Yoshua Bengio and LawZero: Why Fear of Future AI Distracts From Today's Threats

Debates about the future of artificial intelligence increasingly revolve around scenarios in which machines one day become intelligent and slip out of human control. The central argument of this essay is different: there is no need to wait for a hypothetical superintelligence, because the key risks of AI have already surfaced in the real economy and in politics. This is not about a machine uprising. It is about concentrated power, surveillance, pressure on workers and the use of technology in the interests of large institutions.

The author states plainly that he does not consider today's AI systems intelligent and sees no grounds for assuming that impressive statistical models will automatically produce full-fledged intelligence. So he treats talk along the lines of "what do we do if AI becomes conscious" either as a convenient distraction or as a marketing frame that lets people discuss spectacular future catastrophes instead of uncomfortable present-day problems. Hence the reference to "Pascal's wager": fear of an unlikely but grandiose event starts to dictate the agenda more forcefully than consequences we can already observe.

The author's position is not unconditional techno-optimism, however. On the contrary, he admits that he shares some of the concerns of so-called AI doomers, but he locates the source of the threat not in the awakening of machine intelligence but in the people and institutions that already control these systems. He worries that digital tools are concentrated in the hands of corporations whose scale makes public oversight difficult. He worries that technology is used against users and workers: to intensify monitoring, automate pressure, tighten productivity metrics and tilt the balance of power further toward employers and the state.

This is where the main conflict with the more futuristic school of AI safety arises. It was triggered by a public discussion in Montreal, where the author appeared alongside Astra Taylor and Yoshua Bengio. Bengio is one of the most respected researchers in deep learning, a Turing Award laureate and a scientist whose work helped shape the current AI boom. He now takes part in the LawZero initiative, which proposes an international consortium and the development of AI as a digital public good: open, transparent, verifiable and safe. For supporters of this approach, long-term risks are serious enough to justify building protective institutions in advance.

The author, however, disputes not so much the idea of taking precautions as the ordering of priorities. In his view, discussion of a hypothetical superintelligence too easily turns into a convenient abstraction, while existing systems for recognition, recommendation, generation and evaluation of labor are already built into mechanisms of control here and now. They influence hiring, productivity, moderation, access to information and relations among government, business and citizens. When power concentrates in the hands of tech giants and the state readily leans on their tools, the problem stops being science fiction: it becomes political, social and a matter of labor.

This view matters also because it changes the language of the debate itself. Instead of "how do we save humanity from an autonomous supermachine," a more down-to-earth set of questions emerges: who owns computing resources, who sets the rules for data access, which audit mechanisms actually work, and who is responsible for harm caused by automated decisions. This is a less spectacular agenda than apocalyptic forecasts, but it is what determines whether technology will deepen inequality or serve the public interest.

That is the practical takeaway. The debate over AI's future matters, but it should not replace the conversation about model transparency, platform accountability, workers' rights and the limits of digital surveillance. If public attention is fixed solely on a rare scenario of future catastrophe, corporations and governments will gain even more room for decisions that make life worse today.

The essay offers a simple but uncomfortable lens: the main question is not whether AI will someday become too smart, but who is using today's AI, in whose interests, and with what consequences for society.