's argument is the necessity of implementing strict guardrails before AGI is fully unleashed. He warned that if humans fail to establish these safety measures, an autonomous AGI force may implement them itself. Once such a force acts independently, Diller cautioned that there would be no going back, highlighting the irreversible nature of the potential outcomes if the technology escapes human control.

Diller did not specify which other AI leaders he considers insincere, leaving the identity of potential bad actors unconfirmed. His assertion that trust is irrelevant is framed as a philosophical stance on risk management rather than a factual prediction of specific AI behaviour. Furthermore, the specific mechanisms by which an AGI might self-implement guardrails were not detailed in his remarks.

Ultimately, Diller's message underscores a shift in focus from the character of individual executives to the systemic risks of the technology itself. He argued that the issue is not stewardship, but rather dealing truly with the unknown variables that accompany the creation of AGI. As the technology nears its final form, the call for rigorous safety protocols remains paramount, regardless of the personal reputations of those leading the charge.