How much potential does artificial intelligence have to be tapped in the field of education?人工智能在教育领域究竟有多少潜能待被挖掘？
As the highest award of the 2021 World Artificial Intelligence Conference (WAIC2021), the results of the SAIL Award were announced yesterday. Five projects, including a "robot scientist" who can independently complete scientific research tasks, and a flexible brain interface that can stably link biological brains to computers for a long time, won the first place in one fell swoop.
This year's SAIL Awards, its review committee gathered influential scientists and entrepreneurs in the field of artificial intelligence (AI) from around the world, and is committed to setting a benchmark for artificial intelligence to lead the future development. According to reports, this SAIL Award takes AI-enabled digital transformation of cities as the starting point, tracks technological development, industrial empowerment, and governance processes, and pays special attention to "breakthrough innovations" to encourage representative directional technological breakthroughs and application innovations, ongoing or AI projects that will change life in the future continue to emerge. The five award-winning projects well reflect this orientation, refreshing everyone, and providing a rare window to explore the potential and charm of AI's future development.
Robot scientist: It invented new materials
Traditionally, manipulators and robots have often been found alongside factory assembly lines and have been used to replace industrial workers. However, "robot scientists" at the University of Liverpool in the United Kingdom have changed this, giving robots a new mission to enable them to carry out chemical research and prepare new materials autonomously.
This is a mobile robot with an AI intelligent brain. Compared with traditional robotic arms, its "hand" and "brain" functions are more balanced, so it can perform more complex tasks. It is deployed in traditional laboratories and can solve highly complex materials research topics. To be precise, the mission of robot scientists is not to "automate instruments", but to "automate research". With a flexible robotic arm and custom grippers, it can handle a variety of laboratory equipment; it can safely share a room with human researchers, and it will move and avoid obstacles on its own. More importantly, based on a specialized algorithm model, robot scientists can automatically advance research, such as optimizing the initial chemical formula. In one project, it finally screened out a photocatalyst with 6 times higher activity.
The emergence of robot scientists will help promote the paradigm shift of laboratory automation, showing the great prospects of the new research and development model of "AI + new material synthesis". Some analysts believe that this invention is expected to affect the global research and development market of more than 1.2 trillion US dollars per year.
Flexible brain-computer interface: facing the "three major problems"
Brain-computer interface, in simple terms, refers to creating a direct connection between the brain of a human or animal and an external device, so as to realize the information exchange between the biological brain and the electronic device. Brain-computer interface is a hot field of current AI research. It is the core key technology for recognizing important brain functions and diagnosing and treating serious brain diseases. It is of great significance for my country to promote the frontier research of brain science and grasp the initiative of neurological diseases and AI research.
The "Craniotomy-Free Minimally Invasive Implantable High-Throughput Flexible Brain-Computer Interface" project, which won the SAIL Award, was completed by the research team of the Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences. For the three major problems faced by the brain-computer interface - "insufficient bandwidth", "major implantation trauma" and "poor stability of long-term recording", it has given innovative technical solutions. In terms of bandwidth, the team has realized a single-electrode device integrating a 2640-channel brain-computer interface device; in terms of trauma, it does not require a craniotomy during implantation, and only leaves a small hole of no more than 1 mm; plus this team's innovative research and development The flexible implant and flexible electrode technology can avoid damage to brain tissue to the greatest extent, and extend the continuous stable working time in the brain from about 2 months in the past to at least 8 months.
In order to meet the needs of practical applications, the project team is also speeding up the improvement of other supporting technologies while tackling the flexible nerve electrode array, including the development of robots that can automatically complete the minimally invasive implantation of flexible electrodes, and large-scale neural signal encoding and decoding technology. The team hopes to carry out human brain verification as soon as possible to benefit patients with high paraplegia, gradual freezing disease, aphasia, etc., so that they can use "ideas" to control robotic arms and keyboards, synthesize voice, and promote a better integration of human brain and AI.
Data flow AI chip: the first "Chinese design"
CAISA, developed by Shenzhen Kunyun Information Technology Co., Ltd., is the world's first commercial data stream AI chip. Based on the self-developed architecture, it has greatly improved chip utilization efficiency, and achieved industry-leading indicators such as computing power performance-to-price ratio, measured performance, and processing delay. At present, a variety of data stream AI computing platforms developed based on this chip have been applied in smart safety supervision, smart oil fields, smart manufacturing, smart grid and other fields, greatly reducing the cost of implementing AI applications for customers, and providing support for the underlying AI computing power. The new option fills the technical gap in the field of domestic data stream AI inference chips.
Intelligent design engine: AI learns piano, chess, calligraphy and painting
Qin, chess, calligraphy and painting are the traditional "four arts" of our country. In terms of music synthesis, Wenping chess, etc., AI is quite impressive, but can it do calligraphy, painting, and seal cutting? The "Digital Creative Intelligent Design Engine" developed by Zhejiang University endows AI with more artistic flavor, which is expected to further detonate a large number of new applications, and will explore more landing scenarios.
This research has built a high-quality library of traditional cultural basic materials, including traditional Chinese painting (more than 1 million pieces), seal cutting (more than 300,000 seals), calligraphy (more than 70,000 characters), folk music (more than 10,000 pieces), etc. ;At the same time, the team proposed the theoretical innovation of visual knowledge and AI multiple knowledge expression, so that AI can independently learn from the traditional art treasure house, thus building a digital creative intelligent design engine of traditional cultural elements, which has supported 7 intelligent design modules so far . Among them, "Moran" is an AI-assisted traditional Chinese painting creation module, "Yuyin" can automatically add music to short videos, "Dianshi" allows AI to learn seal carving, and "Lanting" realizes the original "Lanting Collection" based on the existing calligraphy master. sequence".
