A tech world skirmish over whether to limit open-source AI models has escalated into a full-blown battle royale.
What happened: Silicon Valley has split into warring factions over how to regulate increasingly sophisticated AI models.
On one side, OpenAI and Anthropic are reportedly lobbying U.S. lawmakers for tighter controls over so-called open-source models, which they argue are now too powerful to be freely distributed in a safe manner.
On the other side: virtually everyone else, including the CEOs of Nvidia, Meta, Microsoft, IBM, and Alphabet, who have come out in support of open models, which can be freely downloaded and modified, arguing that restricting them would slow AI innovation and hobble smaller companies.
Zoom out: Looming over the debate is the geopolitical contest between the U.S. and China to achieve AI dominance. While OpenAI and Anthropic built an early lead with their closed-source models, Chinese companies’ latest open-source alternatives have improved quickly, and are capturing market share on the basis of being cheaper and ‘good enough’ for many businesses.
Anthropic and OpenAI claim that Chinese labs have closed the capabilities gap by “distilling” their technology — essentially, a form of copying — and have, in so doing, stolen intellectual property.
Of course, some (like Microsoft CEO Satya Nadella) have pointed out that this argument is a bit rich coming from companies that built their product using vast troves of books, music, video, and other publicly available data.
Why it matters: There’s a great deal at stake in how this issue gets decided. Depending on which side you fall on, restricting open-source models could be a path to safer development of the technology or a bad mistake that cripples the U.S. AI ecosystem and raises costs for businesses by limiting competition.
