Anthropic CEO says AI growth is exponential. Anthropic research says otherwise.

Key Points:

• Anthropic CEO Dario Amodei claims in a recent essay that AI capabilities are increasing exponentially, referencing "scaling laws" that predict rapid cognitive growth with more computing power, potentially leading to "Powerful AI" within a few years.
• However, this assertion appears to conflict with statements from Anthropic's own model system cards, which indicate that recent AI progress is largely due to human research rather than AI-driven acceleration, and no evidence supports a sustained AI-attributable doubling in progress pace.
• The key paper Amodei cites to support exponential growth, a 2020 study co-authored by Jared Kaplan, has faced criticism from experts like Gary Marcus, who argue that scaling benefits are diminishing and certain AI capabilities such as reasoning and truthfulness are not improving exponentially.
• Despite Amodei's optimistic outlook, recent AI model releases, including GPT-5, have not demonstrated the anticipated superintelligence or exponential cognitive leaps, casting doubt on the near-term arrival of transformative AI as predicted.
• The debate highlights ongoing uncertainty within the AI research community about the pace and nature of AI advancement, with calls for further clarification from Anthropic on these conflicting positions.