I recently attended a very interesting talk, - Exploring the Impact of Artificial Intelligence: Prediction versus Judgment, - by University of Toronto professor Avi Goldfarb. The talk was based on recent research conducted with his UoT colleagues Ajay Agrawal and Joshua Gans. In addition to an in-depth paper aimed at a research audience, they’ve explained their work in two more general interest articles, one in the Harvard Business Review and the second in the MIT Sloan Management Review.
In their opinion, “the best way to assess the impact of radical technological change is to ask a fundamental question: How does the technology reduce costs? Only then can we really figure out how things might change.” For example, the semiconductor revolution can be viewed as being all about the dramatic reductions in the cost of arithmetic calculations. Before the advent of computers, arithmetic was done by humans with the aid of various kinds of devices, from the abacus to mechanical and electronic calculators.
Then came digital computers, which are essentially powerful calculators whose cost of arithmetic operations has precipitously decreased over the past several decades thanks to Moore’s Law. Over the years, we’ve learned to define all kinds of tasks in terms of such digital operations, e.g., inventory management, financial transactions, word processing, photography. Similarly, the economic value of the Internet revolution can be described as reducing the cost of communications and of search, thus enabling us to easily find and access all kinds of information, - including documents, pictures, music and videos.
How does this framing now apply to our emerging AI revolution? After decades of promise and hype, AI seems to have finally arrived, - driven by the explosive growth of big data, inexpensive computing power and storage, and advanced algorithms like machine learning that enable us to analyze and extract insights from all that data. Agrawal, Fans and Goldfarb provide an elegant answer to this question in their HBR article. “Machine intelligence is, in its essence, a prediction technology, so the economic shift will center around a drop in the cost of prediction.”
What are the key implications of significant reductions in the costs of predictions? First of all, it will lower the costs and improve the quality of goods and services that already rely on prediction, such as weather forecasts and personalized marketing. And, - as was previously the case with arithmetic, communications and search, - we will be able to use predictions in all kinds of new applications.
Machine translation is one such application. A recent NY Times Magazine article told the story of how Google used data-driven prediction based on deep machine learning to significantly improve Google Translate, one of its more popular online services. According to the article, Google achieved a quantum improvement in the quality of its machine translations when it switched to its new deep-learning-based system, - an improvement roughly equal to the total gains the old Translate system had accrued since its inception in 2006. Google expects that the new translation model will become the common foundation for translating between different language pairs, rather than needing 150 different models as has previously been the case, as well as enabling a variety of innovative applications based on natural language processing.
Over time, we’ll discover that lots of other tasks can now be reframed as prediction problems. Equally important, as machine predictions become inexpensive and commonplace, the value of human tasks that leverage and complement prediction will rise. Judgment, in particular, will become more valuable.
Decisions typically involves two main activities: predictions and judgement. Judgement is the part of decision-making that, unlike prediction, cannot be explicitly described to and performed by a machine. Whereas predictions are generally based on information, judgement is based on indescribable factors like intuition, unconscious feelings, or analogies with somewhat similar situations from our past.
“Judgment is the ability to make considered decisions - to understand the impact different actions will have on outcomes in light of predictions,” write the authors in the Sloan Review article. “Tasks where the desired outcome can be easily described and there is limited need for human judgment are generally easier to automate. For other tasks, describing a precise outcome can be more difficult, particularly when the desired outcome resides in the minds of humans and cannot be translated into something a machine can understand.”
Predictions will continue to improve. As data availability expands, predictions become more valuable and can be applied to a wider variety of tasks. At the same time, our understanding of human judgements will also improve. Advances in machine learning will find ways to analyze the relationships between decisions and outcomes, and use that new information to quantify some aspects of human judgements and make them thus suitable for machine predictions. “By breaking down tasks into their constituent components, we can begin to see ways AI will affect the workplace. Although the discussion about AI is usually framed in terms of machines versus humans, we see it more in terms of understanding the level of judgment necessary to pursue actions.”
As AI technologies advance, machine predictions will increasingly replace human ones in an increasing number of tasks. What roles will humans then play that emphasize their strengths in judgment while recognizing their limitations in prediction? The article offers three interrelated insights:
Prediction is not the same as automation. “Prediction is an input in automation, but successful automation requires a variety of other activities. Tasks are made up of data, prediction, judgment, and action. Machine learning involves just one component: prediction. Automation also requires that machines be involved with data collection, judgment, and action… Prediction is the aspect of automation in which the technology is currently improving especially rapidly, although sensor technology (data) and robotics (action) are also advancing quickly.”
The most valuable workforce skills involve judgment. As predictions continue to improve, “employers will want workers to augment the value of prediction; the future’s most valuable skills will be those that are complementary to prediction - in other words, those related to judgment… [It] seems likely that organizations will have continuing demand for people who can make responsible decisions (requiring ethical judgment), engage customers and employees (requiring emotional intelligence), and identify new opportunities (requiring creativity).”
Managing may require a new set of talents and expertise. “Today, many managerial tasks are predictive. Hiring and promoting decisions, for example, are predicated on prediction: Which job applicant is most likely to succeed in a particular role? As machines become better at prediction, managers’ prediction skills will become less valuable while their judgment skills (which include the ability to mentor, provide emotional support, and maintain ethical standards) become more valuable.”
Our understanding of the interplay between predictions, judgement and decisions is just at the beginning, opening up many exciting research directions. “As the range of tasks that are recast as prediction problems continues to grow, we believe the scope of new applications will be extraordinary,” write the authors in conclusion.
Great article, very good outline of the potential impact of the combination of machine learning, artificial intelligence, and deep learning. As the understanding of brain functions and processes improves, I expect more "cognition" to be embedded in decision making.
Posted by: Steve Knode | May 05, 2017 at 09:31 AM