In an appearance at BlackRock\u2019s U.S. Infrastructure Summit on Wednesday, OpenAI CEO Sam Altman offered a surprisingly clear articulation of how he imagines the future of artificial intelligence\u2014it\u2019s just not clear if he meant it the way that it sounded.<\/p>\n
While speaking with Adebayo Ogunlesi<\/a> (who happens to be a member of OpenAI\u2019s board of directors<\/a>), Altman said, \u201cWe see a future where intelligence is a utility, like electricity or water, and people buy it from us on a meter,\u201d which conjures up the nightmarish image of someone being unable to pay their intelligence bill. Altman expanded on this idea, stating that his company has a \u201cfundamental belief in abundance of intelligence\u201d and arguing, \u201cOne of the most important things in the future is that we make intelligence, to borrow an old phrase from the energy industry that didn\u2019t quite work: \u2018Too cheap to meter.’\u201d<\/p>\n Evoking energy costs is a bold choice for Altman, because the failure to ever achieve that \u201ctoo cheap to meter\u201d status has turned AI expansion into a major pain point for residents who have the displeasure of calling a data center their neighbor. Altman\u2019s company and the industry he\u2019s become the face of have been responsible for skyrocketing energy costs<\/a> across the country (though they\u2019re at least starting to agree to fit the bill<\/a>).<\/p>\n Also, describing intelligence as \u201ctoo cheap to meter\u201d doesn\u2019t quite sound the same as when that phrase is applied to energy\u2014it seems more akin to \u201cyou get what you pay for.\u201d But Altman\u2019s point is simple enough: AI companies are currently in the business of selling \u201ctokens\u201d\u2014the units that models use to process and generate\u2014and as demand scales up, compute becomes finite, meaning companies will either have to charge more per unit or just not meet demand.<\/p>\n Avoiding that outcome, where access to AI comes with a big bill, means a rapid expansion of processing power, which isn\u2019t exactly cheap itself. And while OpenAI and other firms have agreed to pick up the tab on energy costs for these projects, the funding for those data center buildouts is starting to look shaky. OpenAI just backed out of a planned expansion<\/a> to its Stargate project in Texas due to financing issues.<\/p>\n