Please give AI some tolerance.

"To be honest, I really don't like the word' artificial mental retardation'."

In the chat with Nuggets, algorithm engineer, who is engaged in computer vision, said many times that he hated the word for a long time, almost instinctively. Even if it's just ridicule, it seems to him to be ridicule.

This kind of ridicule is like a passer-by, sneering at the child who just learned to climb: the child is too stupid to even walk.

He even admitted that if a colleague around him used this word to laugh at himself, he would deliberately keep a distance from him, because such self-mockery was a "disrespect" for his work and professional knowledge.

There are not a few engineers who have his technical hobbies. Among the many employees asked by Nuggets, they all expressed similar views: when asked about the level of artificial intelligence, similar expressions were based on "weak artificial intelligence".

A commercial Commissioner in charge of brand communication and public relations revealed that if words such as "artificial mental retardation" are used in external communication, it will "directly affect performance appraisal" because this unprofessional expression is likely to lead to negative communication effects.

In conversations with these people, Nuggets found that in the AI circle, practitioners have a clear understanding of AI, and the negative expression of AI in propaganda is more rigorous.

However, outside the circle, various AI accidents have made the public have many doubts about the true ability of AI. There is a lot of talk about artificial intelligence becoming artificial mental retardation, and the voice of singing artificial intelligence is often seen in newspapers.

On the surface, this is just a public opinion dispute about AI. But its essence is the competition between enterprises and the public for the right to speak AI, which will directly affect the promotion, landing and application of AI.

"If the public cannot form an effective understanding of new technologies, the promotion of new technologies will be very slow." A graduate student of Communication University said that the public's ability to accept new technologies is progressive step by step, and this process is easily influenced by public opinion, and negative public opinion has a "blasting effect", which may directly destroy the previously established "trust foundation".

For example, the public trust in autonomous driving is very weak. After many accidents, this trust has actually been exhausted.

Relevant research reports show that the activation rate of Tesla FSD, a self-driving brother, is less than 10% in China, and even quite a few people have not opened AP service. Even among those who have opened it, few people will use the AP function.

Although there are objective reasons for this phenomenon (such as insufficient roadside data and limited algorithm ability), from the perspective of public opinion communication, one mistake of autonomous driving is more serious than ten accidents of traditional cars, which also hinders the further landing of autonomous driving.

Then, how to establish an effective public understanding of AI and promote the faster and wider landing of AI?

After the interview, Nuggets said that media reports, corporate propaganda and popular science education are the three most important ways. And all kinds of "cognitive education" around the public are doomed to be a protracted "tough battle."

There is an interesting paradox in the application of artificial intelligence: when an AI technology has become very popular, people generally don't think it is AI.

For example, in the 1980s and 1990s, a black-and-white TV set may be an epoch-making symbol, which needs manual frequency modulation. But now remote control color TV has become a standard, and people don't think it is intelligent. For another example, in recent years, community parking lots have popularized the recognition of license plates and brushed faces into communities, but people rarely associate it with AI, even though various recognition algorithms, chips and so on are actually used.

In the public's cognition, artificial intelligence should reach the level of robots in movies, or think and act like people.

"The public is sometimes too optimistic or even overestimated about artificial intelligence." Yang Li, an associate professor at the School of Information, China Metrology University and the head of artificial intelligence major, believes that as a new technology facing the society, people's understanding of AI is not comprehensive, and they think that AI should be omnipotent. This kind of cognition is inconsistent with reality.

In the Nuggets' view, the public's understanding of artificial intelligence is relatively shallow, mainly in two aspects:

This kind of shallow cognition can be easily induced. Under some irrelevant propaganda, it exaggerates the ability of AI itself and makes the public blindly "confident" or overestimate AI.

"The layman looks at the excitement, and the expert looks at the doorway."

Yang Li said that taking face recognition as an example, five years ago, people might think it was mysterious and advanced, but after the popularity of consumer electronics, many people felt that face recognition was no longer difficult. When he told the students about face recognition, the students all thought it was a very mature technology. "It's not new, it's not difficult."

But in fact, face recognition is still a long way from high intelligence, and it is difficult to capture effective face information in many complex scenes. Moreover, face recognition works well in small-scale (small database) scenes, but when the database is very large, the accuracy of recognition is not so high.

"Due to the lack of professional knowledge, it is easy for the public to simplify complex problems, but people engaged in AI research are very cautious about this. Ordinary people think that simple technology, practitioners may think' this won't work, that won't work', in short, it is the feeling of sighing at the mountains and being a living horse doctor. "

Nuggets found that due to the lack of professional general education, the public's understanding of artificial intelligence is relatively simple, and most of them come into contact with AI through media reports and corporate propaganda. Only a few people will spontaneously study relevant books and take courses to enhance their understanding.

