- Anthropic’s chief executive claims AI capabilities are on an exponential curve, yet the firm’s own research says otherwise.
- In an online essay, Dario Amodei repeats the word “exponential” six times and cites old scaling‑laws papers to argue that a year or two of continued growth could yield a so‑called Powerful AI. At the same time, system cards for Claude Mythos and the newer Fable 5 model state that observed gains are attributable to human work and that no sustained AI‑driven acceleration has been seen. The Fable card explicitly notes a lack of a 2× speedup in progress.
- The clash matters because investors and policymakers are listening to the CEO’s hype while the technical data suggest the scaling argument is losing steam. If Anthropic’s internal metrics are correct, expectations of rapid, self‑reinforcing breakthroughs may be overstated.
- In short, the company’s public optimism appears out of step with its own evidence, a reminder that not every headline‑grabbing claim survives scrutiny.