Have you ever read an email and wondered if it was generated by AI rather than carefully written by a person? Mathematicians have faced a similar dilemma for decades—and their experience offers useful insights for everyone.

Figure 1. Machine Mathematicians Reveal Why AI Can Be Trusted.

The story dates back to 1976, when Kenneth Appel and Wolfgang Haken unveiled a proof of the four colour theorem, which states that no more than four colours are needed to shade any map so that adjacent regions differ. Because the theorem is so simple, many expected a neat, elegant proof. Instead, the solution relied on 60,000 lines of complex computer code. Their approach used a machine to systematically analyze nearly 2,000 map configurations, covering every possible case. Figure 1 shows Machine Mathematicians Reveal Why AI Can Be Trusted.

Initially, this method felt unsatisfying to many. Over time, however, mathematicians gradually accepted computer-assisted proofs and worked through the philosophical concerns. As a result, when modern AI tools emerged, the field of mathematics was already well prepared to embrace them.

As we report here, AI is advancing so quickly that it’s catching many mathematicians off guard. While Kenneth Appel and Wolfgang Haken once had to write their own code by hand, today large language models can take on that role, with other software verifying the results—and, by extension, the proof itself. This layered approach helps eliminate concerns about hallucinations, where AI might fabricate information, because robust systems are already in place to filter out errors and confirm what’s valid.

Outside mathematics, things are far less clear-cut. The tech press is full of cautionary tales about AI-generated code going wrong—sometimes with serious consequences. Meanwhile, Gartner has predicted that half of companies that replace workers with AI will end up rehiring people for the same roles within a year.

Of course, the real world isn’t as controlled as mathematics. Still, mathematicians have shown that AI can be valuable—provided we’re both practically confident in its results and philosophically comfortable relying on them. It may just take time for the rest of society to reach that point.

Source: New Scientist

Cite this article:

Priyadharshini S (2026), Machine Mathematicians Show Why We Can Be Comfortable with AI, AnaTechMaz, pp.947