Student Number: 17020120095
Embedded Niu Nose The future of human-computer interaction will be a multi-modal state, inextricably linked to artificial intelligence.
Embedded Cow Nose Intelligence Human-Computer Interaction Artificial Intelligence
Embedded Cow QuestionsWhen the boundaries between humans and machines are blurring, where are the boundaries of human-computer interaction? What will human-computer interaction develop into in the future?
Embedded cow text in the 2013 film "She", the hero after the end of the long-distance relationship, fell in love with a computer operation of the system female voice, this girl called "Samantha" not only has a slightly hoarse voice charming, and funny.
So they became friends and even developed a "monstrous" relationship.
Although the movie is fictional, it proves that the future of user-centered interaction will not just be about ease of use, but also about recognizing the user's expressed intentions and emotions.
The question is, when the boundary between human and machine is gradually blurred, where is the boundary of human-computer interaction? What will human-computer interaction develop into in the future?
In the seventh installment of AI Time's debate, Tsinghua University's Shi Yuanchun, Tian Feng from the Institute of Software of the Chinese Academy of Sciences, Chen Yiqiang from the Institute of Computing of the Chinese Academy of Sciences, and Cao Xiang, the CEO of Xiaoniu Technology and Creativity, came together to talk about issues related to human-computer interaction and intelligence.
In 1945, when the electronic computer was not yet "born", Vannevar Bush published an article entitled "As We May Think", which described the future of personal computers -- a computer called MEMEX.
In 1960, Vannevar Bush published an article entitled "As We May Think", which visualized the future personal computer - a machine known as the MEMEX - and explained the concepts of direct interaction, hyperlinks, and network storage.
In 1960, Joseph Licklider put forward the idea of "man-machine **** life", and under the leadership of George H.W. Bush, through the U.S. National Science and Technology Program strongly supported the concept of man-machine **** life under the concept of graphics and visualization, virtual object manipulation, the Internet, and other research projects under his leadership, personal computers, The iconic key technologies of the Internet were born one by one in the 1960s and 1970s.
Joseph Licklider's leadership in interactive computing not only led to the development of the Fractional Operating System, but also led directly to graphics technology.
Driven by pioneers such as Vannevar Bush and Joseph Licklider, and with the involvement of linguistics, psychology, and computer science***, computers went from having no user interfaces to having graphical user interfaces, and created new industries such as the personal computer, and the interconnected Internet, that benefited the whole of society.
Now cell phones don't need to utilize a mouse, they can utilize new sensing technologies, including AI technologies, which are enabling people to feel more of the world around them, which is part of HCI.
In the future, with the ****same support of new sensing and multimedia technologies, machines will be able to understand us and the environment around us through perception and data processing technologies, enabling more natural and intelligent human-computer interaction.
Cao Xiang introduced, his current work can be said to be a "realistic version of the magic brush Ma Liang", with a piece of ordinary paper and an ordinary paintbrush painting, with a cell phone acquisition, instantly transformed into a three-dimensional animation. Through the technology to create a lower barrier, so that ordinary people can express their creativity is the original intention of the research.
So far it is still the era of pervasive computing, and the future of human-computer interaction will be multimodal, which can be done with the keyboard, voice, but also with gestures, facial expressions, and lip movements. He firstly introduced sign language recognition based on the theory of multi-channel or multi-modal perception for two reasons, one is because the gesture language is too much and too general without clear target boundary, and the other is because he hopes that the technology can serve the daily communication of people with disabilities.
The second work is also related to multimodality, the ultimate goal of human-computer interaction is to make human-computer interaction the same as human interaction. Currently, through multimodality, including knowledge-based perception, the machine is informed of the current state of the person, and then the next step in behavior. In the future, wearable devices can be utilized to infer a person's physiological and psychological situation and then interact.
Chen Yiqiang also believes that the future model of human-computer interaction will be multimodal. Around the "multimodal" concept, he mentioned the current work. One is based on multi-channel or multi-modal perception theory of sign language recognition, facial recognition, gesture recognition and sign language recognition integration, to improve the accuracy of sign language recognition. The second is to make robots aware of the current state of humans through multimodal means.
Feng Tian, who focuses on human-robot interaction for education and healthcare, shared his research results on re-input technologies and related theories. Since input is inevitably inaccurate, it is hoped that intelligent methods will be used to improve and help.
