AI Is Not the Enemy. Ignorance Is.
AI Is Not the Enemy. Ignorance Is.
In a recent interview, Fields Medal-winning mathematician Terence Tao discusses the profound changes artificial intelligence is bringing to mathematics. His concerns are legitimate, but I think they also reveal something familiar about technological progress.
https://www.youtube.com/watch?v=bAnwMdfcMVw
Tao's argument
Tao argues that mathematics is moving from an era of proof scarcity to one of proof abundance. AI can now produce mathematical proofs at a rate previously unimaginable.
His principal concerns are:
Proof is not understanding. AI can produce mathematically correct answers without providing the deeper intellectual understanding traditionally developed through solving problems.
Academic research is being disrupted. Problems that might previously have occupied doctoral students for years can now be solved almost overnight, potentially undermining their research and career opportunities.
Commercial incentives are distorting scientific priorities. AI companies are competing to produce impressive results rather than necessarily advancing human understanding.
Mathematics is becoming industrialised. The traditional model of individual mathematicians making discoveries may give way to large-scale, AI-assisted research.
Uncritical dependence is dangerous. If we increasingly rely on AI to make decisions about critical infrastructure and complex systems without understanding its reasoning, mistakes could have serious consequences.
Importantly, Tao is not anti-AI. He recognises its extraordinary capabilities and wants it to augment rather than replace human understanding.
My argument: We've been here before
I suspect mathematicians expressed similar reservations when scientific calculators became commonplace.
Calculators eliminated much of the intellectual labour previously required to perform complex calculations. Computers subsequently automated vastly more sophisticated mathematical operations.
Yet mathematics did not disappear. The tools changed, and so did the skills required to use them.
AI represents another step in that progression.
The crucial distinction is between producing an answer and possessing the expertise to evaluate that answer.
I regularly use AI to assist with academic research and writing. Having completed a PhD, I have sufficient experience in research methodology, argumentation and evaluating evidence to recognise when AI produces fabricated references, unsupported claims or logically defective arguments.
I don't blindly trust its output. I interrogate it.
The same principle applies to mathematics. A mathematician who understands the underlying principles can use AI to accelerate research while retaining the intellectual capacity to evaluate the results.
Someone without that understanding may accept an impressive-looking but fundamentally incorrect answer.
And therein lies the actual danger.
AI is dangerous to people who refuse to develop expertise
The problem is not that AI makes human expertise unnecessary. It is that AI can create the illusion that expertise is unnecessary.
A person who lacks the knowledge to evaluate an AI-generated answer may mistake fluency for accuracy, confidence for competence and plausible reasoning for valid reasoning.
This is particularly dangerous when people are too lazy to develop their own skills, preferring to outsource not merely the work, but their judgement.
There is an important distinction between using technology to eliminate unnecessary labour and using technology as a substitute for thinking.
Scientific calculators did not eliminate the need to understand mathematics. They eliminated the need to perform every calculation manually.
Similarly, AI should not eliminate the need to understand research, logic, mathematics or scientific methodology. It should eliminate some of the mechanical labour involved in applying that understanding.
The bottom line
AI is dangerous to people who are unwilling to upskill, who lack the expertise to evaluate its output, and who consequently trust it blindly.
The appropriate response is not to resist AI, nor to romanticise the intellectual labour it replaces.
It is to ensure that education continues to develop the foundational knowledge, critical thinking and disciplinary expertise necessary to use these increasingly powerful tools intelligently.
AI should make knowledgeable people more productive, not encourage uninformed people to mistake themselves for experts.
The future belongs not simply to people who know how to use AI, but to people who know enough to recognise when AI is wrong.