Anthropic CEO Predicts Superintelligent AI by 2027, Defends Scaling
Anthropic CEO Dario Amodei insists that scaling up AI models remains the best path to superintelligence, despite rising costs and unpredictable behaviors. He predicts human-level AI could arrive by 2026 or 2027, raising concerns about extreme power concentration.
Anthropic CEO Dario Amodei shares his belief that scaling up AI models is still a viable path toward achieving more capable systems. He explains that this scaling involves increasing the compute used for training, the physical size of the models, and the size of their training data sets, despite a lack of theoretical understanding as to why this approach works so well.
Amodei dismisses concerns about a looming data shortage, suggesting that developers will overcome this limitation through synthetic data generation and extrapolation. However, he acknowledges that computing power is becoming significantly more expensive, predicting that companies will spend billions on training clusters next year and hundreds of billions by 2027 to support this massive growth.
Even with these massive investments, Amodei admits that controlling advanced models remains incredibly difficult and compares the process to a frustrating game of whack-a-mole. Nevertheless, he anticipates that a superintelligent AI surpassing human-level performance on many tasks will emerge by 2026 or 2027, which leads him to worry primarily about the resulting economics and the extreme concentration of power.