Motion target selection is a very important task in human-computer interaction, as we know when we play games, it is more difficult to select a motion target compared to a stationary one, how to improve the efficiency of selection, and at the same time understand the user's ability to select a target. They first did a lot of user research, generated for different speeds and sizes of objects in the movement process of the distribution of the landing point, to establish a model to calculate the probability of the user to select the object. This model can be used to analyze not only normal people, but also Parkinson's patients and others as an aid to diagnosis.
It is worth mentioning that Tian Feng led the team to develop the pen e-learning system won the second prize of the National Science and Technology Progress Award, and with the Union Hospital *** with the National Health Commission issued by the healthcare artificial intelligence application landing 30 best cases of honor.
Shi Yuanchun introduced, in the use of cell phone soft keyboard, 26 letters crowded in the narrow input interface, coupled with fat fingers, pointing wrong experience is too much. This is a typical problem on a natural interactive interface like a touch screen: the fat finger problem.
Based on their research, they proposed a natural user intent understanding framework based on Bayesian inference, modeling user behavioral characteristics and inferring the user's true intent on fuzzy input signals. It doesn't matter if you can't order correctly, the algorithm can guess correctly. Using this technology, Shi Yuanchun's team has researched and implemented input methods on a range of interfaces such as cell phones, tablets, helmets, TVs, and so on, with a substantial increase in input accuracy and almost no need for visual targeting, which in turn also supports blind users to accurately realize soft keyboard input.
Future interfaces will also be extended to sense human manipulation behavior, Professor Shi is developing on the front camera of the phone will be able to: "After sensing the changes in the interface of the human hand, we will be able to use this to make a new 'input method'."
For example, if you hold your hand on any edge or position of the phone, you can enter information, access the interface, and even interact with the table, which can be turned into an operation of the phone."
In response to the question of whether there is a computational model for the construction of interactive interfaces, Tian Feng pointed out that there is traditionally a computational model, and for natural human interaction, there is no corresponding theoretical computational model, but efforts should be made in this direction.
Prof. Shi Yuanchun agreed with the above viewpoints, and pointed out that quantitative evaluation methods exist, but are very insufficient. However, with the help of appropriate sensing technologies, the principles and techniques of quantitative assessment are expanding, as evidenced by a range of applications such as infrared reflectance monitoring of blood flow, blood pressure parameters, and mood changes.
The computing terminals in the future are diverse, and the adapted scenarios and tasks are also different, so it is very difficult to build a completely unified large and comprehensive model, but on specific tasks, the scientific principles behind the technology must have computational models, which researchers should all try to explore.
In addition, quantitative assessment methods theoretically exist, but now it is difficult to say that it is a good method, because the interface is expanding, and the corresponding realization technology, principles and evaluation techniques are also doing expansion and change.
Cao Xiang also pointed out that because of the variety of tasks in human-computer interaction, it is difficult to define efficiency, which is measured more by subjective feelings. In the general direction, there must be a need for quantitative data, artificial intelligence needs data, and human-computer interaction can not be separated from artificial intelligence.
For the mechanism and limitations of modeling, Cao Xiang pointed out that it is relatively easy to model work with a clear task because the goal is very clear, but experiential, recreational, and communicative work is more difficult to model in a computational way because it is mixed with a large number of non-simple human-computer interaction, such as human-to-human interactions.
The prediction of emotions is more effective with big data, and in itself humans don't have a breakdown of what is emotional down to small units. On the flip side, big data analytics or AI analytics, which can make predictions without using explicit segmentation models, can solve precisely non-standard tasks. However, if the problem is solved using big data, the establishment may be a generalized model, involving individuals will also some differences.
Chen Yiqiang believes that human-computer interaction to do a good job, it must be personalized, and will definitely use intelligent methods. In terms of interaction, the keyboard is initially a deterministic interaction, and the mouse is a perceptual level. Moving towards intelligence, the part of speech recognition and gesture recognition has added knowledge-based learning in addition to perception. To the third part, emotional intelligence, knowledge, or cognition, needs to be added before execution. This also corresponds to the three parts of the human brain, namely the central nervous system, cerebellum and cerebrum.
Intelligence, like human-computer interaction, is also hierarchical by person as well, and human-computer interaction, which we also hierarchize from traditional to intelligent, can be understood as the elimination of uncertainty. The higher you go, the greater the uncertainty, especially for the intention to understand, but how do we go to eliminate it, is a discussion on the application of human-computer interaction.
Artificial Intelligence and Human-Computer Interaction, both have the word "human", the relationship between the two, Shi Yuanchun first pointed out that this is the Chinese saying, the English does not have such a word, but the **** of the two is that they are both very early and very clear to talk about the relationship between man and machine.