The only award-winning paper: Showing the face of China's AI
One academic paper won the SAIL award this year. This paper entitled "A New Generation of Chinese Artificial Intelligence" was published in the authoritative international journal "Nature Machine Intelligence". The main authors are three young AI researchers - Zhu Mingjie, Wu Fei, and Lu Cewu. This is the first panoramic overview of China's AI development. It introduces the formation and development of China's new generation of artificial intelligence development planning in a panoramic manner, and points out the challenges faced in basic research, personnel training, and ethics. The article describes The blueprint for the ecological development of artificial intelligence is being transformed into a vivid practice of China's current artificial intelligence technology development.
A virtual submarine might allow students to zoom out to the microscopic level to investigate natural processes taking place beneath the surface of a rock pool, or set the conditions for students to explore a nuclear power plant or even travel to ancient Rome or an outer planet.
"US 2010 National Educational Technology Plan" (NETP/2010 for short) reviews and summarizes the experience and lessons of the application of information technology in the enterprise sector in the past 30 years, and puts forward sharply that "if you want to see a significant improvement in educational productivity, you need to achieve Technology-enabled, major structural change in education rather than incremental tinkering."
At the same time, the "National Medium and Long-term Education Reform and Development Plan (2010-2020)" promulgated by our country gives a new positioning: "Information technology has a revolutionary impact on the development of education, and must be highly valued." The plan is to solve "how to Make information technology truly have a revolutionary impact on the development of education" to find a clear direction.
It can be seen that the deep integration of technology and education is reflected in the level of school education, mainly manifested in the deep-seated transformation of classroom teaching by AI.
The Potential of Intelligent Assistance Systems and Educational Robots to Realize the Technology Convergence Effect
The school education environment is undergoing subversive changes: since the 1990s, information technology marked by multimedia computers and network communications has been widely used in school education, entering our school classrooms, and today’s classrooms are no longer traditional. Consists of blackboards, chalks, desks, seats, books and pens, multi-functional/multimedia classrooms equipped with advanced technology include modern five big pieces and five small pieces.
However, the application effect of information technology in the field of education is not obvious. While the teaching environment is constantly enriched and updated by modern technology, it also faces an unavoidable teaching dilemma brought about by technology: such as redundant teaching content, which is not conducive to Students digest; the teaching content is presented page by page, which splits the internal structural connection; once the program controlled by the software is locked, the teaching will be severely restricted; the fixed configuration of the seat leads to a single teaching activity; the equipment is updated slowly, and the technology is difficult to meet the students' inquiry needs ; The limitations of the electronic whiteboard itself lead to the lack of interaction between teachers and students, and between students and students.
In the past, "information technology" was only regarded as a tool and method, and most of them still stayed in the improvement of educational means and methods. In terms of educational efficiency and benefit influence, information technology seems to be dispensable and icing on the cake s things. We can't help but ask:
From the teacher's perspective, is the application of educational technology in the classroom really effective in supporting teaching? With the support of technology, can teachers free up time to optimize teaching and learning, and can they feel whether students are benefiting from it?
From the perspective of students, does technology really support learning effectively, and can students deeply engage in the learning process and gain positive emotional experiences such as curiosity, awe, and satisfaction? The fusion effect of technology and education is worth looking forward to.
In fact, the essence of modern educational technology is not just "what teaching to use", but also how to teach, that is, to use modern advanced information technology to optimize the educational and teaching process to achieve the triple goal of improving educational and teaching effectiveness, efficiency and effectiveness.
In terms of specific manifestations, in addition to the promotion and improvement of the educational environment, and the enrichment and optimization of teaching resources (including various learning tools and software), the most critical thing is to realize educational ideas, teaching concepts, and teaching methods in an information-based teaching environment. And a fundamental change in the structure of classroom teaching. The AIED (Intelligent Assistance System/Educational Robot) technology has the characteristics of flexibility, inclusiveness, personalization and effectiveness, which will be effectively combined with education and teaching to the greatest extent, and specifically exert its potential in the following fields:
AI will unleash teacher productivity and act as a catalyst for changing the role of teachers
As we all know, due to technical limitations in the field of expert systems/knowledge models, early information and communication systems have been limited in subject resources and knowledge fields, and can only be applied to natural subjects with clear knowledge and logical structures, such as low-level mathematics and physics.
With the development and application of AI (artificial intelligence) core technologies such as knowledge representation methods and natural language systems, the field of subject resources has been expanded, and the development of expert systems in social subjects such as history and language has made great progress in knowledge construction. The teaching system will gradually include applied disciplines such as electronic engineering and management, and the intelligent teaching system will also be more widely used in education, training and vocational education.
In the transformation of teachers' roles, the increase of artificial intelligence technology should become a catalyst for the transformation of teachers' roles. Traditional teaching links such as lesson preparation, classroom lectures, Q&A tutoring, and homework correction will no longer be exclusive to teachers: for example, tutoring Q&A tasks can be performed by virtual agents. Instead, assignment marking can be supported by a learning partner or system. The intelligent assistance system/educational robot can help teachers free themselves from tedious and time-consuming repetitive tasks by taking on some of the tasks of teachers, and devote more energy to innovative and inspiring teaching activities.
In terms of technical support and cognition, with the continuous maturity of artificial intelligence technology, teachers can not only get timely support from technical personnel, but also have a deep understanding of the functions of artificial intelligence systems, and make reasonable evaluations and judgments on new artificial intelligence products. At the same time, the teacher's ability structure will also be deconstructed and reconstructed. The application of AI technology will stimulate teachers' research ability, raise the most valuable questions related to data, and guide students to use data analysis methods.
