**In an era where human wisdom can be shared, “creating” the underlying AI: Series The Next Innovators (3) Cougars Atsushi Ishii**
*In order to change the world, there are people who doubt common sense, open up a pathless path, and continue to challenge to create a future that no one has imagined. The third installment of the series “The Next Innovators,” which approaches the source of that energy, is Atsushi Ishii, a cougar. He asked about his grand plans to develop a virtual human agent that is close to humans and to realize an era in which the wisdom of humanity can be shared by all humankind.*
Nowadays, everything around us is connected to the Internet, and the era of artificial intelligence (AI) processing the innumerable data sent from it is approaching. When the number of data and AI that processes it continues to increase, what will be the interface that connects the almost infinite world of machines and the world of us humans? Atsushi Ishii believes that the fourth interface after PCs, smartphones and voice assistants will be humanoid AI.
Cougars , co-founded by Ishii and CEO, is developing a humanoid AI that the company calls a “virtual human agent (VHA).” This agent not only interacts with people in words, but also communicates with facial expressions, gestures, and movements, taking into consideration the situation, emotions, relationships, and even past memories of the other person. It is said to support.
In order to “create” this VHA, Cougars is bringing together technologies such as artificial intelligence (AI), IoT, and blockchain in addition to human science, and each of these technologies has achieved remarkable results. In the field of AI, we are focusing on research on game AI, voice recognition, and image recognition. Team won the 2nd place.
In the field of blockchain, the Ethereum-based concealment protocol “zkCREAM” originally developed by Cougars has been selected as the official program of the Ethereum Foundation. In 2010, Ishii was appointed as the representative of the Japan branch of the blockchain global community “Enterprise Ethereum Alliance (EEA)” in which Microsoft, Bank of Tokyo-Mitsubishi UFJ, JP Morgan, etc. participate, and blocked as one of the 13 core members in the world. He is involved in sharing knowledge about the chain.
Beyond these technologies, Ishii envisions an era in which humans can grow through communication with AI and share the wisdom of humankind with all humankind. We asked him about the curiosity that started this epic dream and the roadmap to its realization.
Inquisitive mind about “human structure”
â??â??Cougar is a company that handles a wide range of technologies such as AI, IoT, and blockchain, but I feel that it is difficult to explain in one word. What do you explain to outsiders that the company is doing?
The answer is simple, trying to understand the structure of humans. I have a personality like a lump of inquiry, but there are two of them that I don’t understand the most. One is the universe and one is the human structure. The former has a history of refusing to join SpaceX and founding Cougars in the past, so I left it to SpaceX and am now working on the latter question.
As part of this, Cougars is creating humanoid AI to elucidate what human communication is. There are several elements necessary to create this humanoid AI, such as recognition of the five senses by IoT, machine learning, game AI to make you feel familiar, and blockchain to ensure reliability. is.
â??â??In a sense, you’re interested in something very human, and that’s where various cougar studies are derived.
I agree. I believe that the overwhelming difference between humans and other living things lies in the ability to collaborate at a very high level. Communication is required to make it happen, but this includes not only text and images, but also human facial expressions and gestures. I think that all of these things can be combined to do something else that can’t be done.
â??â??Communication for humans to collaborate makes a difference between humans and other living things.
For example, a newborn baby doesn’t know anything and can’t talk, but first he looks for someone. Even if you don’t understand the language, you try to convey your intentions to the people around you by acting. I feel that this kind of person-to-person communication is the principle of action that is engraved at the root of DNA.
Sometimes the words of the school teacher left an impression on me, and sometimes I remember what my parents said. With that in mind, I think that what is taught and communicated through communication with people has the power to change people’s behavior in the positive direction.
To understand such communication, Cougars is developing a Virtual Human Agent (VHA). It may be close to expressing that you are both an assistant and a coach, but it is an image that there is an agent who always understands you and promotes growth after knowing your characteristics. .. If it’s a school, it’s a teacher who is in charge of each student, and if it’s a job, it’s just like the AI â??â??assistant “JARVIS” that appears in “Iron Man”.
Brain science x machine learning x game AI
â??â??When you try to deepen your understanding of human-to-human communication as you just mentioned, I feel that in general, you tend to move toward brain science and biology. Why was Mr. Ishii interested in AI?
I think brain science is very effective in elucidating the mechanism of the brain, and I am constantly researching the latest research content. However, when developing VHA, Cougars combines machine learning and game AI with biophysics, including brain science.
So the reason for game AI is that games are the only medium in which humans interact with characters that don’t actually exist. There is always a character in the game, and the situation changes depending on the player. I don’t think there are many other media like this.
Another reason is that we use game AI and machine learning properly. That’s because the purpose of machine learning is to make predictions based on statistics, while the purpose of game AI is to make the character feel there, or to make the world feel real. .. That is why the player feels that the character is there, communicates, and tries to reach his goal.
â??â??That will lead to the elucidation of communication for collaborative work that I mentioned earlier.
Machine learning is needed in addition to brain science and game AI because AI needs to understand the world. Game AI requires little understanding of the game world. Because you can tell where the enemy is from the coordinates.
In reality, on the other hand, AI must first understand the situation around it. It’s very difficult, but with machine learning such as image recognition, you will be able to understand things statistically in the form of a car as a car and a desk as a desk.
â??â??It means that experience is input as data as machine learning.
I agree. Eventually, for example, when you see a cat, you recognize it as a cat, evoke semantic memory such as what kind of animal the cat is, and episodic memory such as memories with the cat, and make decisions and act based on that. You will be able to do it. To that end, we are fusing game AI with machine learning and brain science.
Using these technologies, I would like to realize a world where anyone can easily create a VHA, and that VHA will be close to that person and support that person. It’s like choosing a character for a game, creating your own virtual human, and you’ll be able to consult immediately.
