Artificial intelligence is a new technology, not a new economy.

Artificial intelligence is a kind of prediction technology, prediction is the input of decision, and economics provides a perfect explanation framework for the trade-offs involved in any decision.

In fact, the new wave of artificial intelligence has not brought us intelligence, it has brought a key component of intelligence – prediction. Even if you have never been involved in the programming of convolutional neural networks or studied Bayesian statistics, but you want to understand what artificial intelligence means to you, but you are still at a loss. Perhaps this article can help you.

We need to know what kind of forecast is most important to the company. How will the further development of artificial intelligence change the prediction on which we rely? With the rise of personal computers and the Internet, all trades and industries have relocated their jobs. In response to the progress of prediction technology, how will our industries relocate their jobs?

However, in the face of the development of artificial intelligence, there is no absolutely perfect strategy. More data means less privacy. The faster the speed, the lower the accuracy. The stronger the autonomy, the weaker the control. The most suitable strategy for your company or occupation should depend on how you balance the weight of each factor in each trade – off.

Internet leads to lower distribution, communication and search costs

We have become accustomed to the report that ” artificial intelligence is about to change our lives” with mobile phones flooding the screen. Although some of us are technophiles celebrating the endless possibilities of the future and others are technophobes mourning the lost good times, almost all of us are so accustomed to the continuous drumbeat of technological news that we are almost numbly chanting, ” the only constant is change itself.” It was not until we really stood in the middle of the wave that we suddenly realized that this time the technology was somewhat different.

Since 1996, the number of academic papers and research on computer science has soared more than 9 times. Academic research and research are usually the forerunners of new intellectual property rights and patents. The Scopus database has over 200,000 ( 200,237 ) papers in the field of computer science using the keyword index of ” artificial intelligence” and nearly 5 million ( 4,868,421 ) papers in the field of computer science.

Since 2000, the number of active AI start-ups has increased 14 times. These companies cover many industries – drug development, customer service, manufacturing, quality assurance, retail and medical devices. This technology is very powerful and versatile, creating important value in a wide range of applications.

The concept of artificial intelligence has gradually entered our sight. It’s packed with apps for your mobile phone, it’s optimizing your grid, it’s replacing your stock broker; It won’t be long before it will probably carry you around or deliver express to you. With the influence of artificial intelligence gradually spreading to all walks of life, technology advocates no longer call AI new technology, but begin to call it ” new economy”. Politicians, business executives, investors, entrepreneurs and major news organizations began to use the word, and everyone began to talk about the ” new economy”.

In fact, the word ” new economy” has been repeatedly talked about when the Internet was popularized, but in the Internet era, we have not seen a new economy or a new economics.

Admittedly, the Internet has brought about many important changes. Goods and services can be circulated digitally; It is easier to communicate with each other. To find information, click the search button to search with one click. However, all things such as commodity circulation, communication and information search can be done before the advent of the Internet, which enables them to do it in a cheaper way.

In other words, the rise of the Internet means that the cost of distribution, communication and search will fall. For example, from an economic point of view, Google only makes search cheaper through the Internet. When search becomes cheap, those enterprises ( e.g. yellow pages, travel agencies, etc. ) that provide information retrieval and make money by other means feel a serious survival crisis. At the same time, those professions that rely on being discovered ( such as self-publishing writers, sellers of rare collections, local filmmakers, etc. ) have been given opportunities to flourish.

The cost changes of some economic factors have greatly affected the business models of some enterprises and even the structural systems of some industries. However, the economic laws have not changed, and everything can still be understood from the perspective of supply and demand. We can still make use of ready-made economic principles to formulate strategies, provide information for policies and predict the future.

Artificial Intelligence Causes Cost Reduction of ” Forecast”

Now, let’s look at artificial intelligence. The economic significance of artificial intelligence is precisely because it makes something important cheaper. Considering the computer itself, ” it has no ambition to create. It can do whatever we command it to do, and it can do according to analysis, but it does not have the ability to predict the relationship or truth that needs analysis. ” Therefore, computers are still unable to think, so intelligence, reasoning or thought itself will not become cheap. On the contrary, what will become cheaper is something very common. Just like computing, you don’t even realize how common it will become and how huge its price drop will have on our life and economy.

