Artificial Intelligence Rises Again
The year 2016 was a special year for artificial intelligence. At the beginning of the year, AlphaGo triumphed over Lee Sedol, a 9-dan rank player, and brought artificial intelligence technology, which has once again risen in the past decade, to the stage and into the public view.1 In the past few years, technology giants have successively established artificial intelligence laboratories, invested more and more resources to capture the artificial intelligence market, and even completely transformed into artificial intelligence–driven companies, to intensively plan for the future of artificial intelligence. The Chinese government and other countries’ governments all regard artificial intelligence as a strategic force driving the future, and are introducing strategic development plans and promoting overall progress from the national level to prepare for the artificial intelligence society that is about to arrive. This revolution will not be contained to laboratory research. At the same time, academic research and commercialization are promoting the conversion of artificial intelligence into products and services, so that the public can truly feel its existence. Especially in areas such as image and speech recognition, natural language processing, and other applications based on deep learning algorithms, they are rapidly becoming industrialized, and the racetrack has already been laid out.
Although we often talk about AI (artificial intelligence) in different settings, we have found that the “artificial intelligence” currently being hotly debated around the world is not exactly equivalent to the artificial intelligence defined by earlier academics. Artificial intelligence researchers and product designers, business people, policy makers, and the wider public generally use the term “artificial intelligence” in different contexts. On the other hand, like the previous terms “cloud computing,” “big data,” and “machine learning,” the term “artificial intelligence” has been used by marketers and advertising copywriters without restraint. In the eyes of different groups, “artificial intelligence” seems to be both a panacea and a time bomb that causes massive unemployment.
As a technical term, “artificial intelligence” dates back to the 1950s. John McCarthy, an American computer scientist, and his colleagues proposed at the Dartmouth conference in 1956 that “letting machines achieve this type of behavior, that is, doing the same thing as humans,” can be called artificial intelligence. In the following 60 years, artificial intelligence experienced “three ups and two downs,” three times experiencing a rise, and two times falling into a valley. In addition to the continuous evolution of the direction of the technology itself, artificial intelligence has also gained many different levels of meaning due to the flexibility of interpretation. Before AlphaGo defeated Lee Sedol and Ke Jie, most people’s impressions of artificial intelligence were limited to what they saw in the movies. For decades, a series of films such as Artificial Intelligence, The Matrix, Her, and The Incredibles have described humans’ yearnings for and fears of “artificial intelligence.” The concept of artificial intelligence is not only scientific common knowledge, but also a form of popular and commercial culture. The gap in understanding between a small group of AI experts and the public users of this “black box” technology is growing. So, during today’s artificial intelligence renaissance, do we understand what AI means? What are its capabilities and limitations? Compared to the past, has the meaning of artificial intelligence changed?
Gaps in Understanding About AI
In order to assess people’s understanding of artificial intelligence, Tencent Research Institute conducted an online survey from May to June 2017. Through the Tencent Questionnaire platform, we sent questionnaires to different groups directly or indirectly related to artificial intelligence, such as R&D personnel, technical personnel, product personnel, and researchers in the legal, policy, humanities, and social science domains. We received a total of 2968 responses from people from all walks of life. According to the survey data, the following questions were answered in turn: How do different groups of people understand and conceive of artificial intelligence and do differences exist? How do people accept and trust artificial intelligence in different areas? What issues need to be paid attention to in the process of artificial intelligence research? Are managers aware of the capabilities and limitations of artificial intelligence?
Among the surveyed respondents, the ratio of male to female was about 2:1, and the overall level of education was relatively high (see Table 1.1). Among them, 11.9 percent of people were engaged in occupations directly related to artificial intelligence, 45.7 percent of people were engaged in occupations indirectly related to artificial intelligence, and 42.4 percent of people were engaged in occupations unrelated to artificial intelligence. Practitioners directly or indirectly engaged in artificial intelligence included scientists, technical personnel, product and design staff, law and policy practitioners, humanities and social science researchers, media professionals, and entrepreneurs. We recognize the limitations of the above data and do not attempt to infer the overall situation in China.
Table 1.1
Composition of respondents
Variable | Percentage | |
|---|---|---|
Sex | Male | 67.5 |
Female | 32.5 | |
Education level | Undergraduate | 50.7 |
Master’s | 37.2 | |
PhD | 3.8 |
The survey covered five important topics in the field of artificial intelligence: the understanding and recognition of artificial intelligence, future predictions about artificial intelligence, trust and acceptance of artificial intelligence, threat of artificial intelligence, and legal and research responsibilities for artificial intelligence.
