Policymakers in Washington are now engaged in a wide range of discussions regarding how to prevent AI technology from spinning out of control. But for all the focus on various dangers, we're losing sight of how the AI economy should be structured — how humanity can be best poised to take best advantage of this new frontier. Put another way, we're not paying enough attention to the question of which companies should be allowed to harness AI's potential.
While AI will surely become less expensive as specific industries find ways to harness these new tools in their own particular ways, the present costs of building the large language models that power today's generative AI are so prohibitive that most of the innovation is being driven not from the bottom up by small startups, but instead by Big Tech. That marks a departure from the ordinary patterns of creative discovery, if only because innovation is driven typically by disruptive new companies challenging the incumbents. Rather than upending the old order, there's a real possibility in this case that disruption will help the big get bigger, with challengers struggling to gain any real traction.
Perhaps more worrisome, the overwhelming focus on public fears of AI may spur policymakers to undermine what many of us call "open source" AI altogether — upending the collaborative model that has enabled a global community of innovators to work both iteratively and rapidly to build and improve the underlying technology. That should be a concern for everyone because, while we should not minimize the grave risks that come with the possibility that AI could get into the wrong hands, we must realize that failing to explore this frontier expeditiously will undermine its potential to improve health, education, and many other aspects of our lives in the near- and medium-term.
With the downside risks primarily in mind, many have predictably called for more government-managed innovation, with bureaucrats doling out AI licenses to a select group of companies — in most cases, our existing tech giants. Doing that would be abandoning the formula that has led to America's rise in favor of the top-down approach to industrial policy that China prefers. On this side of the Pacific, we've adhered to the notion that the best ideas can't be orchestrated — that they are born from the almost random collision of ideas. At its most effective, Washington works not to control any rapidly evolving industry, but to level the playing field so that the best innovations are able to scale.
If that were happening in the realm of AI, new startups would already be sprouting across the county. But a recent Brookings Institution study found that nearly 60 percent of AI jobs are currently based in Silicon Valley. That's worrisome. Policymakers need to ensure that entrepreneurs across the country have the opportunity to participate in the race to put this new technology to good use. That means ensuring that those with unique expertise who live outside Silicon Valley have a way to join the AI ecosystem. In short, AI shouldn't be a vehicle used to ensure that Silicon Valley-based Big Tech can extend its dominance.
America has not always remained true to its embrace of bottom-up innovation — and the innovation economy has suffered as a result. The iterative experimentation that eventually birthed the commercial Internet was delayed until the 1990s not because the technology wasn't ready — the underlying infrastructure had been invented in the 1960s. Rather, the problem was that AT&T convinced policymakers that it was too dangerous to give outside players access to their closed access system. The power of the Internet was unleashed only in the 1980s, when the government used antitrust policy to break up Ma Bell, and then erected a regulatory structure that forced telephone companies to open up their networks to competitors — enabling companies like mine, America Online, to start.
These lessons of the Internet should be applied to this new frontier. To maximize the benefits of AI while reducing the risks, the government should look to put guardrails in place, but not in the form of licenses that only enable only a handful of players to compete. Rather, congressional legislation should have a bias towards unleashing innovation all across the country. This includes ensuring the continued development of open-source AI platforms, and also taking steps to ensure that the big tech AI platforms adopt the same open access provisions that were applied to the phone companies nearly a half-century ago, when the Internet was being birthed.
It's important now that Washington find a way to establish ground rules that allow entrepreneurs to participate in the development of AI, and ensure that the current generation of tech leaders aren't able to pull up the ladder behind them. At a minimum, Washington must be committed to ensuring that medical researchers in the Carolinas, and burgeoning ag-tech firms in Arkansas, and clean energy startups in the Mountain West are able to harness the power of AI and then, if their ideas win in the marketplace of ideas, that their creators are able to harvest the benefits. This new technology should not be a wedge that further separates the tech world from the rest of America. It should be a bridge that connects the two.
Steve Case, a co-founder of America Online, is chairman and chief executive of Revolution, a Washington, D.C.-based venture capital firm, and author of "The Rise of the Rest: How Entrepreneurs in Surprising Places Are Building the New American Dream." He will participate in Sen. Schumer's Forum on AI on Oct. 25.