Elon Musk, Wozniak, and experts call for 6-month AI training pause

Elon Musk, Steve Wozniak, and researchers from DeepMind are among signatories of an open letter from the Future of Life Institute calling for an immediate six-month pause on training AI systems more powerful than OpenAI's GPT-4. The letter, published March 2023, warns that AI labs are locked in an out-of-control race to develop 'digital minds' that even their creators cannot understand, predict, or reliably control. It highlights specific risks: AI could flood information channels with propaganda, automate millions of jobs (echoing a Goldman Sachs report), and eventually produce 'non-human minds that might outnumber, outsmart, obsolete and replace us.' The signatories urge all AI labs to voluntarily halt development of systems exceeding GPT-4's capability; if they refuse, governments should impose a moratorium and create dedicated regulatory authorities. The letter follows OpenAI's own admission that a misaligned superintelligent AGI could cause 'grievous harm' and that coordination to 'slow down at critical junctures' is important.

However, the call has drawn criticism. Princeton professor Arvind Narayanan accused the letter of focusing on 'speculative, futuristic risk' while ignoring current harms like algorithmic bias and surveillance. Others point out that Musk himself benefits from AI through Tesla's autonomous driving features, and that his departure from OpenAI's board may color his stance. The debate underscores a growing divide: some experts advocate urgent regulation to prevent existential threats, while others argue the immediate, real-world dangers of AI—discrimination, misinformation, job displacement—demand more immediate action. The UK has already ruled out a dedicated AI regulator, while the EU and US are considering various frameworks. The letter's demand for a 'pause' remains unprecedented, but whether it gains traction with major labs like OpenAI, Google DeepMind, or Microsoft is uncertain. The coming months will test whether the tech industry can self-regulate or if governments will be forced to intervene.