After Geoffrey Hinton had said that he believed that AIs could be designed to be non-threatening towards humans by instilling a maternal instinct towards them, another AI godfather has said that he’s optimistic that superintelligence can be controlled.
Yoshua Bengio, the Turing Award-winning computer scientist and pioneer of deep learning, has emerged with a surprisingly hopeful message about artificial superintelligence—a stark shift from the sleepless nights that plagued him just two years ago. Speaking recently, Bengio outlined a technical path forward that has not only restored his optimism but also prompted him to launch a new nonprofit organization dedicated to building fundamentally different kinds of AI systems. His vision centers on creating machines that lack self-interest, goals, and agency—more akin to “really smart encyclopedias” than autonomous agents.
“I’m actually more optimistic about our ability to control superintelligence than I was, say a couple of years ago,” Bengio explained. “So a couple of years ago, I didn’t sleep well and I thought I need to do something about it. Can we fix it? Can we make it not harm people? And so I’ve been working on this, and last year I realized that there should be a way forward that I’m going to say a few words about.”
The core of Bengio’s optimism lies in a fundamental reconceptualization of how advanced AI should be built. “The idea is that we can build machines which are not like us, which don’t have a self, which don’t have a goal, but which understand the world and can make very good predictions in it,” he explained. Drawing an analogy to the natural world, he added: “If you think a little bit about it, the laws of physics don’t care about you. They don’t care about me. They don’t care about being elected or having more compute power.”
“And we can build AI in the future—I’m convinced, and I’ve started to assemble a team of researchers and engineers to do that—that will be just like this, which will not be a new species, which will be like a really smart encyclopedia if you want, that can help us in scientific discovery, that can help us solve even, you know, design better tools for our democratic institutions, which will be sorely needed,” Bengio continued. “And that even can help with the untrusted agents that companies will want to put on the market.”
Bengio crucially wants to build AI systems without a Self. If AI systems aren’t conscious and have no feelings, emotions, wants or desires, he seems to be arguing that they will have no will of their own to seek to dominate humans or desire more power. They will exist merely as tools, and do what humans ask them to do.
Bengio’s shift in perspective comes at a pivotal moment in the AI safety debate. While his fellow “godfathers of AI”—Geoffrey Hinton and Yann LeCun—have taken different stances on AI risk, with Hinton leaving Google to warn about existential dangers and LeCun arguing that current concerns are overblown, Bengio appears to be charting a middle path: acknowledging the risks while actively building technical solutions. His approach of creating “tool AI” rather than “agent AI” aligns with growing calls from researchers for more controllable AI architectures. If Bengio’s vision succeeds, it could offer a template for how humanity might reap the benefits of superintelligent systems—scientific breakthroughs, improved institutions, oversight of commercial AI—without creating entities whose goals might diverge catastrophically from our own.