From the perspective of communication, if the audience's access to information is restricted, then the controller of the information channel has the "control right" of information dissemination, forming a situation of "public opinion monopoly", and information is easily "distorted" after repeated dissemination.

In fact, this "distortion" is inevitable. In the process of the spread of artificial intelligence, two groups, inside and outside the circle, were formed. Because artificial intelligence itself belongs to a profession with a high threshold, the connection between the inside (enterprise) and the outside (ordinary audience) is mainly realized through the media.

However, the problem of media propaganda is that many practitioners are either trained or cross-border transformation, and few media people really understand AI. Moreover, with the changes in big data and Internet technology, the media itself has further sunk into various platforms, creating a large number of self-media and forming a mixed situation in the media industry. In the traffic-oriented environment, various news reports emerge one after another, which has a "magnifying effect" (for example, the headlines are too shocking), so that there is an "error" between the information received by the public and the actual information.

At the peak of artificial intelligence, many AI companies put advertisements, soft articles and products in order to gain financing and popularity, creating the illusion that artificial intelligence can already land on a large scale. Later, when AI was cold, the public's ridicule of AI can be regarded as a kind of "self-attack" that was too intense before.

Of course, the circle also noticed the limitations of mass media. Many companies have opened up publicity channels on important social platforms, but due to content differences (such as being too vertical and product promotion) or channel differences, they do not meet the C-end attribute. Most AI companies cannot directly establish effective contact with the public.

Therefore, under the communication chain of "enterprise-media-public", it is difficult for the public to establish an effective cognition of AI under the condition of uneven information due to the defects of the mass media itself. However, enterprises have to rely on mass media to promote AI, which is an important reason for the "cognitive difference" inside and outside the circle.

"In the final analysis, there are still too few AI talents." In Yang Li's view, talent is the core force to promote industrial development. At present, AI is in the climbing stage, and the problem of technology itself is the fundamental factor that leads the public to question AI. The spread of public opinion has aggravated this influence to some extent.

No matter the deep development or horizontal spread of AI, only AI talents can "correct their names" for AI, but at this stage, domestic AI talents are extremely scarce.

"There are really too few applied talents." Yang Li lamented that when AI went from castles in the air to fields, "there are really not many people who understand technology and industry."

In the Report on Talent Development in Artificial Intelligence Industry (20 19-2020) issued by the Ministry of Industry and Information Technology (hereinafter referred to as the Report), it is estimated that the effective talent gap in China's artificial intelligence industry will reach 300,000, which is only the data of two years ago. In fact, in the past two years, according to Nuggets' observation, the demand for talents in AI enterprises has continued to be strong, and the gap of applied talents in the entire AI industry has further widened.

As a technology/knowledge-intensive industry, AI has a high entry threshold for talents and attaches great importance to academic qualifications and work experience.

According to the report, only 1 1.9% of the positions released by AI enterprises in 20 19 have received college education; Only 5.4% positions accept job seekers with less than 1 year working experience; Only 3.3% of the positions accepted the invitation of fresh graduates.

This means that to engage in the AI industry, a bachelor's degree is basically required. At the same time, because most AI companies lack the manpower, funds and motivation to train fresh graduates (at least one year), the demand for fresh graduates is not strong, but they prefer those with knowledge reserves and practical experience. This "innovative" recruitment demand has aggravated the shortage of talents.

In addition, AI has strong professional requirements for talents, especially for positions such as algorithm research and application development. More than 60% positions require a professional background in computer and mathematics.

Under the constraints of various linear conditions, the already scarce AI talents are more "tight".

HR, an AI startup, told the Nuggets that it was difficult to recruit people. "After the screening of majors, schools and work experience, there are very few qualified people. In addition, what the company wants is people who can come in to produce immediately, so it is difficult to recruit outstanding talents. If you take the school move, excellent graduates will be signed by the Internet and star AI companies early, and the rest will be more inclined to big companies. Screening to screen, there are really not many choices. "

In addition to the lack of applied talents combined with the industry, in Yang Li's observation, another talent gap in AI is the theoretical research talents who can "settle down and do basic work".

According to the Artificial Intelligence Report 2022 published by Stanford University, although China ranks first in the world in the number of cited papers in AI journals, published conference papers and artificial intelligence patent applications, it lags far behind Europe and America in the number of cited papers in AI conferences. Moreover, some innovative basic theories and cutting-edge scientific and technological research are mainly in Europe and America.

