Faced with Nilay Patel, Microsoft's co-founder believes that competition between laboratories makes self-regulation impossible and that only public authorities can impose safeguards.
As a guest on The Verge's Decoder podcast, Bill Gates extends the alarmed tone of his essay at the end of August on "the turbulent era of AI". He refuses to be labelled a catastrophist but believes that the collective response remains far below the speed of progress: in one year, models have learned to write code, flush out flaws, reason in biology, build relationships with children and shake up employment. He evokes twenty years of upheaval, at the risk of cutting society in two between those who live at the technological frontier and those who reject it.
The immediate danger: models without guardrails
For him, the short-term risk lies not in the choice of one laboratory over another, but in models from which all protections can be removed, which open the door to cyberattacks and bioterrorism. He brushes aside the debate on AGI or the consciousness of machines: the only useful question is whether we have created a tool capable of paralyzing banks or an electricity grid.
On governance, Gates believes that the structures imagined by OpenAI, Anthropic or DeepMind to curb commercial logic could only work with one player at the border. As soon as there are several of them competing with each other, the responsibility lies with the state, as for the automobile or aviation industries, which must rely on the industry's expertise to impose protections on each model. He also insists on the sincerity of the warnings issued by people in the sector, too often mistaken for marketing.
Read the full article on the original source
Go to the source →