Anthropic's co-founder and CEO Dario Amodei has called the recent AI Action Summit in Paris a "missed opportunity," stressing that the pace of AI technology progression demands a greater sense of urgency and focus on several critical topics. In a statement released on Tuesday, Amodei expressed his concerns, which were further elaborated during an on-stage interview with TechCrunch at the developer-focused event in Paris, held in partnership with French startup Dust.

Amodei defended a third path that steers clear of pure optimism and pure criticism on AI innovation and governance, respectively. As a former neuroscientist, he drew parallels between understanding real brains and artificial brains, highlighting the importance of keeping pace with the rapid development of AI models. "Our understanding has to keep up with our ability to build things. I think that's the only way," he emphasized.

The tone of the discussion around AI governance has shifted significantly since the first AI summit in Bletchley, UK, partly due to the current geopolitical landscape. US Vice President JD Vance's remarks at the AI Action Summit, focusing on AI opportunity rather than safety, underscore this shift. Amodei, however, is keen to avoid antagonizing safety and opportunity, believing that increased focus on safety presents opportunities in itself.

At the Anthropic event, Amodei recalled the original Bletchley Summit, where discussions centered around testing and measurement for various risks. He argued that these efforts did not slow down technological progress, but rather helped better understand AI models, ultimately leading to improved models. Amodei reiterated Anthropic's commitment to building frontier AI models, emphasizing that the company's focus on safety should not compromise the promise of AI technology.

The conversation also touched on Chinese LLM-maker DeepSeek's recent models, with Amodei downplaying the technical achievements and questioning the public reaction. He expressed concerns about the geopolitical implications of non-US-based labs dominating AI technology. Amodei also disputed DeepSeek's claimed training costs, labeling them "just not accurate and not based on facts."

Looking ahead, Amodei teased upcoming releases from Anthropic, including models with reasoning capacities. He highlighted the company's efforts to solve the model selection conundrum, where users struggle to choose the most suitable model for their needs. Amodei envisions a smoother transition between pre-trained models and models trained with reinforcement learning, ultimately aiming to create a single, continuous entity.

As large AI companies like Anthropic continue to release better models, Amodei believes this will open up opportunities to disrupt industries worldwide. He cited examples of Anthropic's work with pharma companies, where Claude has been used to write clinical studies, reducing the time required from 12 weeks to three days. Amodei foresees a "renaissance of disruptive innovation" in the AI application space, with Anthropic committed to supporting and driving this progress.

For more information on the Artificial Intelligence Action Summit in Paris, read our full coverage.