GenAI and job displacement

It goes without saying that generative AI is disruptive and students and recent graduates may often find it difficult to know what to study or how to prepare yourself for the job market.

I don't have a ready-made and simple solution. Also, it is never wise to make predictions especially not about the future (attributed to a Danish politician in the earl 20th century).

But that being said, there is some agreement amongst the experts in the AI and education sphere about the following advices.

1) Learn about the AI technology but with the technology. Learning about genAI is not like learning how to operate a dishwasher - by reading the manual and pressing the right buttons. It may be wiser to see it as learning to communicate with a rapidly developing non-human intelligence. And of course there are many things genAI cannot do or does incorrectly, but the models are growing faster and faster and the race is on. This post's date is November 2025 and ChatGPT3.5 was only released in November 2022. That means that for example medical students who were first years students when ChatGPT3.5 was released are only third years now. Engaging with genAI and learning to communicate with it is good advice, because as they say: "If you are not ate the table, you'll end up on the menu.

2) Understand the human- technology synergy. Reflect and think about what is uniquely human about you and what value does that being human add. In a rapidly changing world, being able to express and use your specific value is essential. Only then will you be able to work with genAI to produce outcomes that are better than either the AI or you could do on its own. The aphorism goes: "You will not be replaced by genAI, but you will be replaced by someone who expertly uses genAI". For students: if you outsource your learning to genAI (have it write your assignments) your learning will be shallow and your future employer can then also easily outsource you to genAI.

3) Use genAI ethically and sensibly. Of course there simply are things that are not done because they are intrinsically wrong, deception, cultural appropriation, bullying, etc. They are your responsibility when you are using genAI and you are accountable for your own actions. AI is a very powerful technology and 'with great power comes great responsibility'. But also, if genAI consumes energy and water, it should be used correctly (shallow prompts produce shallow answers), for the right purposes (if you want a thinking partner use AI, if you want to know a fact look it up elsewhere) and not frivolously.

4) Critically evaluate information. In the past information was expensive, hard to come by and scarce and now information is cheap, easily accessible and abundant. But information is not the same as relevant knowledge. We are surrounded by shallow, incorrect or even deliberate misinformation. That is not unique to genAI, by the way, political spin, advertising, influencers, many documents in the internet and even more traditional media or fora can mistakenly or even deliberately contain erroneous information. But while with other media you can check the provenance of the information or the information owner has (reputational) stakes in the the veracity of the information, with genAI this is much more opaque. So, always reflect on two things. (A) when using my common sense, is this information likely to be true or should I double check it, and (B) how critical is it for me whether the information is correct or not; what is the criticality of the information?

5) In order to do this, well-organised knowledge bases are essential. Isolated mono-disciplinary knowledge, canonical knowledge, is likely to become less valuable (simply because much can be found) but a deep understanding, associations between discipline, learning to see deep structured and have efficient problem representations is becoming more important. For example, in my discipline understanding the relationship between selection into medical school processes and primary prevention (prevalence of the 'disease', sensitivity/specificity of the diagnostic instrument and NNT are important consideration in both) would help me to write much better prompts for genAI to help me solve problems than any mono-disciplinary canonical knowledge could. So, if you are just studying text books or even just use genAI as a chatbot you are only aiming for that mono-disciplinary knowledge, but if you use genAI in different modes and functions to really actively engage with a topic and make connections, you are better preparing yourself for the future.

6) Always consciously manage your agency. Stay in control and be very deliberate about what part of the process you are happy to delegate to genAI. Of course, genAI is not like a calculator where the machine does nothing but only slavishly following your input and it has the affordances of being able to do things automatically. It can query back to you: "Would you like me to write an essay on this for you?" and it can perform quite complicated, multi-step processes (like a literature review). But you need to be able to hand over that agency and therefor still be able to maintain control and assume full accountability for the final result.

These are absolutely no fail-safe suggestions and we don't know what the future will hold, but in preparation for this future, these tips are certainly worthwhile taking on board.