In terms of management and collaboration, new teamwork and management capabilities will be cultivated. In addition to the common teaching assistants, each teacher will be equipped with an artificial intelligence assistant, and the collaboration will take on the repeatable, procedural, memory-based, and repetitive tasks in the teaching process. In the teaching module of practice, teachers will focus on emotional interaction, personalized guidance, and the development of creative thinking, so as to continuously empower teachers' efficiency.
Artificial intelligence technology will stimulate students' learning potential
The new generation of learners who grew up in the digital environment put forward higher demands on the learning environment. The fixed, single and limited learning space and scene will be broken, and the intelligent assistance system/educational robot will make mobile learning and real-life learning possible.
Artificial intelligence can use simple methods to enhance virtual worlds, allowing it to interact with students and respond to students' learning behavior requirements in a more natural way. Numerous studies have also shown that immersion in intelligent virtual reality can improve education, allowing students to construct a personal understanding of the world they are exploring.
The application of intelligent assistance systems/educational robots will continue to innovate learning methods in the process of continuously expanding knowledge fields and types of knowledge, and new learning modes such as collaborative learning and gamified learning will become possible.
For example, artificial intelligence can act as an intelligent virtual agent, that is, play the role of teacher, expert, or mentor or learning companion in learning "tasks", accessing professional expert systems, so that all learners receive the learning that best suits their needs Resources; artificial intelligence technology can also use the information of individual participants to form a learner community that is most suitable for completing specific collaborative tasks in the form of a learner model. It is also possible to design game-like learning scenarios, as the story unfolds in the game, obtain information from the game and the player, and based on this information, use artificial intelligence algorithms to determine which appropriate learning actions to take.
AIED (Intelligent Assistance System/Educational Robot) can match learning activities with learners' cognitive needs and emotional states by imitating learners' cognitive and emotional states, ensuring deep involvement of students in the learning process; not only stimulating learning interest, Maintain curiosity and be satisfied by learning achievement, while promoting the positive transformation of negative emotions (frustration, confusion, confusion, abandonment) to positive emotions (curiosity, focus); intelligent assistance system/educational robot uses dialogue to engage students in Socra A grounded, open-ended learning model that promotes reflection and self-awareness; uses metacognitive scaffolding to provide dynamic assistance to increase learner motivation and engagement; uses social simulation models to enable students to relate to the target language through understanding cultural and social norms Use objects for fuller exposure.
One day in the fall of 2018, Ashtosh Garg, CEO of recruiting service Eightfold.ai, came across a resume that intrigued him.
The master of the resume is an up-and-coming data scientist. He focuses on mining patterns in data to help companies make decisions, such as how to accurately target advertisements. Interestingly, the term "data science" doesn't appear on his resume.
Instead, his résumé reads: UCLA graduate in physics, working as an analyst at Barclays. His LinkedIn profile on the professional networking site said he had never been a data scientist, but Eightfold's software flagged him as a suitable candidate. He closely resembles the four data scientists whose software Garg specified as a template in key areas of math and computer skills.
Garg says the core idea behind such selection is to focus on "what skills they have," not job titles. “That way, you can find people who are competent enough without the corresponding real-world experience.”
The technology will clearly appeal to employers eager to fill jobs amid a tight labor market. For example, data scientists are the talents that companies such as Google, Facebook, and Amazon are vying for.
Now, recruiting services such as Eightfold rely on complex algorithms to match workforces with jobs. As a result, many employers may soon have access to a large number of potential candidates and may even hope to fill difficult roles.
Columbia University economist Bur Cogier has done a special study on the application of artificial intelligence in recruitment. Algorithms can help some job seekers, but there is a potential downside: It could lead to lower wages, he said.
"Employers seem to be able to get unusually good candidates who perform equally well without having to go all out," says Cowgill.
Employers and online agencies have been using algorithms in talent acquisition for years, but often in a crude way: Computers identify keywords in resumes that match the text of job descriptions.
This is more efficient than doing a manual search on sites like LinkedIn, but it also has an obvious downside: Candidates can manipulate the screening process by inserting words into their resumes that a computer might search for when matching them. Conversely, poorly worded descriptions can cause computers to overlook qualified candidates.
But now, recent advances in artificial intelligence in deep learning techniques have made computers more powerful than ever. It goes beyond simply matching words to being able to identify skills and talents that a job seeker doesn't explicitly list on their resume.
Garg gave an example. For example, Eightfold found through data analysis that about 90% of software engineers at Intuit financial software company in Mountain View, California, mastered the programming language Java. That means the computer's conclusion that a software engineer at the company has Java skills is reliable, even if he doesn't put it on his resume.
By doing a similar operation on the entire resume, the software can build a detailed profile of the job seeker. It can extract normally unnoticed information from items such as education and hobbies, such as the high ability of certain majors in non-famous universities (such as the natural language processing major at the University of Massachusetts Amherst), chess Aces are more likely to be good at programming, and basketball players are more likely to be good at sales.
To improve its calculations, Eightfold cleverly struck a deal with its clients to use its HR software. The software imports employee data in anonymized form, including information about how employees from different backgrounds perform in different jobs and their earnings. As a result, Eightfold is better able to predict employee performance.
Such software tools are especially effective when employers are interested in broadening their search beyond candidates from traditional backgrounds. Recruiters can specify certain criteria (such as industry and location, and even the likelihood that a candidate will accept a job offer) to obtain a relatively unique resume.
Claudia Villanueva, a recruiter for AdRoll Group, a digital advertising and marketing services firm, said she recently used Eightfold when she was looking for candidates for financial planning roles. The top candidate had worked in an area the tech industry typically shuns, such as manufacturing, but Eightfold's algorithms decided she was the winner.
