The global alarm unleashed by the heads of North American artificial intelligence companies by proclaiming in unison the dangers of their models leaves many questions in the air.
The AI barons advocate that it evolve more slowly and more regulated while they step on the accelerator to win the race of its development. They continually release insufficiently tested products that are increasingly more powerful… and more expensive.
Sam Altman, CEO of OpenAI, has warned about threats from AI since his company made a splash in these technologies four years ago. But it has not stopped turning a consortium focused on research into a multinational with an unbridled appetite for investments and profits. Additionally, his company was slow to disclose or outright concealed attacks on its applications from various entities.
The other great lord of AI, Dario Amodei, CEO of Anthropic, has been even more critical than Altman with his own business, although the first alerts were raised by one of his models, Fable 5. It was users who discovered that it was as powerful as it was insufficient in security barriers. The North American Government suspended it in June until it was rescheduled.
Anthropic has released a 154-page report on security incidents with its agents, including the production of biological weapons. They started in December and didn’t meet until last week.
Along with these contradictions, AI laboratories seem interested in a framework that gives legal approval to their products not only to defend humanity but to protect themselves from possible lawsuits. And it would also help business if regulation relaxes the rules on competition and they can agree to spend less to outdo each other.