https://hnavi.co.jp/knowledge/interview/roboken1812/ I
shared it.
2019.02.21
An interview column that introduces a system development company that is attracting attention. This time, Roboken is working on the commercialization of "pointed" technologies and services based on AI (artificial intelligence) R & D. We asked Mr. Munehiro Terada, CEO of the company, who thinks that "AI will not grow unless it imitates the human brain," about the business created by cutting-edge technology development and the thoughts on AI development.
■ "Selection and concentration" of AI business
Recognize small biological images by deep learning
--You are working on a wide range of commercialization of AI (artificial intelligence).
Mr. Terada, Roboken For example, we created a conversation platform that makes it easy to create chatbots that utilize AI, and a tool that can automate office operations, the so-called RPA (Robotic Process Automation). I have expanded my reach in various ways, but I came here to "select and concentrate" on the AI business.
Specifically, we will promote the AI business with three core areas: "small biological image recognition" by deep learning, "neuro AI" that recognizes human emotions, and "medical data mining" that aims to improve the efficiency of medical office work. I made it.
Small biological image recognition is a technology that uses AI-based image recognition technology to discriminate small insects, molds, bacteria, etc. at the same level as experts in the field. For example, it can be widely applied to the food hygiene field such as prevention of insect contamination in food factories, and prevention of infectious diseases transmitted by insects such as mosquitoes.
Neuro AI is a study to understand the emotions and mental states of a person by reading and analyzing the vital signs of a person with a sensor. If this technology advances, when hiring human resources, the interviewer will be able to grasp the situation of the person who came to the interview, so I think that it is also attracting attention as a human resources technology.
Medical data mining is research and development related to medical support, such as predicting the timing of patient medication by accumulating and analyzing vast amounts of medical data for various patients.
――Chatbots and RPA are not included in these three core areas, aren't they? I feel like it's a waste.
Mr. Terada That 's right. Actually, there are many other things that have come to me besides the conversation platform and RPA (laughs). For example, we have also developed packaged software that utilizes the in-house developed "stock price forecast AI (Pendulum)".
In addition, we have our own technology such as "epiRobo" that analyzes and edits "episodic memory" in the same way that people memorize something. "EpiRobo" is a technology that uses AI for the function of the part of the human brain called the "hippocampus." The hippocampus, in cooperation with the cerebral cortex, has the function of processing "episodic memory" that includes information such as time, place, and emotions of human memory. "EpiRobo", which is an AI version of the "episodic memory" processing function of "Kaiba", can store and edit some kind of episodic memory on the memory of a computer, for example. It may be a technology close to the "Differential Neural Computer" developed by Deep Mind in 2016.
-- So that's it. In other words, you have narrowed down the "pointed technology" that you have been working on to three cores by selecting and concentrating.
Mr. Terada Small bioimage recognition utilizing existing deep learning technology is a technology that can discriminate with high accuracy of 90% or more. In the application example of this technology, the contamination of foreign substances by food manufacturers is sometimes talked about, but when some insect legs were mixed, now experts are talking about what kind of insect and which part, it was mixed. We are analyzing whether it is before or after heating. It seems that a skilled expert can analyze up to about 97% accurately, but now the problem is that there are few experts and there are not enough people who want to do it. Therefore, our initiative is to automate with AI-based biometric image recognition technology.
With this technology, you can monitor hygiene in real time and determine if mold or bacteria are growing. Of course, by going through the network, it is possible to analyze the state of factories far away, so it can be used even in remote areas and closed environments. From these features, not only food hygiene management, prevention of infectious diseases transmitted by insects such as mosquitoes, but also protection of cultural properties that want to avoid worm-eaten and mold, management of warehouse environment, human health management, protection of rare species of organisms It can be widely applied to such as.
Mr. Munehiro Terada, CEO of Roboken Co., Ltd.
■ "Neuro AI" that accurately recognizes emotions from facial expressions
――Actually, the range of applications is wide. What kind of technology is "Neuro AI", the other of the three cores?
Mr. Terada This is a technology that estimates the emotions and state of mind of a person by analyzing biological information such as subtle changes in the facial expression of the person when AI faces the person's face. This technology is expected to be put to practical use in the field of "HR Tech" in the field of human resources recruitment. At the time of the interview, the sensor reads the vital signs and facial expressions of the person who came to the interview, collects them as information, and analyzes them with AI. This will give you an idea of the emotional and mental state of the interviewee. For example, you could ask all the interviewees the same question, see how they change their emotions and mental status, and decide whether or not to hire them. Also, for example, in the near future, at the time of a face-to-face interview such as a live interview, you may be able to read the emotions of the applicant and find out if you really want to work there.
■ Applying sharp AI technology to improve the efficiency of medical office work
――The two cores so far are certainly "pointed", aren't they? Not only that, both have a fairly wide range of applications. What is the third medical data mining?
Mr. Terada Regarding this, we are now collaborating with a medical company to receive medical data and analyze it. The analysis accuracy and results required by the hospital, which is the end user, have been obtained. We have completed commercialization, and we are planning to expand to medical institutions nationwide through medical companies. To be correct, it is in the category of "data mining for medical office work", but since it has not been commercialized yet, I cannot tell you the details.
Beyond this data mining of medical office work, we are looking to develop a monitoring platform and a telemedicine platform for the elderly, which are equipped with functions such as wobbling and fall prevention. It acquires vital data of the elderly with a sensor, accumulates it, and analyzes it with AI. Based on the results obtained in this way, it is possible to predict when wobbling is likely to occur and to measure the optimal timing of medication. It can be expected to be applied both in hospitals and in the scene of medical treatment at home.
■ AI will not grow unless it imitates the human brain!
What is Roboken's path to AI development?
――You have been working on the commercialization of various ideas and ideas based on AI, haven't you?
Mr. Terada I think that AI will not grow unless it imitates the human brain. Therefore, my foundation is to promote R & D of AI that imitates the human brain and commercialize it. With that in mind, Roboken has always hired people who are strong in brain science.
However, the human brain has not yet been elucidated. Therefore, I wonder if there will be any efforts to elucidate the human brain with AI in the future. Originally, it may be something you would do at a university or research institution. When it comes to working on that as a business, there are some difficult parts. Still, since I'm doing AI, I strongly want to work on it with such a dream.
0 コメント:
コメントを投稿