下载掌阅APP，畅读海量书库
立即打开
This book will cover the spectrum of concerns regarding AI from the perspective of marketing, business, and consumer persuasion. In view of the traditional methods of keeping people in line, the concept of AI in manipulation is relatively new, and we cannot demonstrate the evolution of manipulation through artificial intelligence without first describing its underlying mechanisms. While a human may not be able to solve an overly complex calculus problem, a supercomputer using AI with a strong algorithm adhering to, say, the architecture of the human brain can solve problems spanning many connected parts and different data. These processes or sets of rules to be followed in calculations or other problem-solving operations by computers are collectively called “artificial intelligence.”
AI is“the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings. The term is frequently applied to the project of developing systems endowed with the intellectual processes characteristic of humans, such as the ability to reason, discover meaning, generalize, or learn from past experience.”
We first need to clarify what we mean by the word “intelligence”in order to define AI in an easily digestible manner. For purposes of this book, “intelligence”comprises mental activities such as learning, reasoning, understanding, grasping truths, seeing relationships, considering meanings, and separating facts from beliefs. Below, we explore the differences in intelligence between a human being and an AI for each of the following categories：learning, reasoning, understanding, grasping truth, and seeing relationships.
Learning. A child can learn the difference between a fork and a knife when his mother shows him a knife for the first time. The child is able to quickly and concisely process this new information. The AI in your iPhone that scans your photo gallery would have to see four million pictures of a boy in order to recognize one.
Reasoning. A music student can recognize when she hears Beethoven playing. She can be inspired to create another song using the notes that she heard. An AI, however, has a limited set of operations it can perform. In Shanghai for example, AI connected to cameras can hear the sound of your horn, take screenshots of your license plate and report it, and arrange for a fine to be sent to your home address(it is forbidden to use your horn in Shanghai). This AI cannot process information in a multitude of ways, much less interpret sound beyond recognizing it.
Understanding. A girlfriend understands something is wrong when her boyfriend does not respond to her texts for a long time. The AI in your Facebook messenger, which tracks and scans all of your texts, will only see a technical bug.
Grasping Truths . The same girlfriend will grasp different truths when the boyfriend continues not to answer. She will remember that they had an argument the day before. She will assume he is not answering because of the argument or perhaps another reason. She can do something that an AI cannot：she can interact with different quantities of data and combine knowledge from different sources, whereas an AI will see a glitch or possibly human error in the coding.
Seeing Relationships , Considering Meanings , Separating Facts from Beliefs . The boyfriend can separate facts from beliefs with regard to relationships. He will confirm specific data to his girlfriend such as whether he is at the hospital with a broken cellphone from a minor car accident. He will think that his girlfriend will be worried, perhaps about yesterday's argument. He could also apply truths to particular situations or separate facts from beliefs, which an AI cannot.
There are various forms of Artificial Intelligence that brands can use to collect private data on consumers to better understand human behavior(and push them to use services). As mentioned above, the more data you collect, the better AI works. According to Data Never Sleeps, the world is collecting data at a rate of 2.5 quintillion(10 18 !)bytes per day. In 2018, Google alone processed 3.8 million searches per minute. The Google computer learns from the big data collected through your smartphone, your laptop searches, your Google Home, or through innovative installations(all these devices for collecting AI-supportive data are among those objects comprising the“Internet of things”). Extensive data collection will be key to our thesis regarding the manipulation of people.
Alan Turing, the father of AI, showed that any computation of data could be implemented digitally. Naturally, the idea of building an intelligent machine that could mimic the human brain could not be too far of f. Obviously, imitating nature does not work all the time；for example, some scientists sought to imitate the movement of bats when designing planes, but all the planes crashed.
Artificial intelligence is not currently capable of fully imitating natural intelligence. It is still in its infancy. Engineers seek to model intelligence through different dimensions：
·visual-spatial(understanding the environment for navigating；a GPS is much more precise in locating a spot than a human-brained taxi driver)；
·bodily kinesthetic(a surgeon performing an operation that can be repeated by a robot；the AI of a surgeon robot today has far more skills than a human)；
·creative(Picasso painting a canvas, or you deciding to cook a chicken with Coca-Cola, just to be creative or spontaneous；AI would need an intrapersonal consciousness for creating, even by accident)；
·intrapersonal(Gandhi was for example aware that he created a movement of non-violence to achieve independence for his country)；
·interpersonal(interacting with many people in a business meeting on Skype, dealing with different pieces of information)；
·logical and mathematical(AI can be much faster in mathematics than a human brain, depending on the system running the Ai)；
·linguistic(understanding a language). Natural language analysis is a common activity of Ai. Alibaba speakers and Amazon devices equipped with Amazon's virtual assistant, like Amazon Echo or Alexa, can listen to your orders and process and analyze human
language beyond mere code—basically, they can understand and communicate using human speech. The goal is to have computers that can naturally interact with people in a way that would pass the Turing
test, an examination by which an AI is deemed conscious when an observer cannot tell the difference between an artificial and natural intelligence.