“The tech industry sometimes has a hard time breaking free from technical experience, so it keeps hiring the same people,” Villanueva said. “To me, this experience is an interesting example of how we can break out of the old mold a little bit.”
In the case of the Barclays analyst mentioned at the beginning of this article, Eightfold’s software sees him as a rising data scientist partly because, despite having no stated data science background, he has Master's degree, and mastered computer programming languages such as C++ and Matlab.
"We analyze tens or hundreds of millions of profiles to understand the progression of people's careers," says Garg. "You can easily predict what they might do next." New or unpopular domains with limited data do not work well.
The survey shows that only a few companies have adopted deep learning tools, and other companies even rarely express their willingness to explore. The University of Pennsylvania recently hosted a conference with nearly 20 employers. A post-meeting report cites a variety of problems, including the difficulty of assessing employee performance and the perceived unfairness of computer-based hiring, firing and promotion decisions.
Perhaps the biggest concern is actually a fundamental methodological issue, the skeptic's worry that the computer is actually only studying people who are already working for a company or have come to apply for a job, rather than the general population. Cogier believes there is a solution, and cutting-edge technology companies are using artificial intelligence to fill job vacancies.
According to Garg, the advantage of services such as Eightfold is that it has access to many companies' internal data.
"We're able to build a much deeper model than a single company could build on its own," he said.
I saw a piece of news two days ago, and I had mixed feelings in my heart.
An uncle cleaner walked into the mobile phone store and said with a shy face that he wanted to buy the latest Apple mobile phone.
The clerk saw that the uncle was dressed in ordinary clothes and didn't look like he needed a brand-name mobile phone, so he kindly asked.
The uncle said that the mobile phone was bought for his son.
Because of this matter, the father and son quarreled for a long time, and finally he compromised.
In front of the clerk, the uncle reluctantly took out the money.
Tremblingly, he took out some hundred-yuan bills from his pocket, and the rest were all small change, even coins.
The clerk couldn't bear it, so he called and communicated with the uncle's child, asking him to change to a cheaper old model.
But the child insisted that it must be the latest model.
The father usually saves money for the sake of the child, but the child does not understand the father at all.
I feel chilled thinking about it.
In the comment area of the video, some people accused the child of being a white-eyed wolf, while others said that the father lived too humblely.
But one of the comments made me think deeply: In fact, why does this father need to raise children no matter what the conditions of the family are.
That's right, this uncle only has a salary of 3,000 yuan a month, and after daily expenses, there is not much money left.
But this mobile phone, which costs tens of thousands, is bought as soon as it is said.
This reminds me of a passage from Shanghai Academy of Sciences:
"The love that Chinese parents give their children is not too little, but too much."
Many parents in life, obviously not rich, tighten their belts to give their children a good life;
It is obviously very difficult, but there is still no bottom line to meet the unrealistic demands of the children.
In the end, the parents who tried their best gave birth to a selfish and lazy giant baby.
02
A reporter from Xinhua Daily Telegraph once interviewed a poverty-stricken county that had just taken off its hat.
The reporter found that most families in the county have the following attitudes towards their children:
"Do everything possible to create a comfortable and good living environment for children, give children what they want, wear brand-name clothes, use trendy electronic products, and use mobile phones to chase stars every day..."
But in the end, the children turn a blind eye to the parents' hard work; the parents give desperately, but the children will never be satisfied.
That's how it was at my cousin's house.
The cousin and his wife are just ordinary working class, but the daughter goes to an aristocratic school with a tuition fee of more than 100,000 yuan a year.
The couple usually save money on food and expenses, and are reluctant to change their clothes for several years, but they buy sneakers worth a thousand yuan for their children.
But they bothered about their "rich" daughter, but they fell into a vicious circle of comparisons, blaming their parents for their incompetence, unlike their classmates' parents who allowed her to live in a luxury house and have a driver to pick her up when she went out...
This domineering daughter caused the young couple to complain endlessly.
Educator Mikawa Ling once warned parents: Don't mold children into another class in the family.
Blindly squeezing yourself to give will only result in children asking for more.
A 25-year-old unemployed youth in Fuzhou, who has been idle all year round, wants to eat and drink well. If he has no money, he asks his family for it.
His mother was there for him.
Later, he owed foreign debts due to gambling, so he forced his mother to sell the house.
The mother sold the big house and replaced it with a smaller one. The son, who was short of money, chased after him not long after, forcing her to continue selling.
When his mother refused, he beat and kicked her.
This heartbroken mother fed her son with bran and swallowed vegetables, but turned her son into a cold-blooded snake.
Look at this old lady in Weifang, Shandong.
After her wife died, she and her son lived together.
For more than 20 years, when her son did not work, the grandmother let him stay at home, spending only 3,000 yuan of her pension.
She is reluctant to buy even meat on weekdays, but when her son asks her for money to buy game equipment, she gives it.
This son who was "raised by her rich" has achieved nothing in his forties, and spends all day at home either playing games or sleeping.
The poor old lady is over 70 years old, and she still has to run around on the street to find a job for her son.
There is also the "most popular old man" Li Wenhong.
At the age of 17, he gave up the college entrance examination and "engaged in creation" at home. His talent could not support his dream, and he started a life of gnawing on the old for more than 30 years.
Facing the overwhelming abuse on the Internet, his parents tried their best to defend him:
"It's only natural for my son to spend our money, so it's your turn to say it?"
Psychological counselor Wu Di once said: Many young people nowadays have such a mentality. Their family background is not superior, but they are lazy and vain.
And the root of all this lies in the parents.
Parents indulged their demands when they should teach their children to be grateful, and chose to arrange for them when they should be independent.
