top of page

AI x JOB DISPLACEMENT

  • Writer: Jeremy Connell-Waite
    Jeremy Connell-Waite
  • Jun 1
  • 2 min read

Updated: Jun 3


How many jobs will AI really take? It might not be as many as you think…


You don’t need to look very far to find some scary stats about AI and job displacement. A few years ago the UK's CIPD said “9 out of 10 jobs will need to reskill by 2030”. 🤯


The World Economic Forum thought it might be closer to 50%, but a new study just released shows even the more conservative estimates might be too high.


Why?


Michelle Yin from Northwestern University co-authored a report for the National Bureau of Economic Research, which studied how accurate some of the job displacement stats actually are.


Hypothesis: “Every major forecast about which jobs AI will eliminate comes from asking AI to rate itself. We found the answer depends entirely on which AI you ask. If the measurement is unstable, the policies built on it are too.”



To test this Michelle took all 705 jobs (as categorised by the US occupational job coding scheme) and used different LLM’s to make a judgement on job displacement.


There was a huge amount of disagreement.


eg. According to OpenAI’s GPT-4 economists were 10% exposed to job displacement. GPT-5 thought 50% of economists were at risk. Claude thinks it is over 80%.


(Stating the obvious, but it’s only when different models appear to agree and tell the same story that results tend to be accurate).


⬇️


TAKEAWAY: Don’t believe all the click-baity headlines and stats you read on Linkedin or in the news.


The truth, as always, is usually far more complicated - and it gets increasingly messier (and biased) when you only ask your favourite LLM to help you with the answers.


If you’re interested in the future of work, and some solid studies around job displacement and AI-bias within research, here’s a few useful links:



--


An interesting take from Andrew Ng:



bottom of page