New methods are being sought that will enable artificial intelligence to mimic the functions of the human brain. We made the first step with deep learning . Deep learning is a classification and learning system working with artificial neural networks that allows a computer to acquire human cognitive capacities. In its present forms, artificial intelligence is able to recognize objects；however, while a baby only needs a couple of examples to create a noun association and develop links between them, artificial intelligence requires millions of associations. Hence, in order to teach a program how to recognize a car, for example, we would need to train it using millions of labeled car pictures. Through training and engaging its neural network, AI will be able to recognize new car pictures by association.
In 2012, the most significant deep learning progress was made when Google Brain discovered the concept of a cat by itself. The learning for this discovery was not supervised. In this three-day experiment, the computer analyzed 10 million random unlabeled You Tube screenshots. Nobody told the program that the images were of cats；it learned by itself how to differentiate between cat heads and human heads. With that progress, machine learning made a breakthrough.
To our present understanding, AI is an innovative concept which will experience four phases：
1. The first phase was from 1960 to 2010. During this period, we had traditional programs with manual algorithms.
2. The second phase started in 2012 with deep learning and the first programs surpassing humans in their capabilities, for example in facial recognition. Deep learning allows a program to learn how to visualize the world through the help of virtual neurons making calculations. Deep learning progresses through implementation and grants tremendous control to those who possess data, such as GAFA and BATX. Thanks to the integration of deep learning, artificial intelligence is even able to compete against radiologists.
3. After having seen the progress made with deep learning, the next step will be for artificial intelligence to integrate a memory and act transversally. With those new capacities, artificial intelligence can not only surpass radiologists but even compete against general practitioners. This third generation will not be created before 2030, say the most optimistic among us, and 2199 for the most pessimistic.
4. The fourth and last phase will integrate consciousness in Ai. With artificial consciousness we will have a strong AI that will be capable of displaying smart behavior, experience self-awareness, have feelings, and have an understanding of its own reasoning.
According to Luc Ferry in 2018, we are right now in the middle of the second chronological step of AI history, called“weak AI, ”as distinct from“strong Ai.”
— Weak AI , or“narrow”AI, is machine intelligence that is limited to a specific or narrow area. An AI with narrow capabilities is fast and ef ficient, and can for example beat a human at chess, Go, or poker. Let's take the example of the game of Go. In January 2016, an AI called AlphaGo played five games of Goagainst 18-time world champion Go player Lee Se-dol. There are more possible end outcomes to Go than there are atoms in the universe, making Go“a googol times more complex than chess”. A “googol”refers to the number 10 100 , that is, the number expressed in decimal notation by the numeral one followed by 100 zeros. Thus, an AI cannot play a normal-length game of Go if its strategy is to consider all of the possible outcomes of the game. Winning at Go therefore requires that an AI have human-like intuition . AlphaGo possesses something akin to that of neural networking, which allows it to be taught by others and itself. After his defeat, Lee expressed that the AI executed“brilliant”moves which“surprised”him because, in his words, they were moves“no human would ever make.” Whereas a weak AI gives a huge amount of capacity to people processing data, this step of deep learning(techniques combining machine learning and learning reinforcement)allows computers and algorithms to control and digest so much information(billions of data points)that it can be used to predict the future. All of the intelligent methods and mechanisms which have been implemented and developed until now are considered weak Ai. Weak artificial intelligence still needs people, such as Google's raters, to help improve its accuracy and to train algorithms properly. But weak AI has great potential to behave foolishly. For example, Tay, an AI invented by Microsoft, became racist itself after analyzing millions of racist chats without understanding the concept of racism. Similarly, Facebook's AI has issues with racism and discrimination, encountering trouble identifying hate speech and harassment before it is flagged by users. The AI can only detect about 14% of harassment, bullying, and hate speech before Facebook users report it, whereas its AI can identify “spam content, terrorist propaganda, violence, and nudity with almost complete accuracy.”