This kind of education method feeds the vanity of the children and takes away the sense of responsibility of the children.
03
Li Bo, a professor at Nanjing Normal University, once said:
"The real education is that no matter how rich you are, your children suffer.
Who do you see who doesn't experience the ups and downs of life? The child needs to have normal tempering, not just give him all kinds of excellent conditions. "
A good education is never to let children soak in a honeypot, but to let them recognize the truth of life as early as possible.
Let the child feel the difficulty of his parents, and sharpen himself in life, so that he can know how to struggle and dare to take responsibility in his own practice.
Lin Shujie, a boy from Zhejiang, and his parents run a processing shop with supplied materials.
Since Lin Shujie was in junior high school, his parents would ask him to help in the store every weekend.
On the one hand, it is to let him share the pressure for the family, and on the other hand, it is to let him feel the difficulty of life.
Because of watching his parents work hard all the year round, Lin Shujie is more sensible than ordinary children.
In life, he is his parents' right-hand man, always rushing to do the dirty and heavy work.
In terms of study, he set precise work and rest time for himself, and even took time to help his sister with homework.
Until the day he received the admission letter from Fudan University, he was still helping to drive screws at his parents' processing site.
When facing the interview, he only had a simple sentence, "Parents are working too hard, I just want to help them reduce their burden."
Simple and hard-working parents can cultivate self-reliant and grateful children.
Peng Yong, the "father of pancake filling in Peking University", asked his son Peng Xiangyu to help him keep the stall every holiday when he was very young.
In addition, his father let him learn to do everything from sweeping the floor, washing the stove, and arranging ingredients.
My father often told Peng Xiangyu that we are just an ordinary family and cannot compare with those wealthy families, but our destiny is in our own hands, and we have to rely on our own hands to earn what we want.
Peng Xiangyu kept his father's teachings in mind.
He not only helped his parents share a lot of housework, but also studied very hard.
In the end, relying on steadfast diligence and extreme hard work, I opened the gate of Peking University.
Yi Shu once wrote in "Golden Years":
Rich people are prodigal sons, and poor families study hard to find talents. In fact, it is not necessarily true, it still depends on tutoring and experience.
Compared with material enrichment, children's quality and personality are more worthy of parents' effort to cultivate.
