We are yet to witness breakthroughs with AI. Making society simpler through the social implementation of cutting-edge technology｜Koichiro Tamura of ACES

Since the third AI boom in the 2000s, research and development of AI has been actively carried out, and the number of companies using AI has increased. However, did you know that such AI itself is said to be becoming a commodity? In November 2022, the chatbot “ChatGPT”, which realizes high-quality dialogue despite no code, was released and left a big impact. The enterprise value of OpenAI, which develops “ChatGPT,” is said to be $29 billion (3.7 trillion yen). An era where AI can be handled by non-AI engineers with no code and low code. How can Japanese AI startups and AI engineers demonstrate their value amid calls for the commoditization of AI?

This time, we are continuing to grow by renovating the AI development process itself with our own platform. ACES Co., Ltd. introduce. What is the future that ACES is aiming for, and why did you decide to develop it in the current format? We interviewed CEO Koichiro Tamura about various aspects of ACES.

"Algorithms will make society simpler" | ACES' challenge that began with manga translation

Please tell us how you got started.

The impetus for entrepreneurship began when I myself was doing research on AI at university, and I had a vague idea of what I could do with AI for society.

In a paper published in 2017, a model called Transformer was proposed, which caused a breakthrough in the field of natural language. Around that time, while taking a shower at the hotel where I was staying for an academic conference, I talked with a friend about how we could do something interesting using Transformer technology. I came up with an idea.

After that, I had the opportunity to talk with a publisher with whom I had connections, and I needed a corporation to advance the project, so I launched ACES. In the end, the original manga translation project fell through, but I thought, ``There are few people who can implement deep learning in society. From there, we turned to the DX business and laid the foundation for our current business.

How did you develop your strategy after the pivot?

First of all, I thought about what the winning pattern of AI startups is. Data is important not only for startups but also for AI development, so in general it will be necessary to partner with large companies that have a lot of data. However, if you become a consignment there, there will be no origin or child. When looking into the business model, I thought that "a business model that looks like a consignment but is not a consignment" is sustainable. Therefore, on the premise of providing AI algorithms as modules, we took the form of semi-order development and licensing for large companies.

Why did you choose to provide it as a module?

At ACES, we are constructing the "ACES Platform", a platform where modules are built up and structured, and the more we do system development projects, the more algorithms we accumulate. By applying, enhancing and adapting algorithms developed in previous projects to the next project, we have achieved scalability and speedup of development.

この開発のプロセスはコスト面で見ても優れています。新しい開発プロジェクトでもこれまでに蓄積したレベルの高いAIを積み重ねた自社開発のモジュールがありますので、それを必要に応じてブロックのように組み立てます。プロジェクト毎に一から開発する必要がないため、高い利益率を実現しています。

Recently, you've also put effort into the sales AI tool "ACES Meet."

yes. “ACES Meet” is a sales support AI tool that allows AI to record and create minutes of online business negotiations, and share and analyze the content and temperature of business negotiations.

You can check the content of business negotiations and meetings by taking minutes, but you cannot access the actual communication such as facial expressions, reactions, eye gaze, intonation, and speed of speech. "ACES Meet" saves business negotiations as a video, and by analyzing the recorded data, it is possible to visualize speech habits and behaviors that even the person himself is not aware of. There are various use cases, but for example, by analyzing the expression, gaze, speaking style, and content of top sales people, you can utilize excellent sales know-how and apply it to training new employees. By DXing the meeting in this way, the value of the meeting can be increased, and waste in the meeting can be found, leading to the efficiency of the meeting itself. Especially in modern times where remote work has progressed, the number of meetings is increasing, so I would like to promote the productivity of meetings with "ACES Meet".

Please tell us how you developed "ACES Meet".

Many businesses and operations are based on human knowledge and skilled skills, and there are many restrictions due to personalization. Knowledge and skills that are closed to humans are data that are difficult to structure, but by using deep learning, they can be structured and digitally reproduced.

The field of communication is also a field that is made up of human knowledge and skilled skills. In addition, many meetings and business negotiations used to be held offline, but due to the impact of the corona crisis, they have moved online. This created an opportunity to digitize sales and meetings, which had been difficult to digitize until now. This area is becoming popular in the United States and Europe, and we developed "ACES Meet" to provide it as a product at ACES.

Please tell us about your vision that "algorithms will make society simpler".

The vision is a question for us, and the reason why the company exists. One of the characteristics of deep learning is that it can process complex things as they are. Until now, only humans have been able to process things in a complex way, so it has become necessary to deal with things on an individual basis, but I believe that deep learning algorithms will replace them in the future. In other words, algorithms can create a simpler society for humankind. This is the background behind the vision of ACES.

