A Glimpse at a Post-GPT Future

This is Atlantic Intelligence, a newsletter in which our writers help you wrap your mind around artificial intelligence and a new machine age. Sign up here.

The GPT era may be coming to a close. OpenAI announced yesterday the full release of a new set of “reasoning” models called o1. As my colleague Matteo Wong explains in a new article—for which he talked with OpenAI staffers and independent AI experts, and pored over research papers—this moment represents a legitimate break with the prediction-based technology that has so far defined generative AI. The release of o1 “has provided the clearest glimpse yet at what sort of synthetic ‘intelligence’ the start-up and companies following its lead believe they are building,” Matteo writes.

To a casual user, the o1 models may not appear so different from the GPT series that has powered OpenAI’s famous chatbot. Type a prompt, get a response—sometimes with quirky or mystifying errors. Beneath the hood, however, o1 operates less like a “parrot” mimicking its training data and more like a maze rat, running through possible responses and automatically evaluating and revising its own output before it presents you with a final answer. This process makes o1 particularly well suited to tasks with verifiable solutions, such as testing computer code for bugs. It also requires a tremendous amount of computing power and energy.

OpenAI has said that the arrival of o1 puts humanity on a new path toward a supposed superintelligence. There’s plenty of room for doubt about that claim. But, in any case, the release and its surrounding rhetoric seem likely to fulfill a core function for the company: attracting more interest and investment at a time when generative AI’s growth appears to have otherwise stalled, and its future is still not altogether certain.

The GPT Era Is Already Ending

By Matteo Wong

This week, OpenAI launched what its chief executive, Sam Altman, called “the smartest model in the world”—a generative-AI program whose capabilities are supposedly far greater, and more closely approximate how humans think, than those of any such software preceding it. The start-up has been building toward this moment since September 12, a day that, in OpenAI’s telling, set the world on a new path toward superintelligence.

That was when the company previewed early versions of a series of AI models, known as o1, constructed with novel methods that the start-up believes will propel its programs to unseen heights. Mark Chen, then OpenAI’s vice president of research, told me a few days later that o1 is fundamentally different from the standard ChatGPT because it can “reason,” a hallmark of human intelligence. Shortly thereafter, Altman pronounced “the dawn of the Intelligence Age,” in which AI helps humankind fix the climate and colonize space. As of yesterday afternoon, the start-up has released the first complete version of o1, with fully fledged reasoning powers, to the public. (The Atlantic recently entered into a corporate partnership with OpenAI.)

On the surface, the start-up’s latest rhetoric sounds just like hype the company has built its $157 billion valuation on. Nobody on the outside knows exactly how OpenAI makes its chatbot technology, and o1 is its most secretive release yet.

Read the full article.

What to Read Next

Earlier this week, The Atlantic published the full script for the Broadway play McNeal, by Ayad Akhtar, which deals extensively with questions of creativity and humanity in the generative-AI era. As the actor Jeremy Strong writes in his foreword:

The magic trick of Akhtar’s play—its triple axel—is its human vision of McNeal within a scaffolding that becomes ever more generated by AI. Without a character like McNeal, and without one of our greatest actors in Robert Downey Jr.—without both a compelling human character and a human actor to give the part density and weight and anguish and pain—we would be left with only the scaffolding. Just the machine, without the ghost, without the tender nerve and sinew of life. As McNeal circles the abyss of, in his words, absolution or annihilation, we feel, within this dazzling cathedral constructed of ones and zeroes, the presence of a broken human heart. The tragedy of a single, fallible human against the backdrop of a new kind of infinity, which knows only efficiency and the global maximum.

P.S.

Spotify—the most algorithmically enthralled of all the music-streaming services—released its annual “Wrapped” feature this week. In addition to informing users of their most-listened-to music throughout the year, as is standard, Spotify also presented them with bespoke, AI-generated podcasts. (In mine, for example, the synthetic hosts rambled about the “serious energy” I was channeling in January, when I listened to a lot of the death-metal band Bolt Thrower.) This year, the entire Wrapped endeavor struck me as remarkably lifeless—a reminder that human art (even the lowbrow) is personal in ways that a program could not possibly comprehend. Last year, my colleague Nancy Walecki wrote a lovely story on Spotify Wrapped explaining just that.

— Damon