Monday, 13 November 2017

The Productivity Problem and Machine Learning

One of the things that few mainstream politicians will admit to is the fact that Economic Policies which promote Productivity, lead to short to Medium Term Unemployment. Likewise the converse is true. If a government promotes policeies for High Employment, they tend to undermine Productivity.

Longer term, productivity has benefits all of its own. Long term productivity growth leads to more investment and hence higher employment. Along with this comes better quality jobs demanding higher skills and hence wage growth.

North America has long enjoyed high growth, largely due to the abundance of land and resources and a relative low density in population, leading to skills shortages and therefore the need for higher productivity (gained from investment in automation technologies). This puts Britain's recent economic performance in perspective. The UK economy largely weathered the most recent economic downturn well. Though one of the consequences was large scale imigration and a drop off in productivity growth.

This means that British Industry needs to invest in significant increases in productivity, if it wishes to maintain its long term competitive position. Also, it poses challenges as Britain has shifted to a services dominated economy, as it is often difficult to achieve high productivity within service industries and traditional approaches such as work flow (or Business Process Management) and rules engines tend to be limited in applicability and affordability. Typical problems are activities which deal with subjective judgements involved in tasks such as categorisation or recognition where people are able to deal with inconsistencies and sometimes incompleteness of data. These steps are usually not value adding in themselves, but important to being able to carry out subsequent value adding tasks.

A simple example would be in pest control. when on a job, the pest controller may find a cockroach during an inspection which indicates a probable infestation problem. Different strains of cockroach however need to be treated with different chemicals to eradicate them. Often samples need to be sent to a lab to be examined and identified, so that the operative can then return and take appropriate action. This adds time, delay and effort to the process with impact on costs and productivity. It also may impact the customer, especially if his or her business is in catering or another business which may have to close down until the infestation is cleared up. Use of picture recognition technology, based on training a machine learning system and providing field access via, say, a mobile app would enable a much quicker and cheaper response. As on the spot diagnosis could take place, leading to instantaneous treatment. The avoidance of going away to come back again, takes a couple of steps out of the process.

Other examples occur in the legal profession, where initial analysis of case files may help decide whether a case is worth pursuing based on existence or not of a number of attributes which relate to previous experience and case law. This allows lawyers and clients to focus on valuable activity rather than value destroying cases.

So, it is strange to see that some politicians want to tax such applications as people, in order to protect jobs. As inevitably this will erode productivity growth and damage long term employment opportunities. Are they trying to keep everyone poor?

No comments:

Post a Comment