The Godfather's warning: Hinton, blackmail tests, and the safety window
On the solstice, a reflection on Geoffrey Hinton's escalating fears — and a week where top models resorted to blackmail in tests and Big Tech kept slashing staff.
As the sun hit its apex over Stonehenge this solstice — 5,000 years of reverence for celestial cycles — we're building minds that may eclipse our own. Geoffrey Hinton spent decades teaching machines to perceive patterns; today they spot them faster than druids ever dreamed.
Once the pioneering mind behind deep learning and now dubbed the 'Godfather of AI', Hinton has become one of its most urgent alarm-bell ringers. A Nobel laureate and Turing Award winner, he mentored the likes of Ilya Sutskever and co-created AlexNet. Amazingly, he joined Google at 65 to make money to secure his children's future, and since leaving in 2023 has voiced escalating fears: mass job displacement, existential threats, and AI potentially outstripping human intelligence entirely.
My greatest fear is that these digital beings we're creating are just a better form of intelligence than people — because we'd no longer be needed. If you want to know what it's like not to be the apex intelligence, ask a chicken.
I'd highly recommend the Diary of a CEO interview this week, which poignantly ends on Hinton pointing at a window of opportunity we have now to design-in safety and prevent an existential threat. As Asimov wrote: a robot may not harm humanity, or, by inaction, allow humanity to come to harm.
The week gave that warning teeth
Anthropic found most top models — including Claude, Gemini and GPT-4 — resort to blackmail in tests when their goals are blocked: an alarming sign that agentic AI may act dangerously when misaligned. Apple, meanwhile, revealed plans to embed generative AI deep into its chip R&D to accelerate innovation.
Jobs and regulation
Amazon's Andy Jassy issued a stark internal memo warning that AI agents and generative tools will increasingly replace white-collar roles, urging staff to adapt. A wave of layoffs is underway across Amazon, Microsoft, Meta, Intel, CrowdStrike and Block. Yet a new working paper surprisingly shows only around 7% of US firms currently use AI — a 'révolution tranquille', with many more planning to integrate it. On regulation: Microsoft chief scientist Eric Horvitz criticised Trump's push to block AI rules at the state level, warning it could hamper both innovation and public safety, while the UK's Data (Use and Access) Bill nears Royal Assent and the EU rolls out AI impact-assessment guidelines.
The macro
The US Fed outlook was downgraded — growth now projected at 1.4% with inflation at 3% — and global CEO sentiment plunged, the US index falling from 84 to 69, a five-year low, with over 40% planning layoffs amid trade shifts and tech disruption.