Just like the parents of Lin Shujie and Peng Xiangyu, they don't have much culture, but they understand the simple truth that "the beloved is so unfortunate, but the young is hard to become a talent".
They never hide the reality that they come from ordinary families, and they don't give their children a life beyond their ability.
Instead, restore the true colors of life, let children stand shoulder to shoulder with them, and strive to create a better future.
Children educated in such a family can be more independent and diligent, and become the leader of their own destiny and the leader of a family.
The 25 Best Ideas feature report published by Business Week begins with the following sentence:
"Treat your children well, but don't give them too much possessions."
When a child grows up, if he gets used to getting something for nothing, he will never learn to be grateful, nor will he have the motivation to fight hard.
Successful family education is not about giving children rich food and clothing, but teaching them to support themselves.
Teaching children to be independent and struggling, and raising a grateful, independent and progressive child with spiritual enrichment is the deepest blessing of a family.
作为2021世界人工智能大会（WAIC2021）的最高奖，SAIL奖结果昨天揭晓，能够独立完成科研任务的“机器人科学家”、能长期稳定链接生物脑与计算机的柔性脑接口等5个项目一举夺魁。
今年的SAIL奖，其评审委员会集聚了全球人工智能（AI）领域具有影响力的科学家和企业家，致力于为人工智能树立引领未来发展的标杆。据介绍，本次SAIL奖以AI赋能城市数字化转型为切入点，追踪技术发展、产业赋能和治理进程，尤其关注“突破性创新”，以激励代表方向性技术突破和应用创新、正在或将要改变未来生活的AI项目不断涌现。最终获奖的5个项目很好地体现了这样的导向，让所有人耳目一新，为一探AI未来发展的潜能和魅力提供了难得的窗口。
机器人科学家：它发明了新材料
传统上，机械臂和机器人常常出现在工厂流水线旁，并以替代产业工人为职责。不过，英国利物浦大学的“机器人科学家”改变了这一面貌，为机器人赋予全新使命，使其能自主开展化学研究，制备新材料。
这是一款具有AI智能大脑的可移动机器人，和传统机械臂相比，它的“手”和“脑”功能更加平衡，因此可以胜任更复杂的任务。它部署在传统实验室里，能解决高度复杂的材料学研究课题，确切地说，机器人科学家的使命，并不是“将仪器自动化”，而是“将研究自动化”。它带有一条灵活的机械臂和定制的抓取器，可以操作多种实验室设备；它能安全地和人类研究者共处一室，会自行移动并避开障碍物。更重要的是，基于专门的算法模型，机器人科学家能自动推进研究，比如对初始的化学配方进行优化，在一个课题中，它最终筛选出了活性高6倍的光催化剂。
机器人科学家的出现，有助于推动实验室自动化的范式转变，展现出“AI+新材料合成”这种新研发模式的巨大前景。有分析认为，这项发明有望影响每年超过1.2万亿美元的全球研发市场。
柔性脑机接口：直面“三大难题”
脑机接口，简单来说，是指在人或动物的大脑与外部设备之间创建直接连接，从而实现生物脑与电子设备的信息交换。脑机接口是当前AI研究的热门领域，是认知重要脑功能、诊治重大脑疾病的核心关键技术，对于我国推进脑科学前沿研究、把握神经疾病和AI研究的主动权意义重大。
此次获SAIL大奖的“免开颅微创植入式高通量柔性脑机接口”项目，由中国科学院上海微系统与信息技术研究所研究团队完成。对脑机接口面临的三大难题——“带宽不足”“植入创伤大”“长期记录稳定性差”，它都给出了创新的技术方案。在带宽方面，团队已实现单电极器件集成2640通道的脑机接口器件；创伤方面，它在植入时无须开颅，只会留下一个不超过1毫米的小孔；加上这个团队创新研发了柔性植入、柔性电极技术，可最大限度避免对脑组织的伤害，并将在颅内连续稳定工作时间从过去的2个月左右延长到至少8个月。
为了满足实际应用需求，项目组在攻关柔性神经电极阵列的同时，也在加快完善其他配套技术，包括研发了能自动完成柔性电极微创植入手术的机器人、大规模神经信号编解码技术等。该团队希望能早日开展人脑验证，造福高位截瘫、渐冻症、失语症等患者，使得他们能用“意念”控制机械臂和键盘、合成语音，进而推动人脑与AI更好结合。
数据流AI芯片：首款“中国设计”
由深圳鲲云信息科技有限公司研发的CAISA，是全球首款可商用数据流AI芯片。它基于自主研发架构，大幅提升了芯片利用效率，在算力性价比、实测性能和处理延时等指标上实现了业界领先。目前，基于该芯片开发的多款数据流AI计算平台，已应用于智慧安监、智慧油田、智能制造、智慧电网等领域，大幅降低客户的AI应用落地成本，为底层AI算力支撑提供了新的选择，填补了国内数据流AI推断芯片领域的技术空白。
智能设计引擎：AI学会琴棋书画
琴棋书画是我国传统的“四艺”。在音乐合成、纹枰对弈等方面，AI已颇具风范，但它是否会书法、绘画、篆刻？浙江大学研发的“数字创意智能设计引擎”，就赋予AI更多艺术气息，有望进一步引爆大量新型应用，并将挖掘更多落地场景。
