“Don’t rely on it too much.”
“It’s going to make us lazy.”
“What happens to our brains if it breaks?”
If you have scrolled through LinkedIn, tuned into a tech podcast, or sat in a strategy meeting recently, you have undoubtedly heard these warnings about artificial intelligence. Critics worry that generative AI will erode our critical thinking, kill our writing skills, and leave us helpless if the servers ever go dark.
If you are skeptical about AI, these fears are completely valid. It feels like we are handing over our minds to machines.
But if this panic feels familiar, that is because we have seen this exact movie before.
Thirty years ago, society said the exact same things about a new, confusing tool called the World Wide Web. Today, we do not just use the internet. We depend on it to run our global economy.
Here is what history reveals about this pattern, backed by real data, and where it ultimately leaves us.
1. The Death of Memory vs. The Death of Creativity
When search engines first emerged, critics warned they would ruin human intelligence. The argument was simple. If you can look up any fact in five seconds, you will stop memorizing information, destroying your brain’s capacity to retain knowledge.
Today, the anxiety has shifted from retrieval to creation. The fear is that because AI can draft an email, write code, or design a graphic in seconds, human creativity will wither away.
The Reality: The internet did not destroy our brains. It just offloaded how we store data. A foundational 2011 study published in Science by Columbia University researchers documented the “Google Effect,” showing that when people know information is saved externally, they are less likely to remember the fact itself, but far better at remembering where to find it. It freed up cognitive bandwidth. Instead of spending hours digging through library card catalogs to find data, we learned how to synthesize data.
AI is poised to do the same for the mechanical side of creation. Instead of spending hours writing boilerplate code or drafting routine templates, we can focus on higher-level strategy, deep problem-solving, and unique human insights.
2. From “Nice-to-Have Tool” to Infrastructure
In the mid-1990s, the internet was widely viewed as an optional hobby for tech enthusiasts. Fast forward to today. If a major cloud provider experiences an outage, global banking stalls, airlines ground flights, and remote work grinds to a halt. We are entirely, unapologetically dependent on the internet.
AI is currently transitioning through this exact same pipeline. Right now, many view it as a neat assistant, perhaps a chatbot that polishes text, summarizes a PDF, or autocompletes a line of code. But as AI integrates into cybersecurity, supply chain logistics, and project management platforms, it will shift from a tool into invisible infrastructure.
Dependency is not inherently a failure. It is the natural byproduct of technological evolution. Historical economic data from the Federal Reserve Bank of San Francisco shows that the massive surge in U.S. Total Factor Productivity in the late 1990s and early 2000s was directly driven by businesses reorganizing themselves entirely around information technology. We did not just use the internet; we restructured our companies around it.
We stopped navigating by the stars when GPS arrived, and we stopped using phonebooks when Google did.
3. The Definition of “Basic Skills” Is Changing Again
Every massive technological shift redefines what it means to be skilled.
- Pre-Internet: High value was placed on rote memorization, manual filing, and knowing physical lookup systems.
- Post-Internet: High value shifted to information literacy, digital research, and speed of communication.
- The AI Era: High value is moving toward prompt synthesis, auditing AI output, and judging which tasks should stay human.
We see this exact pattern in how education handles math. As noted by the American Historical Association in their commentary on the “calculator problem,” early math educators panicked that calculators would ruin basic math skills. Instead, math education adapted. Manual arithmetic was kept to build conceptual understanding, but calculators were eventually required so students could focus on higher-level algebra and calculus.
Knowing how to write a basic piece of code from scratch or format a standard report will become less of a day-to-day task. Knowing how to direct an AI to build it, troubleshoot its errors, and audit its output for ethical and factual accuracy will become the definitive skill of the modern workforce. There is a catch, though: you cannot reliably audit code you could not have written yourself. That is where this story gets more complicated.
4. The Real Test: When the Offloaded Task Is the Reasoning Itself
Every example so far has one thing in common: the internet and the calculator offloaded retrieval and computation, not judgment. You could still forget a phone number and still reason through the math problem. The panic turned out to be misplaced because the actual thinking never left the room.
That is not always true of how people use AI today, and the clearest live example is coming out of the University of Chicago Law School, one of the top law programs in the world. In a July 2026 policy statement, the school announced it would ban laptops, tablets, and phones in its first-year core courses, including Constitutional Law, Criminal Law, and Contracts. Their stated reasoning: AI lets students “produce easy answers but stunt intellectual growth,” and first-year students are, in the school’s own words, at their “nadir” for judging whether an AI-generated answer is actually correct. The concern is not that students are looking things up. It is that they are skipping the effortful struggle that legal reasoning is supposed to build in the first place.
As developers, we run into the exact same split every day, just with code instead of case law. Asking AI to recall a function signature, generate boilerplate, or format a config file is retrieval. It is no different from searching for a syntax reference instead of memorizing it, and it frees up real attention for the parts of the job that actually require judgment. But asking AI to make an architecture decision, diagnose a production bug, or write test logic you do not personally understand is a different category entirely. That is not retrieval, that is outsourcing the reasoning, and it produces the same failure mode UChicago is describing: you ship something that looks correct, without the expertise to know whether it actually is.
What UChicago did next is the part worth paying attention to. They did not ban AI outright. Legal research and writing courses still use it, just after students first write without it. Upper-level papers now require an oral defense, not because AI-written work is automatically wrong, but because the only way to confirm someone actually understands what they submitted is to make them explain it under questioning. That is intentional adaptation in practice, not avoidance, and it is the same standard worth holding any AI-assisted work to, legal brief or production code: use it to move faster on the parts that are genuinely retrieval, and keep your own judgment in the loop on the parts that are not.
What History Tells Us
History teaches us a clear lesson, with one condition attached. Trying to stop society from depending on a tool that multiplies efficiency is a losing battle. Human progress moves relentlessly toward tools that save time, and every foundational shift, from the printing press to the calculator to the smartphone, was met with intense panic that humans would lose their core abilities.
But humanity did not become obsolete in any of those cases by accident. Math education kept manual arithmetic in the curriculum on purpose, specifically so calculators would not replace conceptual understanding. UChicago Law is restructuring its classrooms on purpose, specifically so AI would not replace the effortful struggle that builds legal reasoning. The baseline only rises when the transition is managed deliberately, not when a new tool is simply left to spread unsupervised. We did not stop thinking. We changed what we spend our time thinking about, but only because someone made sure the thinking part did not get skipped.
Our Core Takeaway: Navigating the Shift
Let’s be completely honest: we should not blindly trust AI or ignore its clear risks. Instead, our perspective is simple: avoidance is not a strategy, but intentional adaptation is.
In practice, that means asking one question before handing anything to AI: is this task retrieval, or is it reasoning? Formatting, boilerplate, routine drafting, looking something up: hand it over and take the time back. Judgment calls, architecture decisions, anything you would need to defend under questioning: keep those yours, and use AI to move faster on the work around the thinking, not to replace the thinking itself.
The goal should not be to avoid using AI out of fear of losing our edge. The path forward is to build a healthy, deliberate partnership with it, the same kind UChicago Law is now building directly into its curriculum. By using automation to handle predictable, repetitive tasks, we free up human teams to maximize their strategic, creative, and collaborative potential.
The internet did not replace human workers. It empowered the workers who knew how to use it. AI will be no different, as long as the humans using it keep their own reasoning in the loop.