Yoshua Bengio urges Europe to focus on reliable and safe AI

In short: AI researcher Yoshua Bengio says Europe should compete in AI by focusing on systems that are reliable, safe, and secure.

What's going on

Yoshua Bengio, a professor of computer science and a well known AI researcher, argues that Europe is unlikely to beat the US and China by simply trying to build the biggest AI models. Instead, he says Europe should aim to lead in AI that can be trusted, especially in areas where mistakes are costly.

He says many companies have hit an “adoption wall”, meaning they are not rolling AI out widely, even after trying it. One reason is that modern AI can be hard to predict. It can give answers that sound confident but are wrong, sometimes called “hallucinations” (like a student making up an answer that sounds plausible). Bengio says this is a serious problem for safety critical fields like healthcare, finance, energy, and aerospace.

Bengio points to research and surveys suggesting that while many firms report using AI in at least one place, very few believe their AI plans are truly mature. He argues that without better ways to verify AI behavior, many businesses will stay cautious.

What to watch

Bengio calls for Europe to fund long term research focused on AI reliability, including a dedicated research institution and large, multi year budgets. He also suggests Europe should work with other “middle powers” to share computing resources, data, and talent, and he cites the Germany and Canada “Sovereign Technology Alliance” as one example. The key question is whether European governments will put major funding behind AI that can be checked and depended on, not just AI that is powerful.

Source: Financial Times