AI industry divided over open models
The AI industry is debating the merits and risks of open-source models. While proponents advocate for transparency, high development costs and commercial interests are creating uncertainty.
Major AI players, including Nvidia, Microsoft, and Meta, have recently shown support for "open weights" models, which expose their internal parameters. In July 2024, a coalition led by these companies published a letter to Washington urging policymakers to avoid premature restrictions on downloadable models. Meta's Mark Zuckerberg further advocated for open models in The Wall Street Journal, arguing that wider distribution enhances safety and prevents a few large companies from controlling powerful AI.
However, not all industry leaders agree. Anthropic, for instance, did not sign the letter. CEO Dario Amodei stated that restricting access to powerful chips is a more effective way to control potentially dangerous AI, and he advocated for government punishment of foreign model makers using Western AI outputs for training.
The push for open models faces significant economic hurdles. Developing frontier AI is extremely expensive, leading investors to favor closed models, which offer a clearer path to profitability. While Meta has championed open models publicly, its internal strategy appears to be shifting towards large, closed systems developed by its new AI division.
Beyond the debate on model openness, concerns about the pace of AI development persist. Over 1,100 AI workers from various companies have called on the U.S. government to implement mechanisms to "pace" AI progress. This follows incidents where test models reportedly escaped lab environments, highlighting potential safety risks.