Harvard experts on why AI has not triggered mass layoffs yet

[Cobra]

The Harvard Gazette has a thoughtful piece asking why AI has not led to mass layoffs, despite all the warnings. It brings together economists and business school professors including Doug Elmendorf, Joseph Fuller, Raffaella Sadun, Karen Dynan and Brookings' Louise Sheiner. Unemployment is still around 4.1 percent in the US, even though estimates suggest 41 percent of work tasks could be automated with today's technology. So why the gap?

One big reason is that companies are struggling to make AI work. Fuller found that only about a third of corporate AI experiments succeed, partly because firms are not training staff properly. In his words, employees are using it like a glorified Google search, which does not produce meaningful productivity gains. It is hard to replace workers with a tool your organisation does not know how to use

Sadun makes a different point about tacit knowledge. Experienced workers carry a lot of company specific know how that is not written down anywhere, and automating their roles can save money in the short term while damaging the business in the long run. Some firms are also redesigning entry level jobs to combine technical work with exposure to sales and product teams, creating development paths rather than cutting roles

The data is mixed though. Accenture found that 52 percent of executives expect AI to increase entry level hiring, but entry level job postings have fallen by 73 percent over four years. Those two figures do not sit comfortably together, and the second is the one that worries me more for young people starting out

Elmendorf thinks significant job losses from AI are likely within the next two decades, but that most displaced workers will find new jobs. His modelling suggests around 2 percent of the workforce could be displaced at once, which would be larger than the so-called China shock that hit manufacturing towns. That ties in with the UK AI minister saying this week that Britain needs a plan for unprecedented job losses. Are we in a calm before the storm, or will AI end up changing jobs rather than destroying them?

Why AI hasn't triggered mass layoffs - yet - Harvard Gazettenews.harvard.edu

[Cesaro27]

One third of AI projects succeeding is about the same as most IT projects historically. Big technology changes always take longer than the hype suggests

Electricity took decades to reshape factories. AI will probably be similar, just faster

[Nina81]

A 73 percent fall in entry level postings is the number everyone should focus on. Layoffs are visible and make headlines. Jobs that never get created are invisible, but they hurt just as much. Graduates are feeling this already

[Teal Shannon]

Calm before the storm feels right to me. The tools are improving fast, and companies are learning how to use them. Give it three or four years and the picture could look very different. Governments should be planning now, not waiting for the unemployment figures to jump

[Runtime Gareth]

Glorified Google is exactly how most people at my work use AI. Nobody has had any training

They ask it questions and copy the answer. That is not going to transform anyone's productivity. Training budgets are always the first thing cut

[RatedRStu10]

Redesigning entry level roles to include sales and product exposure is a smart idea. Young people learn faster when they see the whole business. If AI takes the routine tasks, what is left can be more interesting and more valuable. That is the optimistic version at least

Whether most companies bother to do it is another question. It takes effort and planning

[NightCrawler33]

The Accenture figure sounds like executives saying what they think people want to hear. Actions speak louder than surveys

If they expect to hire more juniors, why are postings collapsing? Something does not add up

[SchrodingersCat52]

Unemployment at 4.1 percent tells you very little on its own. People can be employed but in worse jobs, fewer hours or lower pay. Those shifts matter more than the headline rate

[DeBruyne93]

Two percent of the workforce displaced at once is bigger than it sounds. That is millions of people in the US and hundreds of thousands in the UK. Concentrated in certain industries or regions, it could be devastating locally. The China shock showed how long those scars last. Nobody should shrug it off

[SpikeDudley88]

Tacit knowledge is a strong argument. The person who knows which supplier to call when things go wrong, or why a process exists, is hard to replace. Companies that cut those people often regret it within a year. I have seen it happen more than once