Brynjolfsson, E. and Mitchell, T. (2017). What can machine learning do? Workforce implications. Science, 358(6370), pp.1530-1534.

From Digital Culture & Society

Machine learning is when an automated system is capable of accelerating the pace of automation itself. Within the general academia world, there is little consensus on exactly what machine learning excels at doing but there are two important factors to consider when researching machine learning. First, humans are still far from general artificial intelligence, and secondly machines are unable to complete a full range of tasks that humans are able to do. With the introduction of automation in the workforce, questions have been raised on the application of machine learning in the workplace environment. What Brynjolfsson and Mitchell were looking at was exactly how machine learning will be able to affect the work force as it becomes more widely adopted in society and the job industry. They outlined 8 criteria that they believe distinguish tasks that are suitable for machines to learn and automate from humans from the tasks that only humans are able to do at this point in the technology available. Those 8 criteria that machine learning systems must be able to do are listed as follows.

1. The task has well defined boundaries, inputs and outputs for the machine.
2. The task comes with data sets in order to train the machine to be as accurate as possible.
3. The task provides the system with feed-back and definable goals/metrics.
4. The task does not include long chains of reasoning or common sense problems.
5. The task does not need an explanation for the end result.
6. The task needs a tolerance for error and no need for provably correct solutions.
7. The task needs to be mostly consistent in future analysis.
8. The task doesn’t require specialized dexterity, physical skills or mobility.

These criteria were outlined extremely well by the authors of the article, and tasks that fit the criteria above will continue to be taken over by automation and machine learning, using algorithms to do the job cheaper and more efficiently as the technology because more widespread. The authors had a strong sense of the effects that machine learning would have on the work force, including economic factors listed as well.

• Substitution: Machine learning systems can be used to design other systems in order to replace humans on other tasks.
• Price Elasticity: Machine learning can lower prices for different tasks as the cost of technology is lowered.
• Complimentaries: As the demand for automation increases, so does the demand for more programmers = more work.
• Income elasticity: As users have more money to spend, more money gets spent.
• Elasticity of labour supply: Demand of labour for particular industries. Low skill, low wages. High skill, high wages.
• Business Process Redesign: “…both demand and supply elasticities will tend to be greater in the long run than in the short run as quasi-fixed factors adjust.

This article was very well thought out and laid out, it was easy to read and provided lots of information that previously was unknown. The strengths of this article were the depth of the information that the authors provided – they gave criteria that will allow readers to evaluate their own job positions, and provided the tools to users so they are able to realize if they need to potentially find a new position. Along with those criteria, the economic factors that the authors provided also show the larger picture that machine learning has an impact on. These factors are important to consider because if automation moves too quickly, the economic stability of the job industry would collapse as unemployment would rise. The authors provide positive scenarios in which machine learning has affected economies, and show that the majority of people who’s jobs have been automated have just been redeployed somewhere else for new products, goods or services to be created. Brynjolfsson and Mitchell also explained that it is possible for these machine learning systems to increase the velocity of the changes happening with automation in the job industry. Due to their nature, they are learning everything they can while they are processing data. Eventually, these machines may be able to replace tasks of its own with its own software algorithms if it collects enough information and continues to upgrade with the latest technology.

Megan McGuire