Understanding and Recognition of Artificial Intelligence
This section mainly analyzes people’s impressions and understanding of artificial intelligence. In the eyes of the people, what does artificial intelligence mean? In this survey, we did not predefine artificial intelligence in a general or in a narrow sense. Instead, we asked the public a wide range of questions about their first impression of artificial intelligence, their understanding of existing achievements, and their imaginings of the future.
Impressions of AI: When “Artificial Intelligence” Is Mentioned, What Do You First Think Of?
More than half of the respondents mentioned AlphaGo and robots. Common words also included “self-driving cars,” “terminator,” “Siri,” and “big data.” When talking about artificial intelligence, people often confused it with the concept of robots. But the current wave of artificial intelligence is more about the flourishing of deep learning algorithms based on big data. It cannot be equated with previous attempts to create “Artificial General Intelligence” restoring robot forms with human intelligence and behavior.
What Capabilities Does the AI Already Have?
“Artificial intelligence” is a collective term for a group of technologies. To understand the capabilities already possessed by AI, it is necessary to understand the development of artificial intelligence in the current technical field and the problems it can solve, instead of seeing artificial intelligence as a type of general ability. For example, decision-making capabilities involve reinforcement learning. Creativity refers to generative models associated with creation, and will have very good applications in the field of content creation. Affective computing research is trying to create a type of computing system that can perceive, recognize, and understand human emotions, and can make intelligent, sensitive, and friendly responses to human emotions, that is, giving the computer the ability to observe, understand, and generate various emotional characteristics like humans. At present, relevant research has made some progress in facial expression, posture analysis, and emotion and recognition aspects of speech. The machine understands your emotions, but this does not mean that it will have “empathy” like humans.
Future Predictions About Artificial Intelligence
Artificial intelligence has entered a period of rapid development. With the application of artificial intelligence blossoming in all sectors, a society in which human beings and artificial intelligence co-exist harmoniously with each other is nearing ever closer. What kind of artificial intelligence society will we usher in?
Will AI Be Commonplace in Society After Ten Years?
Overall, 47.8 percent of respondents believed that artificial intelligence will be commonplace after ten years. After further analysis, the closer a respondent’s connection to artificial intelligence, the more likely they were to believe that artificial intelligence would be commonplace in the next ten years.
Will Artificial Intelligence Have a Positive Impact?
The survey results revealed that respondents were optimistic about the impact of artificial intelligence on society.
The more knowledgeable a respondent was about artificial intelligence, the more likely they were to believe that artificial intelligence will have a positive effect.
Among the interviewees who identified as being “very knowledgeable” about AI, 82.63 percent agreed that artificial intelligence will have a positive impact on society; among respondents who chose “not much understanding” of artificial intelligence, only 59.30 percent believed that artificial intelligence will have a positive effect on society. Among respondents who had used artificial intelligence products, 73.38 percent thought that artificial intelligence will have a positive impact on society. Among respondents who had not used artificial intelligence products, the figure was 9.1 percentage points lower, at 64.28 percent. Lack of understanding and even misunderstanding of artificial intelligence may cause people to fall into an “ignorant fear” of artificial intelligence.
Will Artificial Intelligence Develop Consciousness?
Consciousness is the most magical mental ability of human beings, and it is also a very mysterious and complex phenomenon. Since the 1990s, many philosophers, psychologists, and neuroscientists have started to study what is known as “machine consciousness.” There are two distinct viewpoints on the existence of phenomenal consciousness.2 One is the mystical view that the only thing common to our neurobiological systems is the subjective experience. This kind of phenomenal consciousness cannot be reverted to a physical mechanism or logical description. It cannot be grasped by the human mind. The other is an eliminativist view that the machine is just a zombie. In fact, there is nothing but machines, objects that cannot have any subjective experience. The argument for machine intelligence itself includes people’s different understanding of consciousness. For the ultimate goal of achieving artificial intelligence, consciousness is a problem that cannot go unnoticed. If “artificial general intelligence” becomes possible in the future, it must be accompanied by the emergence of “machine consciousness.” For the current round of artificial intelligence based on machine learning, this is still a relatively distant research direction.