Prof. Shi Yuanchun believes that human-computer interaction should allow the machine to better adapt to people, to adapt to human nature, to adapt to the human ability to manipulate, the ability to perceive and cognitive ability. From the "human" research content, human-computer interaction and artificial intelligence are different, but the starting point is the same, that is, "human-machine **** life".
At present, the research of artificial intelligence is more reflected in the human identification, language expression and other data-intensive tasks on the processing method, human-computer interaction research is more focused on the human active interaction behavior and perceptual ability to modeling, sensing and establishing adaptive interface technology, human-computer relationship is bound to the direction of the development of the **** life, the research content and methods will affect each other and adapt to the intersection of research content will be more and more. research content will be more and more.
"Doing AI ends up touching HCI, and doing HCI ends up touching AI."
Referring to an opinionated paper in Science China, Tian Feng pointed out that the trend of human-computer interaction and AI in the future will move from alternating sinks to synergistic ****ing progress. A core research point in the national artificial intelligence development plan is human-machine collaboration, and human-machine collaboration is also the future direction of human-computer interaction. From the point of view of artificial intelligence, autonomous driving and other human-machine collaboration, in fact, is the same way.
Cao Xiang pointed out that the research values and starting point of artificial intelligence and human-computer interaction are slightly different. Artificial intelligence fundamentally, the ultimate goal is to enable machines to do what everyone can do, human-computer interaction refers to the cooperation between people and machines, the two are not contradictory, but depends on the context.
Aiming at the contribution of HCI research to AI, he pointed out that firstly, the contribution of AI to HCI research must be recognized. In terms of the general trend, a large amount of manually labeled data in machine learning is the process of HCI. Further, one of the challenges of AI is interpretable AI, and ultimately the worry is whether it can be trusted, and the reason for the interpretation is the desire to be able to use it with confidence.
In a sense, perhaps the solution to the problem of trustworthy AI lies in creating a way for people and AI systems to slowly pass judgment through measurement in an interactive process, and perhaps this is precisely the way that human-computer interaction can help solve the problem of the so-called AI explainability.
For intelligent human-computer interaction, Shi Yuanchun pointed out that the shape of the computer will change in the future, and may even no longer exist, but computer technology will continue to serve us as part of the life of the human-machine ****, the interaction interface, the interaction task will be a big change, but it will be more natural and more intelligent.
She focused on intelligent human-computer interaction into three categories, one is gesture, then voice, and wearable devices, including bracelets and helmets. On these three categories see a lot of new technology and new products, but none of them have yet become mainstream, that is, all have certain problems.
For example, voice interaction, not only is the recognition rate has not reached 100 percent, at the same time, the bandwidth of voice expression and the type of data expressed is still incomplete, and the data related to space is inefficient and not accurate. In addition, there are disturbances, privacy, etc., there are great qualifications, wear even more so.
Chen Yiqiang example wearable devices can be attached to clothes and shoes, human-computer interaction will eventually realize the human-machine **** born. And, with advances in materials and technology, can fully understand the intent of natural human behavior, and even help solve the aging population, Alzheimer's disease and so on.
Cao Xiang, based on his current research, pointed out that it is important to bring everyone's creativity into full play through technology, and that creativity will become an indispensable part of survival and work in the future.
It is very interesting to see how more experiences may be available on the output in the future, such as tapping into more sensory experiences, not just in the visual and auditory realm, but even creating a fantasy world.
Tian Feng said, he is more concerned about how to promote industrial development through the research of human-computer interaction, related to the aging of the population has been in-depth cooperation with the Union Hospital, through the interpretation of the movements of the elderly, to provide quantitative auxiliary diagnosis.
For the cultivation of HCI talents, Shi Yuanchun pointed out that there is a demand in industry, but academia is still confused. There are doctoral students graduated in the industry can not find a very match with the professional position, due to the progress of industry will prompt the academic community to establish a set of scientific methods for the cultivation of talent.
Prof. Shi Yuanchun mentioned, "The talents we cultivate should be able to find out the interaction problem and solve it through scientific methods."
Cao Xiang pointed out that interaction designers, user researchers and other counterparts in the training of the profession, it is not difficult to find a job; difficult to find a job is to take human-computer interaction as a field of study to study the students, because the existing a carrot in the vocational system is not very suitable for interdisciplinary talent, but entrepreneurship is particularly in need of such people.