The Dangers of Outsourcing Your Brain to AI

AI is brilliant for automating things you already know how to do, or to create things based on content you’ve already mastered. But there’s an inherent risk in it. Human brains learn things by reading, processing, and doing, and outsourcing critical components of that process to AI is a disaster in the making, not by being great at writing prompts and distributing their downstream output.

Let’s take an obvious somewhat extreme example. A clever sixth grader with a vague interest in American history could create a somewhat sophisticated analysis of a thorny constitutional law question – let’s say, an analysis of the Presidential immunity decision by the Supreme Court a couple years ago against the background of our own Constitution and jurisprudence and with a nod to the history of Common Law. The sixth grader could start by asking Claude to draft the prompt and just tell it “make the question sound sophisticated, as if it was written by a second year law student, and ask the model to write a 20 page term paper complete with citations and to check its own work against multiple sources.” Then the student could paste the prompt back into the model, and voila – output. Some readers/graders will know it’s not right, but others won’t and will give the student a passing grade, possibly even a good grade, and think “wow that 12 year old is brilliant!”

But we all know that the 12 year old is not brilliant, and in fact, has not learned a thing by using AI this way – because the learning is in the reading, the processing, and the doing. And what’s happened? The 12 year is old is given at least a good-enough grade in the moment and is passed on to a more advanced level of studies. Danger, danger, danger.

Compare that to asking a Constitutional law professor to do the same thing. The professor has such a deep mastery of the content that she *could* do the work on her own, but is outsourcing the work to Claude, perhaps (ideally!) reading and editing and perfecting the model’s output. No harm, no foul.

We have to be discerning in how we use AI based on who is using it.

Let’s move the analogy to the world of work. An entry level employee uses AI to create a project plan by feeding it inputs he collects from the organization and giving the model instructions instead of creating the plan by interviewing stakeholders, understanding business process, etc. The model may or may not get the output right, but the employee won’t have internalized how the project will get done and what the dependencies and risks are, and he probably won’t even be able to answer questions about it. The senior employee doing the same thing gets quicker work done that she could have done in her sleep.

As I wrote about in AI won’t necessarily take your job, but someone who uses it will:

Someday, you’re going to need a batch of new not-junior people. And the way you get those people is by training junior people. Junior people have to learn by doing — not only by being clever at writing prompts. You can’t skip the apprenticeship phase of a career and expect to have seasoned leaders in five years.

Smart companies are figuring out how AI can enhance the learning process for junior people. Companies who ask a junior person to create automations by first understanding and documenting a workflow, stress testing it with human conversations, and THEN vibe coding an application to automate or support or transform that workflow are doing great things for their own development and for the organization as a whole.

If you allow your junior people (or your kids) to outsource their learning to AI, you’ll be in bad shape down the road.