Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction

JEP paper creates framework for thinking about AI’s impact on labor market, how much of a job is driven by “prediction,” “…Machine learning does not represent an increase in artificial general intelligence of the kind that could substitute machines for all aspects of humancognition but rather one particular aspect of intelligence: prediction....We define prediction in the statistical sense of using existing data to fill in missing information. As deep-learning pioneer Geoffrey Hinton (2016) said, “Take any old problem where you have to predict something and you have a lot of data, and deep learning is probably going to make it work better than the existing techniques.”...”
Implication 1, “…First, artificial intelligence may directly substitute capital for labor in prediction tasks. Some tasks, like demand forecasting, are already prediction tasks. Where humans currently perform these prediction tasks, they are increasingly replaced by artificial intelligence. At the same time, other tasks that were not historically viewed as prediction tasks are being transformed into prediction-oriented tasks as machine learning improves and the quality-adjusted cost of prediction decreases. Many parts of the workflow in human resources are being broken down into prediction tasks so that they can then be performed by artificial intelligence tools. For example, in the broad area of human resources, recruiting is the task of predicting—based on resumes, cover letters, LinkedIn profiles, and interview transcripts—which subset of applicants will perform best in the job. Promotion is the task of predicting which existing employees will perform best in a higher-level position. And retention is the task of predicting which star employees are most likely to leave and which of the available incentive options could most effectively be employed to encourage them to stay.…”
Implication 2“…Second, when automated prediction can increase the relative returns to capital versus labor in complementary decision tasks, it can lead to the complete automation of a complementary decision task. For example, human reaction times are slower than those for machines. The returns to a machine predicting a potential car accident a few seconds or even a fraction of a second before a human would predict the accident is higher when the response time of the machine is faster. Thus, automating the prediction task increases the returns to also automating certain decision tasks associated with vehicle control. Sometimes, the artificial intelligence is able to make better predictions than a human could because it has access to different data, such as feeds from cameras, RADAR, and LIDAR around a car.1 Once the prediction task is automated, it increases the returns to automating some of the complementary tasks, such as those associated with vehicle control…”

Ajay Agrawal, Joshua S. Gans, and Avi Goldfarb, "Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction," Journal of Economic Perspectives, Spring 2019, https://pubs.aeaweb.org/doi/pdfplus/10.1257/jep.33.2.31

Implication 4“…The fourth and final type of direct impact of artificial intelligence on labor happens when automated prediction sufficiently reduces uncertainty as to enable new decision tasks that did not exist before. The new tasks can be performed by capital or labor, depending on the relative costs of each (Agrawal, Gans, and Goldfarb 2019). Some tasks that are not economically viable when uncertainty is high become viable as prediction technology reduces the level of uncertainty. This relates to the reinstatement force in Acemoglu and Restrepo (in this issue) where a freeing up of labor as a result of automation increases the returns to technologies that use labor for new tasks. At this early stage in the development and use of machine learning, there are few tangible examples of new tasks that have already arisen because of recent advances in prediction technology…”

Implication 3, “…Third, automating the prediction task, in some cases, may have no impact on the productivity of capital performing a complementary task but may increase the productivity of labor.For example, ODS Medical developed a way of transforming brain surgery for cancer patients. Previously, a surgeon would remove a tumor and surrounding tissue based on previous imaging (say, an MRI scan). However, to be certain all cancerous tissue is removed, surgeons frequently end up removing more brain matter than necessary. The ODS Medical device, which resembles a connected pen-like camera, uses artificial intelligence to predict whether an area of brain tissue has cancer cells or not. Thus, while the operation is taking place, the surgeon can obtain an immediate recommendation as to whether a particular area should be removed. By predicting with more than 90 percent accuracy whether a cell is cancerous, the device enables the surgeon to reduce both type I errors (removing noncancerous tissue) and type II errors (leaving cancerous tissue). The effect is to augment the labor of brain surgeons. Put simply, given a prediction, human decisionmakers can in some cases make more nuanced and improved choices.…”