Have you ever received an email and suspected it was crafted by AI rather than a human touch? Mathematicians have been wrestling with similar concerns for decades, and they have crucial insights to share.
This journey began in 1976 when Kenneth Appel and Wolfgang Haken published their proof of the four-color theorem. This theorem demonstrates that you can color a map with just four colors such that no two adjacent areas are the same shade. Although the theorem seems straightforward, mathematicians anticipated elegant proofs that would uncover deeper mathematical principles. Instead, they encountered 60,000 lines of complex computer code. Appel and Haken programmed a machine to systematically analyze nearly 2,000 different maps representing various configurations.
Initially, there was a feeling that something essential was lacking. However, as the years passed, mathematicians began to accept this approach and effectively resolved many philosophical objections. Consequently, when the latest wave of AI technology emerged, mathematics was ready to embrace it.
As reported, AI is rapidly evolving, pleasantly surprising many mathematicians. While Appel and Haken meticulously wrote the code, we now have large language models capable of verifying whether this code—and thus the proof—holds true, mitigating the risk of hallucinations commonly associated with AI technologies.
“ AI is advancing rapidly and surprising many mathematicians “
Yet outside the realm of mathematics, things are more complicated. The tech industry is rife with stories of AI failures, each carrying varying degrees of consequences. Recently, US research firm Gartner forecasted that half of the companies that replaced human roles with AI will reinstate those positions by year’s end.
While life is not as predictable as mathematics, the lessons learned by mathematicians suggest that AI has potential if we can find practical confidence and philosophical approval in its outcomes. It may take time for other fields to reach this understanding.
Source: www.newscientist.com
