All industries and applications deserve to be redone with AI.
Author | Tang Xiaoyin
Produced by | CSDN（ID：CSDNnews）
At the beginning of 2024, SenseTime held a new product launch conference in Beijing, bringing us the third product of the "SenseRobot" family, the light wing lamp, which incorporates AI into the table lamp. On the premise of satisfying lighting and eye protection, it has functions such as AI correction of sitting posture reminder, AI attention detection, AI light moving with the book, etc., priced at 1699 yuan.
As far as the table lamp itself is concerned, the price of the light wing lamp seems to be much more expensive.
As far as intelligence is concerned, the light wing lamp supports voice reminders, but it does not support voice interaction yet. It's still far from the current boom in generative AI in the industry, so why should SenseTime start with a desk lamp?
Starting from the "little things", SenseTime wants to let artificial intelligence enter thousands of households
Artificial intelligence has been developing for decades, with several ups and downs, and with AlphaGo's victory over Lee Sedol in 2016 and the birth of ChatGPT at the end of 2022, the breakthrough from perceptual recognition to generative has been completed in technology. Nowadays, recognition AI has become popular, which is inseparable from Tang Xiaoou, a leader in computer vision in China and the founder of SenseTime, from academia to the establishment of SenseTime to achieve application implementation.
Ma Kun, the co-founder of SenseTime and the president of Yuanradish, has a deep friendship with Professor Tang Xiaoou for 25 years, and volunteered to follow Tang Xiaoou to do face recognition research. Then in 2014, when Tang Xiaoou used his own family property to finance and SenseTime invited Ma Kun to join, he joined the co-creation of SenseTime without saying a word.
Ma Kun is the co-founder of SenseTime, a visiting professor of Xidian University, and the president of Yuanradish Intelligent Technology Co., Ltd
Adhering to the belief of "AI, people-oriented, born for love", in 2022, SenseTime released the first AI chess-playing robot "Yuan Radish SenseRobot", officially announcing that SenseTime will bring AI technology into all walks of life into the family.
In the meantime, it reflects a real status quo that we see that artificial intelligence seems to have gone everywhere, but in fact, it still has not really entered the family on a large scale or even in a comprehensive way. This is the first point that Ma Kun admitted in an interview with CSDN and other media.
In the family, there are a lot of seemingly "small" big things, compared with the construction of a direct smart home, Yuan Radish is more focused on the children in the family. Ma Kun said, "'Committed to integrating AI into life and escorting the healthy growth of children' is the mission of SenseTime to Yuan Radish."
For parents, accompanying their children to learn Xi is a rather distressing thing, it takes time, and it is often easy to get started, resulting in damage to the parent-child relationship. Ma Kun said that many parents have raised the question of whether they can be a robot like their parents and accompany their children to learn and Xi? For the author, I often wonder whether the development of humanoid robots, embodied intelligence and other technologies can make Doraemon come into reality and grow up with school.
Why can't we realize that one robot can meet the needs of multiple scenarios, Ma Kun told CSDN, "This is everyone's beautiful vision, and general artificial intelligence is also developing in this direction." The current popularity of large models is also based on this, and it is hoped that one model can solve all problems. But the development of technology is regular, and it takes a process, and we have to continue to work hard. ”
What Yu Yuan Radish hopes is to be able to solve the pain points of parents and children in the family in a down-to-earth manner, and for thousands of households, it can be used immediately without being scared away by the high cost of robots of hundreds of thousands.
Ma Kun said that for families, the table lamp is a more frequent and rigid thing than playing chess, which is closely related to children's learning Xi and health, and a large number of users have reported the pain points of learning Xi this matter as parents accompanying their children.
Based on this, the Yuan Radish Light Wing Lamp came into being.
Redo the table lamp with AI
Ma Kun, who prefers to call himself a product manager, algorithm engineer, and heavy user of Yuan Radish, said at the press conference, "Any industry is worth redoing with AI, and since users have needs, we have undertaken the mission of redoing the table lamp with AI." ”
Today, the table lamp has undergone three evolutions:
- In the 1.0 era: to meet the lighting function, mainly tungsten lamps and fluorescent lamps, the biggest problem faced is strobe, long-term use of useful eye fatigue.
- In the 2.0 era: eye protection, mainly LED lights, with higher brightness and no flicker, but it cannot completely solve the problem of myopia. Because myopia cannot be solved by high-quality lighting alone, it is also related to children's eye Xi. Even if the light source is good, the child's sitting posture is not correct, which is still easy to cause myopia and spinal problems. However, if parents accompany their children to learn Xi and constantly supervise their children's sitting posture, it is easy to cause parent-child conflicts.
