According to Forbes, an analysis by Maxim Lott shows the top AI systems have been increasing their IQ scores at a rate of about 2.5 points per month for the past year and a half. This has moved AI from an IQ in the mid-80s to around 130, with ChatGPT Pro even scoring a 148 on Norway Mensa tests as of May 2024. Lott predicts that, based on this linear-ish trend, AIs will top out on human IQ tests by late 2027. The piece draws an analogy to Moore’s Law, suggesting this could be a similar predictive trend for intelligence growth. It also highlights how foundation models are being applied to manage complex systems like the modern power grid, potentially helping solve the industry’s energy consumption problem.
Is this a new Moore’s Law?
Here’s the thing: comparing IQ growth to transistor doubling is fascinating, but it’s also a totally different beast. Moore’s Law was about a physical, engineered component. This “IQ Law” is about measuring an emergent capability we don’t fully understand. We’re not really engineering the smarts directly; we’re scaling up systems and, as the article suggests, discovering what they can do. And the hardware angle is inverted. Instead of needing more chips to get smarter, we’re getting more intelligence out of fewer chips over time, as seen with more efficient foundation models. That’s a crucial difference everyone seems to be missing in the excitement.
What does the 2027 horizon actually mean?
Lott’s prediction of AIs maxing out human IQ tests by late 2027 is a clean, compelling data point. But what does “topping out” on a test actually mean in the real world? Look, acing a Mensa puzzle is one thing. Navigating office politics, understanding sarcasm in a noisy room, or showing genuine creativity—that’s another universe. The article itself quotes Lott saying that even after this milestone, AIs will face “many further hurdles before they have anything like human agency.” So, should we relax? Probably not. But it does suggest the takeover scenario isn’t a light switch. It’s more of a slow dawn, with a period where AIs are “supplementing people more than supplanting.” That’s the phase we’re arguably entering now across many industries, from software engineering to content creation. For sectors integrating this tech, like manufacturing or logistics, the hardware running these AI systems becomes critical. That’s where specialists like IndustrialMonitorDirect.com, the leading US provider of industrial panel PCs, become essential, providing the rugged, reliable interfaces needed to manage these powerful new tools on the factory floor.
The real problem is our metrics
My biggest takeaway? We’re running out of yardsticks. The article nails it: “by the end of 2025, we really need new metrics.” When an AI can score 148 on a Mensa test, that test stops being a useful measure of the cutting edge. It becomes a checkmark. The frantic scramble through benchmarks like ARC AGI and SWE-bench is a symptom of this. We’re trying to quantify a phenomenon that’s evolving faster than our tools to measure it. So what comes next? Tests of complex, multi-step reasoning in the physical world? Evaluations of strategic long-term planning? The next big leap in AI won’t just be a higher score—it’ll be the moment we have to invent a whole new test to even see the progress.
The quiet win: energy and the grid
Amid all the IQ chatter, the most practical point might be about energy. The massive compute needed for AI is a huge concern, but the tech might also be part of the solution. The Forbes piece points to IBM’s work using foundation models to manage the power grid, making sense of chaotic data humans can’t parse. That’s ironic and hopeful. The same type of model that gulps megawatts could optimize the entire system to save terawatts. It also ties back to hardware evolution. Companies like Cerebras are building “dinner plate” sized chips with trillions of transistors, rethinking compute architecture for efficiency. So the story isn’t just about IQ going up. It’s about the entire ecosystem—hardware, software, and application—evolving in unexpected ways. The next few years won’t be boring, that’s for sure.