"Many basic theories of artificial intelligence are put forward by foreigners/institutions, such as deep learning, which is very popular now."

Yang Li said that this has a lot to do with the late start of artificial intelligence in China. To make up for this gap, we should not only strengthen the investment in basic theoretical research, but also establish a standardized AI talent training system to provide continuous talent vitality for AI research.

"The school is the cradle of cultivating talents. Ideally, some students engage in theoretical research after graduation, and more graduates enter the industry to promote the landing of AI through Industry-University-Research linkage. "

Nuggets learned that at present, China's artificial intelligence industry has initially formed a "political and Industry-University-Research integration" talent training and health system, but it is still in its infancy. 20 19 artificial intelligence major was officially approved to be included in the list of undergraduate majors, and many domestic universities began to build their own artificial intelligence colleges (research institutes) or cooperate with enterprises to open AI majors.

However, major universities are still exploring how to cultivate professional AI talents, and have not yet formed an effective paradigm.

20 19 domestic artificial intelligence major was officially approved and included in the list of undergraduate majors. However, the establishment of specialty needs to go through the process of curriculum construction, experimental conditions and professional declaration. Most schools just started enrolling students in the past two years.

In other words, it will take about one to two years for the first batch of AI undergraduates to graduate.

It is not easy to train these freshmen to fill the current talent gap. In addition, whether the comprehensive ability of the first batch of graduates in the future meets the standard is also of great symbolic significance.

"On the one hand, the content of artificial intelligence is very difficult. Many courses offered at the postgraduate level are now taught at the undergraduate level, which is a pressure on students and a challenge to teachers' teaching methods and skills. On the other hand, it is difficult to combine talent training with social needs, so that students can apply what they have learned. "

As a senior scholar in the field of artificial intelligence, Yang Li not only has in-depth research and thinking on AI, but also explores some "methodologies" on cultivating AI talents during his many years of teaching career.

"First of all, we must respect the law of learning." Yang Li told Nuggets that AI itself requires high practical ability, so it can't copy the training mode of traditional disciplines, that is, freshmen focus on theory and juniors focus on majors. Instead, we should combine theory with practice, study first, then practice, learn in practice, and then "spiral up"

In terms of specific measures, he said that students can be encouraged to participate in various learning competitions and research projects in a team way by setting up a "science and technology group".

The advantages of this group model are: the group covers all students, and through teamwork, it forms an internal learning atmosphere of mutual assistance, so that all members can participate in practice and become "interest groups"; Moreover, the duration of the group covers the whole university career of students, and all members can enjoy "welfare". At the same time, mutual help among group members can also relieve the pressure on teachers to some extent.

"Secondly, we should teach students in accordance with their aptitude and stimulate students' curiosity and exploration of AI. "

Yang Li said that students' interest in learning AI also shows an obvious "28 law", that is, 20% of students have a strong thirst for knowledge, while 80% of students have an average interest.

"For these 20% students, you just need to tell him how to do the best, tell him the matters and details that need to be paid attention to in this process, and don't care too much about the rest; For 80% of students, their interest is not that high, so they need more detailed guidance and some "compulsory homework", such as directly assigning tasks for them to participate. "

Furthermore, through the incentive mechanism to stimulate students' creative inspiration. "

For example, in curriculum design, innovation should be included in the grading standard, and students' innovation should be driven by curriculum achievements.

For example, in a case, if students just follow the steps listed by the teacher, their highest score may only be passing, and the rest depends on their creativity and play.

"Most students need teachers to give them some push, and grades are the best motivation." Yang Li said that in order to get a higher grade point, students have to "spend more time" instead of perfunctory, and the final homework "often has many unexpected highlights."

"Finally, teachers and students should form a virtuous circle of benign interaction."

A common problem in undergraduate teaching is that the interaction between students and teachers is weak, or only exists in the classroom, with little extra-curricular contact. It is not uncommon for students and teachers to attend classes and passers-by to finish classes.

In Yang Li's view, if a teacher only regards teaching as a work task, then students will also take a coping attitude. On the contrary, if teachers have a sense of responsibility, students will be influenced by their "leading by example" and be more enterprising.

Therefore, teachers can communicate with students through projects and online and offline interactions. Understand students' needs and give feedback to their own teaching work, and this feedback will eventually reach students through teaching, forming a "win-win situation for teachers and students".

In addition to the methodology of cultivating AI talents, Yang Li also pointed out that the cultivation of artificial intelligence professionals should break the "graduate-only theory".

