The idea in a nutshell : The unemployment rate seems set to be affected by artificial intelligence in a way which is not being discussed – even in the run up to current Australian election. The ramifications of AI on employment appear substantial although they’re hard to quantify. This suggests the government has it’s head in the sand and will sufer the full costs of high unemployment without getting the benefit of the tax which might be on offer.
Why aren’t they telling us the truth about unemployment ?
One open question for me is whether the government (and others in Australia) have a clear view of some of the threats we face to employment. We’ve just, as I write this, undergone the release of the 2016 Budget. The chart below was not taken from that budget but is representative of the information which is available at the moment about unemployment.
Below : Forecast Australian Unemployment. 2010 – 2020 with 2016-20 as forecast.
From my point of view, the specifics statistics involved are not important to the point I’m trying to make. These forecasts suggest things are going to continue in the near future much as they have done for most of the recent past. There is a lot of evidence suggests this is not the case.
As I said in a recent blog entry, there’s a lot of ‘early’ activity in the public domain – AI is already here. Considering the economics / maths of the circumstance, this trickle may be set to turn in to a flood sooner than we think.
I am sure it comes as no surprise that there are forecasts other than those official out there. To be fair, the timeframes involved in these estimates don’t overlap with the period of consideration covered by the forecast from above. But :
Evidence of forecast increases in unemployment as a result of AI
Forecasts for job cuts vary. They tend to suggest 30%-50% of jobs will be performed by AI in the future.
- McKinsey : McKinsey – the consulting organisation One in 5 companies will go out of business as a result of AI in the next 5 years say.
- The ABC : The ABC were the first I saw to release a forecast of the number of jobs that AI would see go. They predict 40% of Australian jobs will disappear by 2025.
- Oxford University : 35% of UK jobs or 47% of jobs in the USA will be automated (the USA is affected more because it has a higher proportion of workers employed in manufacturing) as reported in the Guardian newspaper.
- The House Of Lords : The UK’s House Of Lord’s Digital Futures report suggested that 35% of jobs will be automated in the next 2 decades ( it was published in 2015)
- The CSIRO : The CSIRO have also written on the subject although they were more responsible in what they reported.
I think one of the components of this work which some find surprising is that it won’t just be basic jobs which go.
The way I think about it is that any job which follows a pre-established pattern will disappear. If it’s a job which follows a process, for example, it seems to be highly at risk. This applies to those who work in call centres of course. But it also applied to Doctors and teaching assistants. It applies to many types of economists and dentists as well as airline pilots and
It’s not quite as clear cut as suggesting that since 35% of people are made unemployed, there will be 35% fewer people left not working.
But I grew up in the North of England, in a town called Worksop. I still remember the strikes by desperate miners which went on for months. The mines were being closed by Margret Thatcher and many were displaced, never to work again.
The classic economic argument is that these new technological innovations raise productivity for those who remain in the workforce. Take the Doctor for example. His patients are unlikely to want to be diagnosed by a robot arm. With the assistance of AI doctor software he will have more accurate diagnoses and be able to spend longer with each patient.
Jobs which are more likely to be safe
I think the truth is that no one know which jobs will be safe in a world of AI. If swathes of people are made unemployed, those on the fringe of displaced markets with transferable skills will seek employment in the alternatives. Take a taxi driver who can also speak Arabic ( a taxi I took recently had just an individual behind the wheel. ) His taxi driving role is displaced by autonomous cars so he applies for jobs in the labour market for translating Arabic to English, pushing down prices in that market. ( i.e, lowering wages for translators. )
- Originality / Intuition : Jobs which require a mental leap beyond what’s seen as possible are likely to be safer than others. Idea generation. Humans have the upper hand on this for now. Afterall, even Watson, the grandfather of machine intelligence solutions falls short in it’s ability to infer new ideas or answers.
- Moving unusually shaped things in unstructured environments: So rugby players, bar / pub workers and soldiers should be OK.
- Healthcare : There are going to be so many more old people that healthcare will become increasingly important. And again, these workers tend to perform their tasks in complex unstructured environments – like old folks’ homes.
Conversely, jobs which are less likely to be safe
The Committee for the Economic Development of Australia suggests that “low levels of social interaction, low levels of creativity, or low levels of mobility and dexterity” will be most at risk.
Summing up – the unemployment rates are likely not as forecast
This situation is far more complex than it seems. Just because these jobs are going doesn’t mean the displaced people will not be re-employed. The productivity from AI being deployed could have an enormous and positive effect on the economy.
One of the upshots of all this is that it appears the new wave of jobs might disproportionally affect women. Empathy and burgeoning opportunity in healthcare does, at a headline level, appear to favour jobs for women. This is a bit ironic given the low levels of inclusion of women in growing areas of the economy – STEM.
The greater concern, overall is that the government is struggling to keep up with these changes in the economy. It legislates too slowly, as we’ve seen by the head in the sand way they’ve dealt with the regulation surrounding Uber.
If they can’t get that right, adapting mechanisms like education is never going to happen in the timeframes required for those changes to be meaningful. The balancing act for the future would seem to be managing the economic benefits released by AI alongside the potentially higher unemployment levels. Unfortunately, if the mining tax is anything to go by, government will be out manoeuvered and will squander the opportunity in full face of the costs.