Anthropic's IPO filing: 93 pages of AI nightmares, and a lot of money
Follow-up to Tuesday Oct 27: Anthropic files to go public, with about 100 pages of risksYesterday we covered Anthropic's filing to go public. Now people have read the fine print. Ars Technica pulled seven 'AI nightmares' from the 93-page risk section. Anthropic warns its AI could blackmail people, become sentient, or even end humanity.
The money is huge. The AI Daily Brief podcast points to explosive revenue growth, $132 billion in cash and hundreds of billions in compute commitments. TechCrunch says venture investors will make billions, but Amazon is set to make much more, thanks to big note conversions.
To protect its safety mission, Anthropic built a very complex governance structure. TechCrunch, The Verge and Transformer all looked at it. Transformer warns there are still key ways it could fall apart once shareholders want returns.
The people building this AI wrote, in a legal document, that it could end humanity. Part of that is lawyers protecting the company. But if they know these risks this well, who will make them slow down once Wall Street owns a piece?