The important and common thing that has become cheaper is actually – prediction. Forecast is the process of filling in missing information. Forecast will use the information you currently have ( usually called ” data” ) to generate information that you do not yet have. Most discussions on artificial intelligence emphasize a variety of prediction technologies, which have increasingly difficult and fuzzy names and labels: classification, clustering, regression, decision tree, Bayesian estimation, neural network, topological data analysis, in-depth learning, reinforcement learning, in-depth reinforcement learning, etc. We will skip over the mathematical details behind these methods, but we need to know that each method here is related to prediction: use the information you already have to obtain the information you do not yet have.

Predictions are cheaper, meaning they will become more. The theory of traditional economics still applies to artificial intelligence: when the cost of something falls, we will do it more.

The development of artificial intelligence can save the forecast cost of industries that used ” forecast” originally, such as goods management, demand forecast, budget making, etc. More importantly, as ” forecast” becomes cheaper, ” forecast” is gradually applied to some industries that did not use ” forecast”.

For example, self-driving cars have been in a controlled environment for more than 20 years. However, they can only operate in places with detailed floor plans, such as factories, warehouses, etc. Having a floor plan means engineers can design robots to operate according to basic ” if-then” logic: if someone walks in front of the vehicle, then stop; If the shelves are empty, turn to the next row. But these vehicles can never enter the ordinary city streets. There are too many things happening on the ordinary streets to write ” if – then” codes one by one.

With the development of artificial intelligence, engineers have framed the problem again from the perspective of prediction. Engineers realized that instead of telling machines what to do in every situation, they only need to focus on a prediction problem – ” what will humans do?” Imagine an artificial intelligence robot sitting in a car with a human driver. Human drivers drive millions of miles. He receives environmental data through his eyes and ears, processes the data with his brain, and then takes corresponding actions according to the incoming data: going straight or turning, braking or accelerating. Engineers installed various sensors ( such as cameras, radars, laser locators, etc. ) to artificial intelligence, giving it its own eyes and ears. Therefore, when human drivers are driving, artificial intelligence observes incoming data and human behavior at the same time. When specific environmental data are transmitted, will the human driver turn right, brake or accelerate? The more human drivers are observed by artificial intelligence, the better they can predict the specific actions that drivers will take when receiving specific environmental data.

Artificial intelligence learned to drive by predicting what human drivers would do under specific road conditions. Relying on cheap prediction, we turned driving into a prediction problem. We only need to input enough scenes and the corresponding operations of each scene, and train the machine constantly. It can predict the corresponding operations of each scene immediately after it appears. When the amount of training received by the machine reaches a certain level, it can drive automatically in an uncontrolled environment, even on the streets and highways of the city.

At the same time, once again using the theory of economics, when the means of production such as ” forecast” become cheap, the value of ” forecast” supplements will increase accordingly. Just as a drop in the cost of coffee will increase the value of sugar and cream, for self-driving cars, a drop in the predicted cost will increase the value of sensors that capture data around the vehicle. In 2017, Intel spent more than $ 15 billion on the acquisition of Mobileye, an Israeli start – up. This is mainly to obtain the latter’s data acquisition technology, which allows vehicles to effectively ” see” objects ( parking signs, pedestrians, etc. ) and signs ( lane lines, roads ).

Under Artificial Intelligence, Trade – offs of Enterprise Strategies

So, how does artificial intelligence ( or ” predictive” cheapness ) affect corporate strategy?

Let’s look at an example of Amazon. Most people are familiar with how to shop on Amazon. Like most online retailers, you visit websites, buy goods, put them in ” shopping carts”, pay for them, and Amazon sends them to you. At present, Amazon’s business model is to shop before shipping. During the shopping process, Amazon’s artificial intelligence predicts what you want to buy and then provides corresponding recommendations. Artificial intelligence can accurately predict 5% of what we want to buy. In other words, for every 20 items it recommends, we actually buy one.