AI Trust and Acceptance
Acceptability is the key to implementing artificial intelligence. User trust in artificial intelligence systems is the prerequisite for artificial intelligence systems to produce social benefits. Secure and steady trust requires constant trial and error. Trust requires a system of practice that helps guide the security and ethical management of artificial intelligence systems. This includes coordinating social norms and values, algorithmic accountability, compliance with prevailing legal norms, as well as ensuring the integrity of data algorithms and systems, and protecting personal privacy.
In Which Areas Do You Want to Use Artificial Intelligence?
The results of the survey showed that respondents wanted to use artificial intelligence in smart homes, transportation, elderly/child care, and personalized recommendations.
Level of Acceptance in Nine Major Fields: Are We Prepared?
According to observations of the current landscape of enterprises and research in artificial intelligence, we selected nine common application areas: autonomous driving, virtual assistants, research/education, financial services, medicine and diagnostics, design and artistic creation, legal practices such as contracts and lawsuits, social companionship, and services and industry. Respondents were asked to answer regarding each of the nine areas to what extent they could be handed over to “artificial intelligence”: (1) humans do it themselves; (2) mainly human beings, with support from artificial intelligence; (3) mainly artificial intelligence, with human supervision; (4) artificial intelligence replaces people; (5) unclear.
Areas of high acceptance of artificial intelligence include: services and industry, autonomous driving, financial services, and virtual assistants. Forty-two percent, 41 percent, 41 percent, and 40 percent of respondents, respectively, believed that artificial intelligence should be doing the bulk of the work, with human supervision. Especially in services and industry, 40 percent of respondents believed that artificial intelligence can replace people.
Areas where AI acceptance is relatively low include: research/education, medical and diagnostics, social companionship, and legal practices such as contracts and lawsuits. Fifty-seven percent, 49 percent, 43 percent, and 39 percent of respondents, respectively, believed that humans should do the bulk of work in these fields, supported by artificial intelligence.
The lowest level of artificial intelligence acceptance is in the field of design and artistic creation. Forty-seven percent of respondents believed that humans should perform the tasks in these fields themselves. Only 4 percent of respondents believed that artificial intelligence could replace people in this field.
According to people’s answers to the above questions, it is easy to draw a conclusion that meets the beliefs of the public; that is, the higher the degree of mechanization in the work, the more people want it to be completed by artificial intelligence. For work that requires creativity, people are more confident in human capabilities.
The fact is that, unlike the previous wave of automation that only affected mechanical labor, artificial intelligence has increasingly appeared in research and art. At the end of 2016, Sony released a pop song “Daddy’s Car” created by artificial intelligence. The track was created by Flow Machines, a Sony Computer Science Laboratory artificial intelligence program, and involved discovering special style by analyzing a database with a large number of songs. Artificial intelligence has been able to create poems and songs. In the field of art and creation that people thought could not be replaced by machines, the trend of man-machine integration has gradually emerged. However, Flow Machines head Pacht said that although artificial intelligence can now create “perfect” songs, only musicians can create unique works.
AI Interaction Mode: “Natural Language Communication” Has Become the Preferred Mode of Human-Machine Interaction
Every technological revolution drives the evolution of methods of interaction at the same time. With the rapid development of language recognition technology and natural language processing technology (NLP), speech recognition has gradually become a common interaction mode of intelligent machines. Some analysts estimated in a report that by 2020 dialogue between ordinary people and machines will exceed dialogue between spouses. The report does not indicate whether the reason is the increase in dependence on AI technology or the deterioration of future spousal relationships, but it may also be a combination of the two factors. At present, the transition from “screen operation” to “chat interface” in electronic devices has become a trend. A number of players and products have emerged in the field of voice interaction. There are Amazon Alexa, Google’s Google Assistant, Tencent Cloud’s Xiaowei, and Baidu’s Duer. These products use dialogue as an interactive method to control different smart devices. All technology companies are accelerating this transition and are striving to enter the next generation of artificial intelligence services.
The Threat of AI
When artificial intelligence gains territory in various fields, there are also various hidden concerns about AI. Some people worry that AI will largely replace human labor. Some people worry that the development of AI will not be controlled. Films such as Metropolis and Terminator have such arguments and they express a kind of fear. When a strong artificial intelligence system is created, perhaps its wisdom will be far beyond that of humanity, bringing some unimaginable risks. How sho...