The Lesson of the OpenAI Drama May Not Be What You Think
Hi friends –
If you’re a tech nerd like me, you may have spent the past few days glued to your phone, watching the twists and turns of the OpenAI drama as it unfolded on the app formerly known as Twitter. Certainly, the tale has all the drama of a Shakespearean play, as the OpenAI board fired the CEO, the staff revolted, Microsoft stepped in to hire the ousted players, and even as I write, negotiations continue to possibly bring back the CEO and remake the OpenAI board.
But the tale also pointed to something deeper: which is the cementing of the for-profit model in the artificial intelligence market. OpenAI was the most prominent nonprofit in the space, founded in 2015 with a mission to “ensure that artificial general intelligence benefits all of humanity.” Eventually, however, it transformed into a hybrid nonprofit/for-profit in order to raise the enormous amounts of capital needed to build artificial intelligence products. In whatever fashion it emerges from this fracas it will definitely be more beholden to investors like Microsoft than it was before.
In this week’s article for New York Times Opinion (gift link), I ask whether we want AI to be primarily in the hands of the private sector and whether it’s time to think about what a real public version of AI could look like.
Here’s my argument in a nutshell:
What’s the Problem?: AI is extremely expensive to build. It requires collecting massive data sets and using enormous amounts of computational power to analyze all that data. OpenAI estimates that the cost of training just one of its chatbots, GPT-4, was $100 million.
Even renowned researchers like Fei-Fei Li, an AI pioneer and director of Stanford’s Human-Centered Artificial Intelligence, told me that she couldn’t afford the massive amounts of computing power needed to analyze huge troves of video data. “The public sector is very under resourced,” she said.
Consider that in 2023 the U.S. government is expected to invest $1.8 billion into AI research, while venture capital invested almost $18 billion in AI in the third quarter alone. And access to AI is largely controlled by three companies — Amazon, Alphabet and Microsoft — that comprise two-thirds of the global cloud computing market that is essential to building powerful A.I. models.
To raise the capital it needed, OpenAI established a complicated corporate governance structure, with a nonprofit arm overseeing a for-profit unit. This allowed the company to raise more than $13 billion from Microsoft – with some of the investment in the form of credits to access Microsoft’s cloud computing infrastructure.
Now, the disarray is pushing OpenAI even closer to Microsoft, serving as a nail in the coffin of the most prominent effort to build a noncommercial version of AI.
What Can be Done?: Technologist Bruce Schneier argues that we need an AI public option. “A publicly funded LLM could serve as an open platform for innovation, helping any small business, nonprofit, or individual entrepreneur to build AI-assisted applications,” he wrote earlier this year.
Others say that we need to break up Big Tech’s control over the cloud computing market. Europe is already looking to build its own cloud, and the UK is investigating US cloud providers for abusing their market power.
Barry Lynn, executive director of the Open Markets Institute, said that cloud computing should be separated from Big Tech’s other businesses and regulated as an essential utility.
“This is foundational infrastructure for our entire online economy,” he said. “The fact that there are only three corporations that do this gives them all sorts of power, including the power to exclude competitors or set pricing in a discriminatory way, and it also leads to them not paying enough attention to stability and resiliency.”
What Would That Look Like?: Making cloud computing into a public resource that anyone could use for a modest fee, like public libraries, could spur huge innovation.
Professor Li would be able to build her AI models. OpenAI would not have been forced into the arms of a for-profit company. OpenAI would likely not be the only prominent nonprofit in the AI space.
It has the making of the kind of coup I could support: one where the public took back the power of computation.
As always, thanks for reading.
Best
Julia