Underpinning this is a huge decentralized network that is decentralized by the blockchain and that anyone can enter. This network allows machine learning engineers to publish their own AI models, and character designers to publish 3D models. It’s a place with the freedom of Linux, which was born open source by volunteers from all over the world.
â??â??When the time comes when everyone can create VHA, what kind of situation do you expect to play?
When it comes to work, agents will be able to provide work training in a sense and have an assistant who will also tell you what to do next, so you will be able to grow in a way that suits you. I think. VHA will also be responsible for communicating with people in games, entertainment, smart cities, smart homes, and autonomous vehicles. It’s like a translator between the machine internet and people.
AI that keeps close to you
â??â??You translate by standing between the communication between machines and people. What phase is Cougars currently in to achieve this?
We are now aiming to build “trust”. As with the episodic memory I talked about earlier, communicating with facial expressions, gestures, and movements, and communicating one’s tastes and roles are extremely important in creating trust in AI from people. Is important to. In addition, we are also working on reproducing things that are unique to humans, such as “I’m curious about what that person said.”
â??â??Does it mean that you can make decisions that are rooted in the relationship between people?
It means communicating with the other person using all the elements of the communication ability that a person has, and through that, deeply understand the other person and adapt to the other person. We attach great importance to creating a mechanism to rotate this cycle.
â??â??The difference between the AI â??â??assistant developed by other companies and the VHA developed by Cougar is that.
Yes. The AI â??â??assistants out there today are focused on responding as comfortably as possible to what they are asked. For example, when you ask Alexa or Siri something, it’s basically a statistically safe answer. However, we are focusing on working together continuously toward some purpose. In that sense, it may be a bit like a game.
â??â??I think the technical hurdle is to be together continuously.
I agree. For example, the VHA we are developing does not have wake words such as “Hey Siri” and “Alexa”, so we can talk to them suddenly. We don’t use wake words every time we communicate with each other, so we adjust to that.
â??â??The lack of a wake word means that you have to understand the physical sense of distance and past communication. Then, the interaction with AI will change from “command” to “communication”.
Yes, that is the point. The existing AI assistants basically do not lead the action, but the VHA we are making now proposes and leads the action to be taken next.
For example, when there is a certain task, after analyzing to some extent the behavior that will be most effective, “It is better to refer to this information at this timing” “There is a tendency like this, so I will work on this It’s better to do it. ” It is an image of proposing the most effective action to be taken from the collective intelligence of the actions of a large number of people.
Eventually, you will try to understand what the other person thinks through communication. By doing so, you will be able to understand that it may be effective if you act in this way, considering the characteristics and abilities of the person.
Beyond Metaverse, NFT, Web3
â??â??When you reach that point, the current AI assistant will go up one or two steps. What is the concept of a decentralized network secured by the blockchain, which is another area?
I originally noticed blockchain when I was doing research on machine learning. As I proceeded with his research, I was always worried that AI could be tampered with by changing the learning data. I decided to use the blockchain as a means to stop it.
Also, about four years ago, we have announced the idea of â??â??making VHA an NFT so that everyone can utilize the one and only AI assistant. On the other hand, the words Metaverse, NFT, and Web3 are really used now … Of course, each concept is great, but as the word goes on and the number of services and projects around the world grows, it costs too much for users to understand which ones are strong, credible and valuable. I’m worried about that.
In that case, the user thinks, “Let’s use the one I’ve heard for the time being,” and eventually heads for a centralized world where the name is what it is. In short, it’s the opposite of the basic concept.
â??â??It should have been an attractive mechanism to be able to secure trust regardless of the name recognition, but that goes in the opposite direction.
Of course, there are also good points. For example, when shopping at a major convenience store, consumers do not think “Is this product okay?” And shop with peace of mind. That’s because we know that the myriad of products on the market are all credible.
In the same way, it’s okay if everything is really credible and distributed. But now the cost for users to understand what it is is too high.
â??â??In that case, what remains after overcoming the enthusiasm of Metaverse, NFT, and Web3 will function as the foundation of society. How do you think the technology that Cougars is developing now contributes to society in an era of overcoming that enthusiasm?
First of all, I think that the completely decentralized and reliable AI network that I am thinking of will be able to guarantee the reliability of data and AI. For example, the reason we can eat the food in front of us with peace of mind is because we know where and how it comes from. I want to do that at the data level.
The other is to allow VHA to choose how to utilize reliable information according to the person. It’s more like providing what you need to grow, rather than entertaining that person.
Sharing the wisdom of mankind with all mankind
â??â??What do you think is the society in which such technologies have permeated all over?
First of all, since we are in an era where we can keep records of human communication through digitalization, I think that human wisdom will remain after excluding privacy-related data. This is very important in that it reduces the risk of being in a very disadvantageous situation due to the place of birth, environment and circumstances. Communication is information sharing, education, and fun, but there are still plenty of situations in which we cannot grow without such communication.
â??â??It means eliminating the disparity between the presence and absence of communication.
I agree. In addition, if the population continues to increase, food and resources will be in short supply in the future. Then, on a slightly larger scale, the time will come when human beings who live on the Moon and Mars will become multi-planetary species.
Then, it will be necessary to share information within Mars. I hope to support communication there. Elon Musk says he will bring a million people to Mars, but knowledge will need to be shared. It may finally be possible to cooperate with SpaceX.
â??â??It means that you can share the basic part of human wisdom with 1 million people from the beginning.
that’s right. Moreover, you will be able to convey information according to the person’s knowledge level.
â??â?? Then, you will be able to take what you have shared on Earth as it is to Mars and the Moon and share it.
Yes. I definitely want to talk about that as Elon Musk.
https://wired.jp/article/the-next-innovators-3-atsushi-ishii/