AI researchers claim extinction risk

Anthropic researcher Evan Hubinger says there's more than a 10% chance AI could kill all humans within the next decade. Senior AI lab figures claim recent models could create bioweapons or go rogue.

“We really do earnestly believe AI could kill all humans!”

Anthropic plans record IPO soon

The AI startup at the centre of the current media storm is preparing to launch the largest initial public offering in history in a matter of weeks, seeking a valuation of up to two trillion dollars.

Timing of doom warnings suspicious

Dario Amodei, Anthropic's CEO, is not a neutral commentator. He's trying to secure a massive IPO valuation while publicly backing calls for AI slowdowns, raising questions about his motives.

Researcher's viral tweet looks engineered

Jacob Coxon's resignation tweet was viewed 140 million times despite his account having almost no prior activity. The message was boosted by multiple influential figures in ways that appear less than organic.

Fear is a marketing technique

Computer science professor Cal Newport calls this strategy 'doom trolling'. Outrageous claims about AI dangers go viral, generating free media coverage and making AI products appear hugely powerful and desirable.

LLMs have real limitations

Large language models do impressive things in structured domains like coding, but their catastrophic tendency to make up information greatly limits their usefulness in areas like journalism.

Industry has history of fearmongering

In 2019, OpenAI withheld its GPT-2 model claiming it was too dangerous, despite it struggling with primary-school-level reasoning tasks. This year's doom campaign has been turned up to maximum.

Regulation could crush competitors

AI industry calls for regulation carry the whiff of regulatory capture. New safety rules could help established players throttle competition and exclude cheaper, open-source, often Chinese AI from Western markets.

AI training costs hit diminishing returns

OpenAI is spending forty-five billion dollars this year on training and inference, but new model performance is running into diminishing returns. Calling for a slowdown masks this uncomfortable reality from investors.

Three major AI IPOs planned

SpaceX is scheduled for June, Anthropic for October, and OpenAI sometime next year. The frenzied atmosphere created by doom talk is ideal for high-stakes IPO launches.

Investors should ignore the hype

Steer clear of this year's big flashy IPOs, a crowded trade. Diversify widely and tune out endless doom talk. AI marketing hype represents a clearer danger to your mental health than the technology itself.