— Strong AI is a form of machine intelligence that is equal to human intelligence and can solve problems the way humans do. A strong AI would be clever but not always ef ficient in the moment, which means it would act like a human, as well. Strong AI can do very complicated tasks by use of an intellect comparable or equal to that of humans. “Key characteristics of strong AI include the ability to reason, solve puzzles, make judgments, plan, learn, and communicate, ”and should have capabilities like“consciousness, objective thoughts, selfawareness, sentience, and sapience.” Although strong artificial intelligence has not yet been implemented or developed, there is widespread fear of strong artificial intelligence because of its characteristics being comparable with the human mind. Computers have already generated copyright-protected creative content and have assisted, albeit to a limited degree, in the creation of patented ideas by using AI techniques. On an individual scale, strong AI is considered to be dangerous(we will see the different types of AI and their various impacts in the area of surveillance below). Such a fear does not come as a shock given that movies such as Terminator , Frankestein , Avengers ： Age of Ultron , and I , Robot have shown the theoretical dangers and potential catastrophes of an AI escaping human control. In a survey of over 4,000 Americans, more than 70% fear robots taking over their lives. There is also more concern than enthusiasm about the prospect of machines performing jobs done by humans. To the human workforce, strong AI presents a major threat.
The importance of AI innovation is huge with regard to the expansion and survival of the companies of the Western world. AI will be one of the means of protection against emerging countries. But for Western companies or countries, AI is a technological innovation that is quite high-risk due to its complexity and expense, and not simply in terms of research and development. So, except for the fascination that we have for AI after watching such movies as Ready Player One , why are so many people obsessed with innovation in the field of artificial intelligence? There are many reasons. One is that AI represents the future of the economic world. Onethird of the GDP of a country such as the US comes from products that did not exist as recently as 25 years ago. Second is that, given that AI is developing at a rapid pace, the idea of a powerful, self-aware AI scares the human race. Such revered prophets as Elon Musk, the famous boss of Tesla, warn of the dangers of Ai. Musk even created a group called Open AI( https：//openai. com )to research and evaluate the safety and societal issues surrounding Ai. But even if AI is developed in a way that is safe and manageable, this will not stop companies from collecting private data. Contrary to what many people think, AI is not a hollow promise for the future but a reality that is already pervasive in our daily lives.
There is a growing lexicon of terms surrounding AI, many of which are confusing. The field of AI has yet to settle upon even a specific structure. However, there are several main areas of current research that illuminate some of the dif ferences between approaches：natural language processing, genetic programming, machine learning, deep learning, expert systems, computer vision, computerized speech recognitionand, above all, artificial neural networks(ANNs). ANNs create algorithms that simulate the operations of neurons similar to brain activities where neurons interact with each other； memorizing inputs, learning correlations, and making decisions. Increasing the speed and calculational ability of computing allows neural networks to include more layers and more neurons, which gives them higher capabilities in learning and analyzing big data. Larger neural networks are capable of digesting and analyzing bigger, more complex data, and can detect more nuances in big data and make discoveries and predictions. Artificial neural networks thrive on data volume and speed, so they can be used within real-time or very near real-time scenarios. When you go for example to the airport, you leave what are called initial inputs (basic information)when you check in. These inputs are processed by algorithms and linked to additional layers to compare your data with the data of other passengers and to establish patterns. Essentially, this AI uses an analog of the mechanisms of the brain and neurons specialized in cell transmission of nerve impulses, doing millions of small calculations to interpret your path in the airport. After only a couple of repetitions of the same operation, the machine has learned the various paths so well that it can place along your path targeted advertisements or an invitation to go to the duty-free section. It is a machine(which is) learning . I have met with some of the teams working within JCDecaux(a world leader in digital billboards)who are developing a project of machine learning for calculating more sophisticated routes through the airport in order to sell more luxury products in duty free. Most AI techniques that require“training”the system with large data sets use some form of artificial neural network.
By analogy, using neural networks is an old technique in marketing. For example, by the end of the 1970s, it was possible to track the eyes of a customer while watching an advertisement. Tools of neuromarketing simply followed the pupils of the viewer. In short：the dark circular opening in the center of the iris varies in size to regulate the amount of light reaching the retina. By following the pupil, eye-tracking can ascertain whether the eye watched the body of the model during a cosmetic advertisement, or the product itself. Advertisers could finally change the position of the product in the advertisement and on the screen for a bigger impact.
The “Internet of things”(Io T)—all the devices such as phones, watches, smart glasses, light bulbs, TVs, security systems, vehicles, wearables, voice assistants, smartphones, and much more that can be connected to the Internet and to one another—currently consists of 27 billion devices, a number that is expected to nearly triple to 75 billion by 2025. The Internet of things will be able to identify people through facial recognition, which will enable owners of big data to identify you everywhere you go. For example, a camera in London can record you 300 times per day(420,000 cameras, so one camera for around 15 people). The typical American household contains five connected devices and 77 percent of Americans go online daily. In the near future, people will check in to hotels, spas, pools, restaurants, or supermarkets while cameras observe and analyze their facial reactions, as is seen in many science fiction movies from Blade Runner to Minority Report .
