When you’re building cutting-edge AI technology, regulation isn’t just a compliance issue – it’s an existential threat to innovation itself. In a recent episode of Category Visionaries, Contextual AI CEO Douwe Kiela shared a provocative perspective on AI regulation that challenges conventional wisdom about how to make AI safer.

The False Target

Most discussions about AI regulation focus on the underlying models themselves. But Douwe argues this fundamentally misunderstands the nature of AI technology: “We also didn’t regulate like the Linux kernel… or like a browser. And I think a language model is very similar to those pieces of technology where it can be used for good things and for bad things.”

This comparison to foundational technologies like operating systems and browsers reveals a crucial insight: attempting to regulate the base technology rather than its applications could have far-reaching unintended consequences.

The Hidden Power Play

The push for aggressive regulation of foundation models isn’t just about safety – it’s about market control. “The debate around regulating just the model itself and not the application of the model is really driven by a kind of cynical group of people who are, I think, trying to lock down the market in their favor,” Douwe explains.

This dynamic creates what he calls a regulation paradox: “If we overregulate the market too early, then we’re just going to have a few incumbents kind of take the entire market because they can afford to hire hundreds of lawyers and a company like Contextual just can’t afford to do that.”

Applications Over Architecture

Instead of trying to regulate AI models themselves, Douwe advocates for a more nuanced approach focused on specific applications: “Rather than the fundamental technology itself, we should regulate the application of the technology.”

This perspective aligns with how we handle other powerful technologies. We don’t regulate programming languages or databases – we regulate how they’re used in specific contexts like healthcare, finance, or consumer applications.

The Startup Stakes

For AI startups, the regulatory approach taken could determine who survives and who thrives. Heavy regulation of foundation models would force companies to invest heavily in compliance before they can even begin innovating. This creates an artificial moat that only the largest tech companies can cross.

The alternative – regulating applications rather than foundations – would maintain a more level playing field where startups can innovate on the technology while still ensuring safe and responsible deployment in regulated industries.

Looking Ahead: The 2024 Crucible

The stakes are particularly high as we enter 2024. “The hype train is going to stop at some point and so the tide is going to run out and a bunch of people are going to get caught swimming naked,” Douwe predicts. In this environment, over-regulation could accelerate market consolidation, pushing innovation into the hands of a few large players.

A Path Forward

The solution isn’t no regulation – it’s smarter regulation. By focusing on how AI is applied rather than trying to control the fundamental technology, regulators can protect against misuse while preserving the innovation ecosystem.

This approach would:

• Allow startups to build on foundation models without crushing compliance burdens
• Enable industry-specific regulation where it matters most
• Preserve competition and innovation in the AI space
• Focus regulatory resources where they can have the most impact

For founders building AI companies, this perspective offers both a warning and a roadmap. The warning: be prepared for regulatory battles that could reshape the competitive landscape. The roadmap: focus on specific applications and use cases where you can demonstrate clear value and safety, rather than trying to build everything from scratch.

The future of AI innovation may well depend on whether we can get this balance right – ensuring responsible development without inadvertently creating an AI oligopoly through well-intentioned but misguided regulation.