„Google DeepMind“ leader Demis Hassabis, also a Nobel Prize laureate for work in protein science, stated in his article that AI capable of matching the full spectrum of human thought, also called general AI, will most likely be created within a few years.
Read more A foreign bank card held in a woman’s hands
D. Hassabis leads the division where the „Gemini“ AI models were created and which directly competes with „OpenAI“ and „Anthropic“ products, as companies race to develop more powerful systems.
D. Hassabis proposes establishing an AI oversight agency at a time when calls from competitors for mandatory government regulation of this technology are growing louder. D. Hassabis suggests creating a new regulatory body modeled after the Financial Industry Regulatory Authority (FINRA). FINRA is a private oversight organization that supervises Wall Street brokerage firms, with federal oversight. According to D. Hassabis, the new agency would be staffed by a team of leading technical experts, primarily funded by AI industry money, and its task would be to develop safety standards that advanced models must meet before entering the market.
Under this plan, AI developers would initially voluntarily submit their most powerful models for review no later than 30 days before release. Once the testing system proves reliable, compliance would become mandatory.
Read more One mistake can cost a sickness benefit: what you need to know about incapacity for work
„No one in the world really knows what will happen next, and even experts disagree on this,“ writes D. Hassabis in his article. „When there is great uncertainty and the risk is so high, acting with cautious optimism is a wise and correct strategy.“
The new US agency’s rules would apply to any model deemed powerful enough to meet the „borderline class“ criteria – regardless of whether it was developed in the US or abroad. D. Hassabis hopes that international market participants will join the system once it is implemented and operational.