"Studying artificial intelligence must be a graduate student. If you don't go to graduate school, there will be no future. "

Many people hold this view, but Yang Li firmly opposes it. He believes that many postgraduate courses have been decentralized to undergraduate courses. After undergraduate talents are trained into a system, students' theoretical and practical abilities will be able to meet the basic needs of the AI industry. Blindly pursuing postgraduate education will only lead to more and more AI circles, which will not help alleviate the shortage of talents in the industry.

"Of course, graduate education is also very important, but the cultivation of graduate talents may be more inclined to basic theory, and the scale of AI needs more applied talents to promote it."

For example, many traditional manufacturing industries have introduced artificial intelligence, such as robotic arms and automated production equipment. However, due to the lack of applied talents, enterprises do not know how to use the bought equipment, how to maximize benefits, and how to operate and maintain it.

Such a position does not require practitioners to have a very deep theoretical foundation, but needs talents with an AI foundation and an understanding of the industry. In the process of intelligent upgrading of traditional industries, the similar talent gap is very large.

"In fact, when AI goes to all walks of life and lands, the demand for talents will also change. At the undergraduate level, outstanding talents can also be cultivated through theoretical study and social practice related to majors. "

At the just-concluded Winter Olympics, Professor Yang Li led the team to make an intelligent assistant technology, which can review and analyze athletes' movements through video and provide reference for judges to score.

Although it is only a simple behavior recognition, the model is not exquisite, and many AI companies in the market have the ability to develop this technology. But it is gratifying that once this project was put forward, students actively participated in it. Under the guidance of their tutors, they mine data, mark, model, train and test step by step. The whole process lasted for two weeks, and most of the work was done by students. Moreover, during the Spring Festival, some students even deeply regret that their efforts are not enough.

"Empty talk is worthless." In Yang Li's view, others have the ability to do this project, but only they put it into practice. The whole project was completed by freshmen. The process is far more important than the result. They "represent a new force in the field of artificial intelligence."

Doing this project is not always smooth sailing.

Jiang Zhengyang, a project member, majoring in 2 1 Artificial Intelligence from School of Information, China Metrology University, told Nuggets that when the team was modeling, either the network was too big and the training was too slow, or the network was too small to meet the requirements and it was difficult to achieve the expected goals. At the same time, training will also encounter insufficient computing power.

After many failed attempts, the team had to turn to Professor Yang Li, who added a network structure. Under this structure, the model became relatively "light" and the training could meet the expectations.

In the end, the team successfully developed the "snowboarding AI referee technology". This technology can accurately identify whether an athlete grabs the board in complex scenes such as blurred pictures, high-speed camera movement and long-distance panoramic pictures, thus providing a basis for judges to score and helping to "make the Winter Olympics fair".

"Our professional knowledge is limited and we need to continue to strengthen theoretical study. Through this project, we learned about the process, method and difficulty of starting a project from scratch, and accumulated experience. Of course, I am still very happy to finally see the results of the project. " Jiang concluded.

Yang Li believes that it is normal to encounter problems, and the key lies in action and practice. "People will fall down many times on the way to learn to walk, but you can't just learn to climb because of the fall, so you will never walk."

This is not the epitome of domestic AI development.

After a period of no one cares about it, domestic AI began to flourish in 10, and a number of AI companies such as Shang Tang, Shi Kuang, Congyun and Yi Tu were born one after another, which were warmly welcomed by the capital and propped up the hope of domestic AI. However, after the burning of passion, it is followed by various problems such as difficulty in landing the industry, difficulty in commercialization, and difficulty in realizing it.

Today's AI is in the groping period from climbing to walking, and bumps, falls and falls occur from time to time, and it is also ridiculed by the public as "artificial mental retardation".

However, Yang Li is not depressed, but optimistic, because "more and more enterprises and talents are involved in the R&D, promotion and landing of AI", and under the impetus of the "political Industry-University-Research" model, AI will also be unveiled to show its true face, and the public will form a "comprehensive and objective" cognition of AI in the future.

In the process of communication between Nuggets and many AI practitioners, almost everyone is full of hope for AI. Even though AI is still in the stage of "weak artificial intelligence", they still firmly believe that AI has a bright future.

"The vast ocean of artificial intelligence is not just angular, but changes the world." The engineer who bluntly said "artificial mental retardation" from the beginning told Nuggets that even though the road to changing the world was full of bumps, "I persisted because I loved it."

And for some ridicule and doubt of the public, he hesitated and replied:

"Please give AI some tolerance." Leifeng. com