What are we going to do about AI and what’s it going to do about us? Workers and automation (1/2)

With great power comes great unemployment. Or at least, that’s the fear that many have, especially with every new report from a bank [1] or tech company [2], making headlines for predicting the hundreds of millions of jobs that will be lost to automation.

During discussion, our members had mixed views. First, they concluded that the ability of AI to codify skills and then transmit that information far and wide would mean that “know-how” would be rapidly distributed across the globe. But crucially, this isn’t just the spread of familiar ‘know-how’ of the sort we’ve come to use every day, like books, courses and how-to YouTube videos, but also ’know-do’. That might mean packaging years of deep human expertise into an algorithm and then throwing the software into the field to help humans make difficult decisions, or sometimes making decisions for them.

Interpreting patient CT scans is a perfect example. Research teams across the world [3,4,5] are helping to develop algorithms that can spot the signs of cancer more accurately, more consistently and earlier than doctors and other healthcare practitioners working unassisted. And this means earlier, more targeted and effective treatments can be provided, improving patient outcomes and making better use of hospital resources, whilst general-purpose medical assistants accepting mixed data inputs are not far behind [6].

This is great news for both advanced hospitals in developed countries as well as less fortunate areas. Advances like these help to rapidly level the playing field from a healthcare standpoint, increase overall capacity and fill skills gaps in the workforce (and do them faster and more accurately), and ultimately drive down costs.

While this will sound like be good news if you are sick, or buying services, it’s rather less good if you were the person that used to do the job. There is a well-rehearsed argument that this will free up people from menial tasks and provide them with more fulfilling roles. Inter-generationally this might be true, but if you’re a 50 year old radiologist you’re probably a bit miffed at the prospect.

The network members also pointed out that there is a succession and apprenticeship problem. Many of the skilled workers at the pinnacle of their careers – the ones freed up from the menial tasks mentioned above – got that way by working their way up through a system. The repetitive work forms the basis of the training. Not only that, there will be a problem with the vesting process in, say, a legal firm. The monetary benefits of employing large numbers of trainees at low cost and charging out their time at full rates (so you can reap the profit) may be a model open to destructive competitive forces unleashed by AI.

Meaning, we may start to see the pyramids crumble.


[1] https://www.goldmansachs.com/intelligence/pages/generative-ai-could-raise-global-gdp-by-7-percent.html 

[2] https://arxiv.org/abs/2303.10130 

[3] https://www.royalmarsden.nhs.uk/news-and-events/news/ai-could-help-doctors-diagnose-lung-cancer-earlier

[4] https://ascopubs.org/doi/full/10.1200/JCO.22.01345

[5] https://www.fiercebiotech.com/medtech/google-s-cancer-spotting-ai-outperforms-radiologists-reading-lung-ct-scans 

[6] https://www.linkedin.com/posts/vivek-natarajan-a3670118_medicine-is-a-multimodal-discipline-absolutely-activity-7090202293710557184-Vtie/

Leave a Reply