How soon will machines be smarter than humans, really?
August 26, 2025
By Matthew Pietz
This article was written and edited without the use of AI.
AI has been science fiction for so long, many quite reasonably doubt that it could surpass human intelligence anytime soon. Add in tech CEO hype, and the future can seem an impenetrable haze.
But there is a growing consensus among expert forecasters and scientists that Artificial General Intelligence (AGI)—a machine that can do most intellectual tasks just as well as a person—could be five years away.
Here is a well-done summary of the range of expert predictions. https://lnkd.in/g8BNn_2P
There is a lot of much-needed skepticism about AI (for instance, the new best-selling book the AI Con), but it is important to parse the various types of doubt. People persuasively argue that we're in an AI investment bubble, that CEOs have expected too much or tech bros have overpromised, that we shouldn't consider it an automatic savior nor an automatic bringer of the apocalypse, and that it still makes rather silly mistakes.
But all of these claims are separate from the question of whether it is likely that AI will continue to advance until it exceeds human intelligence. At Keranaut we have not seen a convincing argument that anything is likely to fully stop AI progress, though technical challenges and global instability (like possible conflict in Taiwan) could slow it down. Many informed experts predict a 50% chance of AGI by 2030 or 2031. Some predict sooner. And over the last 15 years the consensus date among experts surveyed has gradually moved earlier. In fact, the median prediction jumped forward 13 years between 2022 and 2023.
Stop and think for a moment about what this means. A world full of AI agents as smart as—and, shortly thereafter, smarter than—humans will have seismic effects. Even a 50% chance in five years is staggering.
We tend to think first of our jobs and the economy, which will radically transform, but there will also be huge shifts in education, health care, and governance. We may experience significant anxiety and disorientation along the way, and we may arrive in a world where AIs have more power, and less transparency, than we are comfortable with.
These can be disquieting subjects to ponder, but at Keranaut we feel the time to engage is now. Let’s advocate for more guardrails and human-first development of AI, and let’s thoughtfully prepare for a wide range of disruptions to the way we live now. Let’s navigate the future, together.
