15 Minutes
Cover of Prediction Machines

Technology and the Future · 15 min

Prediction Machines

Ajay Agrawal, Joshua Gans and Avi Goldfarb · 2018

Who it's for

For managers, entrepreneurs, and curious readers who want to evaluate where artificial intelligence can help a business, using the language of economics.

Read it in the app

⚡ Free to download • 🎁 One full book free every day

🎧 Listen to this summary in the app

Every summary has audio narration. Open the app to read it or listen to it.

3 ideas to take with you

1

Consistently recording what happens in a business is the first step toward eventually having data that support useful predictions.

2

The autonomy granted to a machine should depend on the cost of its mistakes.

3

If a decision's bottleneck lies in judgment, a better prediction alone may not solve the problem.

“
When something becomes cheap, we start using it even where it once was not worthwhile.
An idea from Prediction Machines, in the words of 15 Minutes

Preview

Start reading now

Artificial intelligence through economists' eyes

Prediction Machines, by Ajay Agrawal, Joshua Gans, and Avi Goldfarb, published in 2018, attempts to explain artificial intelligence without fascination or panic. The authors are economists proposing a simple perspective: AI's major workplace innovation is dramatically reducing the cost of making predictions.

This view is less glamorous than conscious robots, but it allows practical questions. If prediction becomes cheap, what changes in business? Where will companies use more prediction? What other things become more valuable? The book answers through supply-and-demand logic, something everyone already applies when a price falls.

The trio's goal is to help managers decide where to invest, what to automate, and how to prepare their organizations. The tone resembles a strategy manual, with examples from technology, healthcare, retail, and transportation companies.

What prediction really means

The authors explain that prediction is not only about looking into the future. It means using information you have to fill in information you lack. A physician diagnosing an illness is predicting the cause of symptoms. A system translating text predicts which word comes next in another language.

Machine learning, the technique behind recent advances, works by giving an algorithm many examples and letting it find patterns. Once trained, it receives new data and makes an estimate. More high-quality data generally lead to better results.

This broad definition has an important consequence: many problems that do not look like prediction can be reframed that way. Recognizing a photograph, driving, recommending a movie, and detecting credit card fraud become tasks of predicting an answer from data.

When something becomes cheap

The central economic idea is simple. When something's price falls, people use more of it, including in situations where it previously was not worthwhile. The authors use electric lighting as an example: once it became cheap, we started lighting everything, not just places where candles had burned.

Another example is arithmetic. Computers began as calculating machines. As calculation costs plummeted, they spread to spreadsheets, games, photography, and music, applications that all became calculation problems. Prediction is following the same path.

This leads to the book's forecast: cheap prediction will appear in new, sometimes unexpected tasks. People working in areas dependent on repetitive decisions, such as logistics, lending, and customer service, are likely to feel the impact sooner.

11 more sections and the quiz

Continue in the app

Read the whole summary, finish with the Prediction Machines quiz and follow the guided Journeys. No ads in the app.

About Ajay Agrawal, Joshua Gans and Avi Goldfarb

Ajay Agrawal, Joshua Gans, and Avi Goldfarb are professors at the University of Toronto's Rotman School of Management in Canada. Agrawal founded Creative Destruction Lab, a program supporting technology startups. All three are economists researching artificial intelligence's economic impact. They also coauthored Power and Prediction, published in 2022.

Explore