该项研究建设了一个高质量标注的传统文化基础素材库，包括国画（100万余幅）、篆刻（30万余印面）、书法（7万余字）、民族音乐（1万余段）等；同时，团队提出了视觉知识和AI多重知识表达的理论创新，使得AI可以自主在传统艺术宝库中从师习艺，由此构建传统文化元素的数字创意智能设计引擎，迄今已支撑7个智能设计模块。其中，“墨染”是AI辅助国画创作模块，“余音”可自动为短视频配乐，“点石”让AI学会篆刻，“兰亭”则实现了基于现有书法母本倒推原始的《兰亭集序》。
唯一获奖论文：展现中国AI面貌
今年有一篇学术论文获得SAIL大奖。这篇题为《新一代中国人工智能》的论文发表于国际权威刊物《自然·机器智能》，主要作者是三位年轻的AI研究者——朱明杰、吴飞、卢策吾。这是第一篇关于中国AI发展的全景综述，全景式介绍了中国新一代人工智能发展规划的形成和发展，指出了在基础研究、人才培养和伦理道德等方面所面临的挑战，文章所描绘的人工智能生态发展蓝图正在转变为中国当前人工智能技术发展的生动实践。
一艘虚拟潜艇可能允许学生将学习场景缩小到微观水平，调查在岩石池表面下发生的自然过程，或者给学生设置条件，能够去探索一座核电站，甚至穿越到古罗马或外行星。
《美国2010国家教育技术计划》(简称NETP/2010)通过回顾和总结近三十年来企业部门应用信息技术的经验与教训，一针见血地提出“如果想要看到教育生产力的显著提升，就需要实现由技术支持的、重大的教育结构性变革，而不是渐进式的修修补补”。
同时，我国颁布的《国家中长期教育改革和发展规划纲要(2010-2020年)》给出了新的定位：“信息技术对教育发展具有革命性影响，必须高度重视。”规划为解决“如何使信息技术真正对教育发展产生革命性影响”这个问题找到明确的方向。
可见，技术与教育的深度融合体现在学校教育层面上，主要表现为AI对课堂教学的深层次变革。
智能辅助系统和教育机器人实现技术融合效应的潜能
学校教育环境正发生颠覆性变化：自20世纪90年代以来，以多媒体计算机与网络通信为标志的信息技术被广泛地应用于学校教育，走进我们的学校教室，现如今的教室早已不是传统意义上由黑板、粉笔、课桌、座椅以及书本和钢笔构成的，由先进技术配置的多功能/多媒体教室包括现代五大件和五小件。
然而，信息技术在教育领域的应用成效却并不明显，教学环境被现代技术不断丰富和更新的同时，也面临着由技术带来的、无法回避的教学困境：如教学内容冗杂堆砌，不利于学生消化；教学内容逐页呈现，割裂了内在的结构联系；被软件控制的程序一旦锁定，教学将严重受限；固定配置的座椅导致教学活动单一；设备更新迟缓，技术难以满足学生探究需求；电子白板自身的局限性导致师生、生生互动缺失。
以前，“信息技术”只被看作是一种工具、手段，大多数仍然停留在教育手段和方法的改进上，在教育效率和效益影响力上，信息技术似乎成了可有可无、锦上添花的东西。我们不禁发问：
从教师的视角而言，课堂教育技术应用是否真正有效支持教学？在技术的支持下教师能否腾出时间优化课程教学，并能感受到学生是否从中受益？
从学生的视角，技术是否真正有效地支持学习，学生是否能够深度地投入学习过程并获得好奇、敬畏、满足等积极的情感体验？技术与教育的融合效应值得期待。
事实上，现代教育技术的本质并不只是“使用什么教学”，更应是怎样教学，即运用现代先进信息技术来优化教育教学过程，达到提高教育教学效果、效率及效益的三重目标。
在具体体现上，除了对教育环境的提升和改善、对教学资源(含各种学习工具软件)丰富及优化之外，最为关键的就是在信息化教学环境下实现教育思想、教学观念、教学方式以及课堂教学结构的根本性变革。而AIED（智能辅助系统/教育机器人）技术所具有的灵活性、包容性、个性化和有效性等特征，将最大限度地与教育教学有效结合，具体在以下领域中发挥潜能：
人工智能将释放教师生产力，成为教师角色转变的催化剂
众所周知，早期的信息通信系统由于专家系统/知识模型领域的技术局限，在学科资源以及知识领域一直受限，只能应用于知识和逻辑结构清晰的自然学科，如低阶段的数学、物理等。
随着AI（人工智能）核心技术如知识表征方法以及自然语言系统的发展应用，学科资源领域得到了拓展，历史、语言类等社会学科的专家系统开发，知识建构获得长足的进步，将来的智能教学系统将逐步包含电子工程、管理类等应用学科领域，智能教学系统也将更广泛地应用于教育培训以及职业教育中。
在教师角色转换上，人工智能技术的增加应成为教师角色转变的催化剂，传统的备课、课堂讲授、答疑辅导和作业批改等教学环节将不必再是教师的专属：如辅导答疑任务可以由虚拟代理来替代，作业批改可以由学习伙伴或系统来支持。智能辅助系统/教育机器人通过承担教师的某些任务，帮教师从日常繁琐耗时的重复性工作中解放出来，把更多的精力投入到创新性和启发性的教学活动中。
在技术支持和认知上，随着人工智能技术的不断成熟，教师不仅能获得技术人员的及时支持，还能深度了解人工智能系统的功能，对全新的人工智能产品做出合理的评价和判断；同时，教师的能力结构也将发生解构与重构， AI技术应用将激发教师的研究能力，提出与数据相关的最有价值的问题，并引导学生使用数据分析的方法。
在管理协作上，将培养新的团队协作和管理能力，除了常见的教学助理，每位教师会配备人工智能助手，协作承担起教学环节当中可重复性的、程式性的、靠记忆、靠反复练习的教学模块，教师将集中在情感的交互、个性化的引导、创造性思维的开发上面，不断为教师效率赋能。
人工智能技术将激发学生学习潜能
数字环境下成长起来的新一代学习者对学习环境提出了更高诉求，固定的、单一有限的学习空间和场景将被打破，智能辅助系统/教育机器人将使移动学习、实景学习成为可能。
人工智能可以使用简单的方法来增强虚拟世界，使它能够与学生交互并以更自然的方式响应学生的学习行为要求。许多研究也表明，沉浸在智能虚拟现实中可以提高教育水平，使学生能够构建对所探索的世界的个人理解。
智能辅助系统/教育机器人的应用在不断拓展知识领域和知识类型的过程中将不断创新学习方式，协作学习、游戏化式学习等新的学习模式将成为可能。
例如，人工智能可以充当智能虚拟代理，即在学习“任务”中扮演教师、专家，或者扮演被指导者或学习同伴的角色，访问专业的专家系统，使所有学习者获得最适合他们需要的学习资源；人工智能技术还可以通过个人参与者的信息，以学习者模型的形式，组成一个最适合完成特定协作任务的学习者共同体。还可以设计游戏式的学习场景，随着游戏中故事的展开，从游戏和玩家那里获取信息，并基于这些信息，使用人工智能算法来确定采用哪些适合的学习行动。
