Artificial intelligence for everyone

Original author: -
  • Transfer
At the beginning of January 2019, Forbes compiled the TOP-10 of the main technological books of 2018 , among which was the “Prediction Machines: The Economics of Artificial Intelligence” . The book, written by a team of authors - Joshua Gans, Ajay Agrawal and Avi Goldfarb, blows up the established understanding of artificial intelligence and translates it into a completely different plane. This book is a real must have.

One of the authors of the book, an expert on artificial intelligence, Joshua Hans, a professor of the Rothman School of Management at the University of Toronto (Canada), spends a lot of time every day tracking news in the field of AI, separating HYIP from reality. Today, he teaches MBA students networking and digital marketing strategies, including how companies can successfully compete in their markets through technological innovation.

The editorial team of discussed with Joshua his sensational business book “Forecasting Machines: The Simple Economics of Artificial Intelligence.” Read the interview with him read below.


Today they talk a lot about the possibilities of AI.
But are there any specific examples of the benefits of AI in the modern business world?

I admit: around the artificial intelligence today raised a lot of hype. But in the book, we take a different approach to the consideration of everything that has been created in the field of AI over the past 10 years. We are not talking in the book about the general intelligence that can replace people and all their cognitive abilities - we are talking only about one aspect, namely, about our ability to foresee (predict, foresee).

We usually talk about foresight in the context of forecasting. Like, for example, with the weather - we first collect historical data on wind, precipitation and other factors, and then we make a weather forecast for tomorrow or next week.

But prediction is not always about the future. Computer vision is one of the most successful examples here: when you give an image to a computer and ask what “he sees”, the answer is actually a prediction. The computer as if asks itself: “What would a person think about exactly what is depicted in this picture?” And gives an answer.

Foresight is always about making the best decisions. Thanks to weather forecasts, we can decide which clothes to wear. And when you have an assumption about what is, for example, in an MRI scan, you can prescribe the right course of treatment.

From this point of view, AI is pretty boring. It is simply a better statistical technology. But the tremendous progress in the development of AI leads to the fact that his predictions will get better, faster and cheaper. And this will open up great opportunities that we did not have before.

So how does AI move from advertising to real value?

When we wrote this book, we recalled what was happening with the computer revolution and the Internet revolution. There was also a lot of hype around them, and many companies spent millions of dollars on things that weren’t really well thought out.

We do not want to repeat this error. Instead, we say: “If forecasting helps improve the decision-making process, then let's take the work processes of our organization and select from them all the decisions that we need to take in order to move from the source data to the result, and in this process we will determine where the sources of uncertainty. It is then that you will begin to understand where the AI ​​can be useful for you to reduce the number of uncertainties and make better decisions.

This process has already occurred with computers. People then divided workflows and tasks into separate iterations, and found out where computers would be useful. 20-25 years ago, this led to the phenomenon of reengineering. We suggest doing it again.

You write that everyone has a moment of insight with AI - the moment when everything becomes clear, like a click of your fingers. Is it worth it for each industry to wait for its insight to start using AI?

There are people who are still wondering: “Can AI help our business?” But this may already be a reality. For example, we studied one supermarket chain. They used AI to predict the level of loading of cold warehouses, which, if optimized, could lead to significant savings. After all, food warehouses are expensive, and you need to respect the optimal balance of supply and demand, otherwise you will face damage to the goods.

This network used machine learning to understand what exactly stimulates the demand for yogurt in Canada. They found that the weather was a significant factor in determining whether there was more yogurt in the store at the end of the day or less than expected. Even lowering the temperature a few degrees in general and so cold in Canada changed the consumer demand for yogurt. And it turned out to be something quite incredible! They began to see profits - 5% here, 5% there - and it all adds up to total profits. This is the very moment when people realize: "Oh, it really is important to us."

It is known that AI "eats" data. There are companies that have accumulated huge amounts of data for work, and others that are lagging behind in this area. Will AI benefit companies that have large amounts of data?

In fact, this question is quite difficult to answer. Definitely, for AI, data is needed, but the easiest way to scare everyone away from AI is to say: "They have the data, but you do not." I have no doubt that companies such as Google, Facebook and Amazon are leading in AI at the moment because they have long thought about the data and collected them correctly. Usually, the company, collecting data, does not think about how they will be used, and therefore is not necessarily in the same position.

For AI, you need the right data — well-structured, measuring the right characteristics, clean. It is likely that new companies starting data collection from scratch may end up creating better data for AI.

Where is the potential application of AI in the organization?

This is a complex issue that many organizations face. Now I see AI as one of the functions of analytics systems, since there are still a large number of different elements that require in-depth study of the data associated with them.

But in the long run this should change. First, there is always a choice — outsourcing or building your own systems, and both have their own advantages and risks. It also depends on which areas of the organization should be affected by the AI ​​functions. General AI capabilities make sense in centralized tasks, but more department-specific tasks can lead to a shift in the corresponding AI functions to these departments.

Take, for example, HR: personnel managers are always trying to predict whether a new employee will be productive, or whether to raise an existing employee as an employee. Today, HR departments have accumulated a lot of data that could help these forecasts, but all the necessary information is stored in files and not used.

What should be done to prepare for the introduction of AI?

Beware of the story. Beware of people giving technological gifts. AI is a very specific thing. A deep knowledge of technology and what it can give you will help you understand whether something really worthwhile is trying to sell you and what its potential is. In other words, it’s very important that the organization has people who can help you evaluate whether the potential benefits are real or not, in terms of data science and company operations.

At the same time there is a huge benefit from the experiments. If you have a large organization, then allow individual teams to use artificial intelligence in their activities. As an experiment, and not as a replacement for the main functions, this can be of great benefit. You are obliged to manage your risks, but you should not lose the opportunities provided by the AI.

Also popular now: