When AI does the work of juniors, who will grow the experts?
While the debate on AI often focuses on job displacement, a quieter issue is emerging: if artificial intelligence performs the tasks traditionally used for training, how will beginners gain the expertise needed to become future seniors?
Most of the discussion about artificial intelligence and the labor market focuses on the question of who will lose their job. This is an important question, but alongside it, a quieter problem is developing: if artificial intelligence performs more and more of the tasks that employees train on at the beginning of their careers, how will they become experts?
The problem is not necessarily that beginners will not have work. On the contrary, it is possible that young people will have a technological orientation that will allow them to integrate well into a world where they work alongside artificial intelligence. The question is what work they will do, and whether it will still be work that allows them to learn the profession. This is true for almost every knowledge-based profession:
-
A lawyer at the beginning of their career reads documents, searches for case law, and prepares drafts.
-
A programmer writes code, fixes bugs, and learns from existing systems.
-
An economist or analyst collects data, builds tables, and prepares initial analysis.
These are exactly the tasks in which artificial intelligence is becoming more efficient. However, these tasks were never just a way to produce output; they were also a way to learn. The lawyer wrote dozens of drafts until they learned to identify a good argument; the economist built models again and again until they developed an intuition regarding a number that does not add up; the programmer solved hundreds of bugs until they learned to identify where the problem really lies. Expertise is built from repetition, mistakes, corrections, and exposure to many cases.
If the machine does an increasing part of this path for us, we need to ask how we will accumulate the experience it replaces. Here, a market failure may arise. As artificial intelligence allows experts to perform more of the basic work themselves, a single employer has less incentive to invest in junior employees. But what will happen when, in a decade, those same employers still want people with ten years of experience? Where will that experience come from?
In a study by Prof. Einat Albin from the Hebrew University, a version of the same question already arose among high-tech companies: if fewer juniors are needed today, where will tomorrow's seniors come from? There is no simple solution. But it is clear that training for the world of artificial intelligence cannot be limited to teaching young people to use new tools. Precisely as the machine does more, it is important that the person understands the building blocks of the profession well, so that they can criticize it, identify mistakes, and know when not to accept the answer it offers.
At the Faculty of Law at the Hebrew University, this is the direction we are trying to move in: alongside providing tools for working with artificial intelligence, we continue to teach deep reading of court rulings, building an argument, identifying difficulty, and independent legal thinking. The goal is not to choose between traditional professional training and AI, but to understand that in a world of AI, we will need both. The academy alone will not be able to solve the problem. Employers will also need to rethink how they allow beginners to gain experience, and perhaps even intentionally invest in tasks that are no longer necessary for output, but are still vital for training.
Much of the discussion on artificial intelligence deals with the question of what work it will save us. The harder question is what we stop learning when it does the work for us.
Dr. Gadi Perl is the research coordinator at the Federmann Cyber Security Research Center at the Hebrew University. His doctoral dissertation deals with the regulation of artificial intelligence.
