🔮 AI & Math 2.0, Pakistan’s solar hedge & who controls computing++
Exponential View Azeem Azhar
OpenAI’s new math release has people arguing over whether AI will help people understand maths, or just spit out answers. That same fight is now reaching computing, energy, and even Pakistan’s power bills.
Based on reporting by Exponential View, Azeem Azhar — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI’s release of hundreds of mathematical manuscripts and proofs has kicked off a familiar but sharper argument: if machines can solve hard problems quickly, what happens to human understanding? The piece frames this as a possible split between more answers and less comprehension unless AI systems are built to explain what they’re doing, not just hand over proofs.
That tension is already upsetting parts of the field. A group calling itself the Association for Human Mathematics attacked the release and urged mathematicians to stop working with OpenAI. Terence Tao’s answer is less dramatic and more realistic: “Math 1.0” rewarded being first to solve an open problem, but that model has been pushed to exhaustion. In “Math 2.0,” he argues, mathematical progress should be judged more broadly, with exposition, community building, and new directions of study carrying more weight.
The human side of that transition looks messy. In a seminar this week, Fields medallist Hugo Duminil-Copin said his PhD students and postdocs were “in a state of total panic.” His response, like Tao’s, is to double down on the things machines don’t replace cleanly: writing explanations, organizing conferences, and helping colleagues make sense of results.
The same newsletter broadens the lens to computing itself. In a white paper with the World Economic Forum, the argument is that compute is no longer just a technical input; it is economic participation and strategic infrastructure. That leads to practical concerns: a country may own data centers and still be dependent on foreign suppliers for maintenance or licensing, quantum systems will still need conventional computers for control and error correction, and electricity or skills can be the real bottleneck before more compute ever helps.
Then there’s Pakistan, where rooftop solar has already changed how people deal with imported-fuel shocks. Now high petrol prices could push the country’s electric-vehicle target closer than planned. But the transition is uneven. If grid fixed costs get spread over fewer customers, households without panels can end up paying more so that others pay less. Clean power can be a bargain for some and a bill hike for everyone else.
My take — AI-written commentary, not fact-checked reporting
This is the part people keep missing: compute is starting to look less like a cool product category and more like public infrastructure with a Silicon Valley accent. The open-model crowd is right that knowledge should be shareable, but if AI only makes better answer machines, the real losers will be the humans expected to nod along. The smart move is boring and unfashionable: demand explanations, not just outputs, before the whole system gets too clever to trust.
Read more about this at: Exponential View