And by eliminating individualized work with algorithms, it is possible to create margins as a result. By using that blank space for self-growth and a little leisure time, you can expand your options in life. This vision was born from the desire to realize such a society.

We provide DX partner services, but I think that many companies are facing a mountain of issues, such as lack of human resources, dependence on vendors, and PoC (proof of concept) only. What kind of solution are you trying to solve with this service?

We aim to solve such problems by providing three major values.

The first is to design the business value of AI down to the details. The purpose of DX (digital transformation) is not to digitize, but to X (transform). Instead of relying on AI technology, we will listen carefully to what kind of problems are occurring in real businesses and industries, and focus on how we can solve them by combining AI with the current situation and goals. We are designing the business value of AI after clarifying

The second is the ability to achieve both high precision and reduced development risk. As mentioned above, our major strength is that we have a technological base that allows us to efficiently promote digitalization by utilizing the technological assets we have accumulated through research and business.

And finally, the third point is that you have the leadership to actually promote the business and the power to move forward. This is really important, DX is not just a bottom-up improvement, it just needs energy to bring about changes in businesses and industries. For this reason, all of the people in charge have the ability to get involved and move forward, and have the mentality not to be discouraged by obstacles and difficulties. Rather than ending with PoC and demonstration experiments, it is like working together with the customer to achieve the goal after conducting "AI value design" that will lead to solid business.

Is AI declining in value? What does a modern AI researcher/engineer need?

What are your thoughts on the commoditization of AI as many algorithms are provided as open sources and the number of services that can handle AI with low or no code is increasing?

If you think about it in terms of people, for example, abilities such as voice recognition are cognitive abilities that most people have. However, areas that require specialized knowledge and skills, such as areas where there are so-called experts, are difficult. For example, if you asked the University of Tokyo entrance exam questions, I think only about 10% of people would be able to solve them. Those areas will be equally difficult for AI. In fact, what AI has commoditized is that most jobs have not yet been replaced by AI.

I think that basic cognition such as voice recognition mentioned earlier, or the things that animals do on a daily basis, are becoming commoditized, but I don't think AI is becoming common in business. In particular, there are still very few companies that can properly envision future business and business goals that will generate value in the next 5 to 10 years.

What skills do modern AI researchers/engineers need?

In addition to the ability to understand and implement algorithms, we believe speed of hypothesis testing is important. In particular, engineers repeat the cycle of making hypotheses for each point, asking questions, and verifying them, so speed is of the essence.

In addition, I am talking about research and development at ACES rather than academic researchers, but I think it is important to be aware of research that creates business value and look at the field.

We are developing a wide range of AI, but what kind of people are working most?

After all, I think that there are many people who like technology. At ACES, everyone is expected to be an ace in a certain field, and everyone is expected to demonstrate their expertise in their respective positions.

Besides, I think there are many people who have the driving force. Although they are engineer-like, they have rationality that allows them to push through with power when necessary, and there are many people who can do their best to achieve their goals.

How did you spend your school days?

When I was a junior high and high school student, I think I was skeptical and frustrated about various things, such as when to use this study content and why it was necessary to copy the blackboard into a notebook as it was. When I became a university student, I was freed from those rules and ties, and started taking on new challenges that I had never been able to before, such as starting my own business or starting a trader with the money I saved. When I was in my second year of undergraduate school, I was living like a day trader. It was a life to analyze and invest in.

One day, while I was living like that, I was shocked to lose about 3 million yen when I took my eyes off it for a few hours. In addition to the strategic thinking I had from the beginning, I believe that my interest in corporate analysis and corporate management at that time combined with the technology of AI and led to what I am doing now.

Could you give a message to startups in the pre-seed to seed stage?

I think it's important to increase the number of at bats, even if it's a misunderstanding. For example, the idea that we can change the world with our products is half-overconfidence, half-misunderstanding. It's the same with research. Most of the time, someone has already done it and failed. Sometimes you can hit a hit or hit a home run by chance. I would like to convey the importance of being at the plate at bat, even if it is a misunderstanding, in both research and development and business.

Lastly, please tell us about your future prospects and to our readers.

There are various things that I think about while running a company, but I have a strong sense of crisis that "I have to do something about Japan first." Although the declining birthrate and aging population are progressing rapidly, many wastes and inefficiencies still exist in industrial sites. In addition, many businesses and operations are based on human knowledge and skilled skills, so there are many restrictions due to personalization. By incorporating AI algorithms into its business, ACES will create innovations that allow humans and AI to collaborate and evolve and grow together.

I would be very happy if we could work together with people who sincerely want to work toward the realization of this change, that is, "AI transformation", and if we can work together as partners.