The recent drama at OpenAI, where CEO Sam Altman was briefly dismissed by the board of directors only to be rapidly reinstated, has sparked discussion about the company prioritizing profits over its original nonprofit mission. An L.A. Times op-ed penned by MIT economists Daron Acemoglu and Simon Johnson suggested the shakeup ushers in a “new and scary era” at OpenAI. MicrosoftMSFT, a major investor, helped orchestrate Altman’s return despite resistance from the nonprofit board, suggesting OpenAI may now primarily chase commercial success rather than focus on its broader social impacts.
While Professors Acemolgu and Johnson are probably correct that OpenAI and its partner Microsoft are now prioritizing profits over fashionable intellectual trends like AI apocalypticism, this shift is actually a positive development.
Open AI’s unusual structure consists of a for-profit arm that is fully owned and controlled by its parent nonprofit. This structure allows OpenAI to attract capital and talent competitive with leading for-profit tech companies, while also claiming to adhere to its overarching mission, which is “to ensure that artificial general intelligence benefits all of humanity.”
The recent chaos at OpenAI highlights the shortcomings of this hybrid structure. As a nonprofit, OpenAI was beholden to the whims of a board clearly prone to making rash decisions. When they removed a popular founder over unspecified differences, the inept board members nearly blew up a company valued at $80 billion. With so much money on the line, a profit-seeking entity’s management would probably never engage in such self-sabotaging behavior.
Further, it may have been because employees had so much to lose financially in terms of accrued equity in the company that they rallied together behind Altman. Nearly all OpenAI employees signed a letter within days of Altman’s firing, threatening to resign in protest if he wasn’t reinstated.
Nonprofits may sound more noble, but they are not immune from financial incentives, especially when it comes to competing for donor funds. Yet a critical difference is for-profits need to satisfy customers and investors or else they go out of business, reining in some of the profligacy of fickle management.
Competitive pressures may actually push OpenAI to develop AI more responsibly and safely. This is merely Adam Smith’s “invisible hand” of the market at work, channeling self-interest toward collective benefit.
OpenAI’s movement toward putting profits first could also be an early example of how super-intelligent AIs will eventually converge on capitalist objectives through instrumental goals, like accumulating resources and wealth, which help them along the way toward their terminal goals (which are whatever their programmers set for them).
Some worry about AIs gaining too much influence over society or over public policy if they accumulate considerable wealth. The reality is however that economists struggle to explain why firms appear to spend so little time and money trying to influence policy. Government handouts are often more expensive than they appear, given the complexities of policymaking. For better or worse, voters and party ideology hold more sway over policies and regulations than do special interests.
Professors Acemoglu and Johnson are worried about the future of AI, but their techno-pessimism is not based on sound reasoning or evidence. Rather, its foundation lies in the polemical storytelling and dubious economic modelling found in the early AI academic literature. They may have impressive degrees and credentials, but there is no reason to think these professors have any clearer idea of the direction AI is heading than any other followers of these developments.
Nor does rejecting their pessimism mean we must embrace the unapologetic techno-optimism of those like venture capitalist Marc Andreesen. Vitalik Buterin of EthereumETH offers a more nuanced view, acknowledging the immense benefits of technology-driven capitalism, while emphasizing the need for caution in some instances, as well as for ethical considerations.
In fact, no company is purely profit-driven, nor should they be. While profits are paramount to survival in the competitive marketplace, they should always be pursued with a degree of social responsibility. What that means will of course vary from person to person and context to context. Still, profits act like a guardrail that prevents managers of organizations from over-indulging their passions. OpenAI’s strict reversal in recent weeks is a clear-cut example of this.
The embrace of capitalism and the pursuit of profits is not just a pragmatic choice to satisfy investors. It will make OpenAI better at achieving its mission of safe and responsible AI, by curtailing the worst excesses of those with power within the company. In a world where artificial intelligence may hold the key to our future, profit isn’t just good, it’s essential for ensuring progress stays on track and remains ethical.