As consumer buying behavior continues to accumulate, Amazon’s artificial intelligence collects more information and uses this data to improve its predictions. This improvement is like turning up the volume knob of the speaker. However, they are not raising the volume, but the accuracy of artificial intelligence prediction. They turned the knob to a certain point, and the accuracy of artificial intelligence prediction crossed a certain critical value, which changed Amazon’s business model. This kind of prediction is accurate, and it directly predicts the goods you want to buy and sends them to you ( even without waiting for your order ).

AI can make amazon more successful! Because of it, you don’t need to go to other retailers any more. Amazon will occupy a larger market share and send the goods to you before you buy them, which may prompt you to buy more other things. If everything goes smoothly, Amazon will send the goods to the door before you buy them, so that you will not suffer from shopping. The forecast knob is set high enough to change Amazon’s business model from ” buy before send” to ” send before buy”. Of course, consumers are reluctant to take the trouble of returning everything they do not want. As a result, Amazon will invest in product exchange infrastructure, such as a fleet responsible for distribution, making weekly rounds of inspection to easily recycle what customers don’t want.

If this is a better business model, why hasn’t Amazon done so yet? Because if it is implemented now, the cost of collecting and handling returned goods will far exceed the extra money earned from customers. For example, now we have to return 95% of the goods we send, which will be very annoying to us. For Amazon, the cost of handling returns is much higher than the revenue generated by the increase in market share. The prediction generated by artificial intelligence is not accurate enough to help Amazon generate net profit.

However, it is not hard to imagine that Amazon will adopt this strategy before the technology is accurate enough to bring profit to it, because Amazon has foreseen that it will bring profit as long as the prediction is accurate to a certain extent. If it is implemented one step earlier, Amazon’s artificial intelligence will get more data faster and then improve faster. The earlier you start, the harder it will be for your competitors to catch up. Good prediction will attract more shoppers, more shoppers will produce more data to train artificial intelligence, and more data will bring better prediction, thus repeating the cycle and realizing a virtuous circle. It may be expensive to adopt the new strategy too early, but it may be fatal to the company if it moves too late.

Recently, Google has publicly claimed to place artificial intelligence technology at the center of all work. Google is not alone in making this strategic commitment. In the same month, Microsoft announced its intention to shift from ” mobile first” and ” cloud first” to ” artificial intelligence first”.

However, turning the company into an artificial intelligence-oriented company is not necessarily a correct choice. From an economic perspective, any statement that ” we will focus on XX” means a trade – off. Always give up something for exchange. ” Artificial intelligence first” means to invest resources in data collection and learning ( a long-term goal ), while sacrificing important short-term considerations, such as direct customer experience, revenue and number of users.

Finally, it should be pointed out that the artificial intelligence mentioned in this article is not general artificial intelligence, but a more narrow prediction machine. At present, people are studying how to make prediction machines operate in a wider environment, but human beings have not yet found a breakthrough in general artificial intelligence. Some people think that general artificial intelligence is too far away to spend time worrying about it. In a policy document drafted by the U.S. Presidential Administration Office, the Technical Committee under the National Science and Technology Commission ( NSTC ) said: ” The consensus currently reached between the private sector expert community and the Technical Committee under the NSTC is that general artificial intelligence cannot be realized in at least recent decades. The NSTC’s technical committee’s assessment is that long-term concerns about general artificial intelligence with super intelligence should not have too much impact on current policies. ” Meanwhile, several companies that claim to create general artificial intelligence or human-like intelligent machines, such as Vicarious, Google DeepMind, Kindred and Numenta, have already started deploying artificial intelligence and the future is full of uncertainties.

When we surpass the prediction machine and enter the era of general artificial intelligence and even super intelligence, we will usher in a different era of artificial intelligence. By that time, we can confidently predict that economics will not be so simple.