Facial recognition will help the field of neuropsychology to understand people's basic reactions. Our psychological responses come from our brain, which can be divided into three parts：
—The reptilian brain allows us to carry out unconscious actions, such as breathing.
—The limbic brain(on the left side)is a complex system of nerves and networks in the brain, involving several areas near the edge of the cortex concerned with instinct and mood. It controls memory and the basic emotions(fear, pleasure, anger)and drives(hunger, sex, dominance, successful parenting).
—The cortex(on the right side)—the outer layer of the cerebrum(the cerebral cortex)composed of folded grey matter—plays an important role in one's consciousness. The cortex is responsible for our thoughts, intelligence, imagination and rationality.
Let's say a man asks his girlfriend, “what is your favorite restaurant? ”The girlfriend may look unconsciously to the left to probe an answer from the limbic brain, the brain of instinct and of memory. If the girlfriend is looking to the right, it means that she is using the cortex and her imagination to create a fake answer, a lie.
Now let us consider how one can manipulate or counter-manipulate people knowing information from a facial recognition program. While we can only imagine the scope and capabilities of this form of AI, integrating this technology is closer than you might think. In 2030, there will be enough cameras inside and outside your house to collect and immediately process your data using facial recognition. Analysis of your reactions in a store could be instantaneously carried out and forwarded to the screen or Google glasses of a salesman, enabling him to give you recommendations on what to buy(Maserati is already using this kind of technology, in order to recognize the client inside the store and pass along relevant information to salesmen's iPads). There are various approaches to collecting personal data with AI, whether it is genetic testing or tracking eye movements. Companies and brands can use this type of information in their expansive collections of data in order to obtain a total, comprehensive understanding of you. Using AI to analyze your data can help to change your act of purchasing.
A funny story happened to me in 2005. As a young professor, I gave a lecture at the University of Nice. I said to my students, “did you know that there is a company in California which offers the possibility to use free office tools like Windows in exchange for tracking and scanning your emails? The emails are analyzed, then this company ranks your most-used keywords and groups them into themes. Then, an advertisement linked with these themes is sent to you by email.”My students stared back at me with eyes agape in shock and horror as I told them about this 1984-esque scenario(none of them had read 1984, but they had heard of it). When I asked them what they thought about it, they told me, “This is scandalous!”I asked them if they wanted to know the name of the company. Naturally, my inquisitive students said“yes.”I said one word：“Google.”They opened their eyes even wider. The students could not believe it.
As this story illustrates, no level of shock you may feel about a new way of collecting data is going to change anything about it. In Yuval Noah Harari's book Homo Deus , humans seek immortality by upgrading themselves to“godlike”status through biological and cybernetic engineering, and discover helmets that allow telepathy between people. Telepathy allows communication between two different people by means other than their familiar senses. These helmets are performing basic actions, but they are enough to collect a giant quantity of personal data on one's behavior at home. Along the same lines, in Sweden microchips have been invented to be inserted under your skin, enabling employees of a company to clock a large number of actions during one day. These new ways of collecting data may seem surreal(even when we know that Japanese scientists implanted microchips in real cockroaches during the 1990s), but at the same time：
1)People are becoming more and more engaged with technology, any why would we keep on using laptops or cellphones when we could have all these devices under our skin?
2)How can we expect companies'techniques to change when we know(as with the Google example with my students)that no consumer will complain for more than a few seconds?
Can we still protect our free choice in buying and privacy of thinking? In 2015, the details of the British Secret Intelligence Service's “Karma Police”revealed that consumers are frequently facing credit checks and are subject to online shopping experiments using personalized prices. With such inspection of everybody's Internet use, can it be said now that the manipulation of citizens and society has become a reality? The control of our data through algorithms is an effective way to understand the way we think, what we do, and how we feel. Europe has tried to protect consumers and the privacy of their information with GDPR, the now-famous law that transformed the term “cookies”in our everyday vocabulary. American GAFA seek to sidestep this kind of legislation because they truly think that they can help people with free and friendly services that customers are happy to use every day. In the United States, only a few federal regulations protect privacy, and they all are industry-specific. There is no global regulatory body protecting consumer privacy. The foremost example of federal law protecting privacy in the US is the Gramm-Leach-Bliley Act(GLBA), which controls how financial institutions treat personal data. Countries such as China have likewise started to regulate activities in their market. Right now, we cannot imagine how many legal complications AI is going to create and what it will take to protect humans from these complications.
How could the global manipulation of consumers work without law? Using AI as a system or mechanism to predict human behavior and persuade consumers to buy is essentially a natural continuation of companies'prior coercive marketing methods. Brands cultivate this manipulation by connecting with consumers, meaning there is an unconscious relationship between brands and people. As humans, it is natural for us to become attached to people, animals, even inanimate objects；the same goes with brands.