AIED（智能辅助系统/教育机器人）通过模仿学习者的认知和情感状态，能将学习活动与学习者的认知需求和情感状态相匹配，保证学习过程中学生的深度投入；不仅激发学习兴趣、保持好奇心并通过学习成就获得满足，同时促进消极情感（挫折、混淆、迷惑、放弃）向积极情感（好奇心、专注）的正向转化；智能辅助系统/教育机器人利用对话让学生参与苏格拉底式的开放式学习模式，促进反思和自我意识；采用元认知支架提供动态帮助，提高学习者的动机和参与程度；使用社会模拟模型，通过理解文化和社会规范，使学生能够与目标语言使用对象进行更充分的接触。
2018年秋季的某天，Eightfold.ai招聘服务公司首席执行官阿什图什·加格看到了一份令他很感兴趣的简历。
简历的主人是一位有前途的数据科学家。他专注于挖掘数据中的模式，帮助企业决策，比如如何精准定位广告。有意思的是，他的简历中没有出现过“数据科学”一词。
相反，他在简历中呈现的样子是：加利福尼亚大学洛杉矶分校物理学研究生毕业，正在巴克莱银行担任分析师。他在职场社交网站领英上的个人资料表明他从未做过数据科学家，但Eightfold公司的软件却将他标示为合适人选。他在数学和计算机技能等关键方面与加格指定的软件作为模板参考的4位数据科学家十分相似。
加格表示，如此选人的核心理念是关注“他们有什么技能”，而非职位头衔。“那样，你就能找到虽无相应实际经验但能力足以胜任的人。”
在劳动力市场紧张的情况下，此项技术显然会吸引急于充实岗位的雇主。比如，数据科学家就是Google、Facebook和亚马逊等公司争相抢夺的人才。
现在，Eightfold等招聘服务公司依靠复杂的算法为劳动力和岗位匹配。得益于此，许多雇主可能很快就有机会接触到数量众多的潜在候选人，甚至有望填补高难度角色空缺。
哥伦比亚大学经济学家伯·考吉尔对人工智能在招聘中的应用做过专门研究。他说，算法可以帮助部分求职者，但也存有潜在的不利因素：它可能导致工资降低。
“雇主似乎无需经过竭力争夺，就能得到不同寻常但表现同样良好的合适人选。”考吉尔说。
多年来，雇主和网上中介机构一直在人才招聘中使用算法，但具体方法往往简单粗暴：计算机会识别简历中的关键词，确定是否与职位描述的文本相吻合。
相比在领英等网站上手动搜索，这么做效率更高，但缺点也很明显：求职者可以通过在简历中刻意夹入计算机匹配时可能搜索的词汇，操纵筛选流程。相反，措辞不当的描述可能导致计算机忽视合格的候选人。
但现在，人工智能在深度学习技术方面的最新发展令计算机前所未有地强大。它不再是简单地匹配单词，而是能够识别出求职者在简历中没有明确列举的技能和才华。
加格举例说明，比如，Eightfold通过数据分析发现加州山景城Intuit金融软件公司的软件工程师中有约90%掌握编程语言Java。那意味着计算机推断该公司某位软件工程师具备Java技能的结论是可靠的，即便他并未将此写在简历中。
通过对整份简历执行类似操作，软件便可以构建求职者的详细档案。它能够从诸如教育和爱好等项目中提取到寻常不被注意的信息，比如非知名大学的某些特定专业（如马萨诸塞大学安姆斯特分校的自然语言处理专业）盛产能力不凡的员工，象棋高手更有可能擅长编程，篮球运动员则精于销售。
为了改善计算结果，Eightfold公司巧妙地与客户达成协议，要求客户使用其人力资源软件。该软件以匿名形式导入员工数据，其中包括不同背景的员工在不同工作中的表现以及他们的收入信息。由此，Eightfold能够更好地预测员工表现。
如果雇主有意将搜索范围扩大到传统背景候选人之外的人选，这样的软件工具就格外有效。招聘人员可以指定某些标准（比如行业和地点，甚至候选人接受录用的可能性），获取相对特别的简历。
数字广告与营销服务企业AdRoll集团招聘人员克劳迪娅·维兰纽瓦表示，她最近在为财务规划岗位寻找候选人时使用了Eightfold。那位最佳候选人曾就职于技术行业通常选择回避的领域（如制造业），但是Eightfold的算法认为她是赢家。
“技术行业有时很难摆脱技术经验的束缚，因此一直在雇佣同样的人。”维兰纽瓦表示，“在我看来，本次体验是我们可以如何稍稍打破旧模式的有趣例证。”
在本文开篇提及的巴克莱分析师的案例中，Eightfold的软件之所以把他看作一位冉冉升起的数据科学家，部分原因在于：尽管他没有明示的数据科学背景，但是他拥有统计学硕士学位，并掌握了C++和Matlab等计算机编程语言。
“我们分析了数千万乃至数亿份人员档案，以了解人们职业生涯的变化过程。”加格说，“你能够轻而易举地预测他们下一步可能做什么。”不过，他承认这种方法在数据有限的新领域或冷门领域效果不佳。
调查显示目前仅有少数公司采用了深度学习类工具，其他公司甚至鲜有表达探索意愿的。宾夕法尼亚大学近期举办了一次有近20家雇主参加的会议。会后总结报告列举了各种各样的问题，其中包括评估员工表现的难度，以及计算机作出的招聘、解雇和晋升决定可能令员工产生不公平感等。
也许，最大的担忧实际上是基本的方法论问题，怀疑者担心的是，计算机实际上只研究了那些已经在某公司工作或曾前来求职的人，而非一般人群。对此，考吉尔相信有解决的办法，尖端技术公司正在利用人工智能填补职位空缺。
加格认为，Eightfold等服务商的优势在于能接触到许多公司的内部数据。
“相比单一公司自己建立的模型，我们能够建立更深层次的模型。”他说。
南京师范大学教授郦波曾说：
“真正的教育是，再富也要苦孩子。
你见哪一个人才不经历人生坎坷？孩子需要有正常的磨砺，而不是只给他各种卓越的条件。”
好的教育，从来不是让孩子泡在蜜罐里，而是尽早让他们认清生活的真相。
除此之外，扫地、 洗灶台、摆食材，父亲也让他样样都学着干。
父亲常对彭翔宇说，我们只是普通家庭，无法去跟那些富裕家庭比，但我们的命运掌握在自己手里，要得到什么，得靠自己的双手努力去挣。
父亲的教诲，彭翔宇铭记在心。
他不仅帮父母分担了许多家务活，学习上也异常刻苦。
最终靠着踏实勤勉，与极度的努力，叩开了北大的校门。
亦舒曾在《流金岁月》里写：
有钱人家败家子，寒门苦读出人才。其实都不一定的，还是要看家教，看经历。
比起物质的富养，孩子的品质、人格更值得家长花心思去培养。
就像林舒杰与彭翔宇的父母，他们没有多少文化，却懂得“宠儿多不幸，娇儿难成才”这样朴素的道理。
他们从不掩饰自己出身普通家庭的现实，也不给孩子超出自己能力范围内的生活。
而是还原生活的本来面目，让孩子与他们肩并肩，努力创造更好的未来。
这样的家庭教育出来的孩子，才能更加独立与勤奋，成为自我命运的引领者和一个家庭的领跑者。
▽
《商业周刊》出版的25个最佳观念专题报道，开篇就写着这样一句话：
“好好待你的小孩，但不要给他们太多财产。”
孩子在成长过程中，如果习惯了不劳而获，那么他永远也学不会感恩，也没有拼搏的动力。
成功的家庭教育，不是给孩子锦衣玉食，而是教他们自食其力。
教会孩子独立与奋斗，用精神富养出一个心怀感恩，独立奋进的孩子，才是一个家庭最深的福报。
评论
发表评论