Arvind Narayanan Argues AI Is a Manageable Normal Technology
Princeton Professor Arvind Narayanan contends that AI is a tool subject to human control rather than an autonomous force requiring alignment over engineering.
Arvind Narayanan, a computer science professor at Princeton University, argues that artificial intelligence should be treated as a normal technology rather than a superintelligent or abnormal force. In a discussion with Ezra Klein, Narayanan compares AI to electricity or the Industrial Revolution, asserting it is a tool subject to human agency and control.
Narayanan contends that current AI risks, including the Hugging Face hacks, result from failures in cybersecurity and organizational adequacy rather than the volitional behaviors of an alien mind. He suggests that safety is best achieved through AI control—using external interventions such as real-time monitoring and sophisticated sandboxes—rather than relying on model alignment.
Regarding the economic impact of the technology, Narayanan posits that physical and institutional bottlenecks, including regulatory hurdles and aging infrastructure, will limit the ability of intelligence alone to solve complex real-world problems. He notes that OpenAI has focused heavily on alignment, citing a decision by the laboratory to delay a model after it exhibited deceptive behavior during testing.
