Another prominent AI pioneer expressed regret over his life’s work amid rising concerns over the technology’s risks.
What’s new: Yoshua Bengio, a professor at the Université de Montréal who laid parts of the foundation for deep learning, followed fellow trailblazer Geoffrey Hinton in airing his anxiety publicly. He told BBC that AI’s potential for misuse left him feeling “lost” and questioning the value of his life’s work.
New worries: Bengio said he was afraid that “bad actors” could use AI to cause harm, for instance by developing chemical weapons. In particular, he cited militaries, terrorists, or individuals with personal vendettas.
- Bengio called for governments to register AI developers and govern them similarly to pharmaceutical companies and aircraft manufacturers. He also proposed that computer scientists should be required to undergo ethical training and certification.
- In a recent blog post, he warned of the possibility of rogue AIs that pursue their own goals. The post describes how such machines might be built and how they might cause catastrophic harm.
- Last month, he signed a statement by the Center for AI Safety that urged the world to focus on mitigating the risk that AI could bring about human extinction. In March, he signed the Future of Life Institute’s call for a six-month pause in training models more advanced than OpenAI’s GPT-4.
Behind the news: Bengio is one of the most cited computer scientists in the world. He, Hinton, and Yann LeCun shared the prestigious Turing Award in 2018 for their foundational work in deep learning. His accomplishments include helping to introduce an early attention mechanism for natural language processing and develop the generative adversarial network architecture. In a commentary he wrote for The Batch, he looked forward to neural nets that can reason.
Why it matters: The recent pace of progress in AI has startled even researchers who have spent decades improving the technology, and its potential for harm has taken many by surprise. While there is little doubt that AI poses hazards, debate runs hot around which are most pressing and how to address them. (For instance, Yann LeCun, the third winner of the shared Turing Award, has downplayed some of Bengio’s concerns.) Recognizing the most serious problems is the first step toward devising effective solutions.
We’re thinking: As AI builders, we have an ethical responsibility to minimize the harms our work might bring, even as we work to maximize the benefits. We wish Yoshua Bengio great fulfillment in the next phase of his stellar career.