- Thus, the 3.0 era is coming: entering the era of artificial intelligence, the table lamp should enter the 3.0 era, and it should have the functions of lighting, eye protection and intelligence.
In terms of hardware, the light wing light is equipped with a two-wing floodlight, an AI spotlight, an intelligent perception camera, a metal skeleton, and an NPU AI chip.
In terms of exterior design, it is quite in line with children's love for novelty and cool style. It inherits the aerospace design concept of the "SenseRobot" family. The design of the wings of the main light source is inspired by the solar wings of the space station, and the shape of the spotlight imitates the "water droplet" cosmic probe in "The Three-Body Problem".
In terms of "AI sitting posture reminder", the sitting posture analysis technology based on the 3D bone joint points of the human body can accurately identify 10 sitting postures, with an accuracy rate of more than 95%. When the child is perceived to have a bad sitting posture such as left-leaning, right-leaning, and forward-leaning, it will use voice reminders to help the child correct his sitting posture. The "AI Attention Detection" can intelligently remind children to take a break at the right time and combine work and rest.
These data analyzed and collected by AI technology can be reported in real time on the mobile app. At the same time, the focus data can form a visual daily and weekly report to show the proportion of children's concentration, eye protection curve, and eye fatigue value, so that parents can comprehensively and conveniently observe their children's learning Xi time, learning Xi status and interests.
"AI companion for children, AI helper for parents. Ma Kun gave a precise definition of such an AI desk lamp.
In the era of artificial intelligence, privacy and security are the core of the same. Wing Light adopts four strategies and technologies to comprehensively protect user privacy and security:
- Physical: Comes standard with a lens cap that completely obscures the lens;
- App self-selection: default digital human mode, local analysis, no data storage and no upload, real image mode, local analysis, visual desensitization, encrypted storage and upload;
- Visual desensitization technology: Localization and fuzziness of home environment information, no need to worry about privacy leakage during transmission;
- Data encryption technology: Data encryption storage and transmission are more secure.
Transcript of the interview
CSDN: How does the light wing lamp take AI attention detection and AI eye fatigue analysis as the core capabilities from the perspective of algorithms?
Ma Kun: This is mainly achieved by quantifying the state first, and then modeling the process with a mathematical model. Specifically, first of all, in the implementation of attention detection and analysis, we use a single frame to detect the user's attention and assign a quantitative score to this state. Secondly, in terms of the degree of eye fatigue, it actually involves a continuous change process, and we use a weighted formula to synthesize according to the values in each state to form a curve of the degree of eye fatigue.
CSDN: Many children actually feel very focused when they do handicrafts, but from the actual results you demonstrate, it is that they are not focused, why do they have such results?
Ma Kun: Yes, our current concentration test mainly focuses on Xi scenarios to meet parents' concern for their children, especially when writing homework and learning Xi. Of course, there are many concentration scenarios, such as the concentration that can be defined for crafting, building blocks, and even playing with mobile phones.
CSDN: You mentioned that the use of light wing lamps does not require networking, and the iterative optimization of AI relies on a large amount of data training behind it, under this premise, how to ensure that the performance of AI is getting better and better?
Ma Kun: Whether it's sitting reminders or other functions, all the analysis work is done locally. For users, there is really no need to connect to the Internet when actually using it. An internet connection is only necessary to upgrade OTA. For users, it can be downloaded online when there is a model upgrade in our background to bring better performance. A mature product should be able to enjoy new features without feeling it at all.
CSDN: We are currently in the era of large models and AGI (general artificial intelligence), our original expectation is whether SenseTime Yuan Radish is going to let "general artificial intelligence into thousands of households", but from the current point of view, there is still a certain distance, why is the light wing lamp not directly combined with the application of SenseTime's RiRixin large model system this time, from the long-term planning, what is the consideration of AGI?
Ma Kun: General AI, to be more specific, can be understood as generative AI. Unlike the previous AI that perceives the world, it changes the world out of nothing, and its business functions and expectations are different. At present, whether it is playing chess or desk lamps, they are both more functional, and generative is not needed for the time being, because generative AI mainly has a greater influence on content-based products. But in the future, we are also thinking about combining the ability of large models to do the function of answering questions and solving questions with children's human-computer interaction. And right now, we're incubating consumer-grade generative AI-related products.