Engineer Round-Table Talk: What we do “to bring AI to every being in the world”
--
At AI inside, we work with the mission of “to bring AI to every being in the world and contribute to a richer future society.” We interviewed three of our engineer VPs who work toward this ambitious mission to have them share their honest thoughts on day-to-day work- from the future society they have in mind, to the development environment and the mindset that helps them realize their vision.
Mitani: VP of SRE Unit (left)
Joined AI inside in November 2020. After graduating from Tokyo Institute of Technology, he gained experience in a system development company, including assignment to a development base in Myanmar and system development using machine learning. Currently leads the infrastructure development and operations team.
Iwasaki: VP of Learning Center Unit (middle)
Worked as an embedding engineer after studying image recognition and pattern recognition in graduate school. Joined AI inside as a researcher in 2018 and then moved to application development, where he was responsible for developing AI-OCR models and the inference infrastructure of “DX Suite.” Currently leads the “Learning Center” development team.
Tanaka: VP of AI on Things Unit (right)
Worked as a system engineer and has a wide range of experience in application development, infrastructure, security, etc., at an independent system integrator company. First started working for AI inside in February 2020 as a contractor, then joined as a full-time employee in July 2022. Currently involved in the edge device “AI inside Cube” and the surrounding infrastructure technology.
The idea to “Deliver AI” is like the “GHOST IN THE SHELL” comic!
- First, we’d like each of you to tell us what your units are working on.
Tanaka: Our AI on Things Unit focuses mainly on “delivering” AI, based on the company’s mission of “to bring AI to every being in the world and contribute to a richer future society.” Our job is to figure out what we can do with technology to provide an environment where all our customers can use AI anywhere they want. We’d like to make it possible to deliver AI to people who do not have internet access, considering that the current digital platform is accessible only with the internet.
Iwasaki: Our unit is responsible for the “Creating” AI part, and we work on developing the “Learning Center,” which can develop and operate AI without the use of code. AI is not just something we create, so it is essential to “democratize AI”-we aim to create tools that can be easily used by non-experts, from data maintenance to operation, to ensure that AI fits into the workflow so that that AI can be delivered to every corner of the world.
Mitani: Our SRE Unit also works to “deliver” AI. Our current aim is to ensure the stable operation of “DX Suite” and “Learning Center” and to develop infrastructure that will satisfy our customers. Looking toward the future, we are working on constructing “Leapnet,” a virtual distributed AI network.
Tanaka: There must be a lot of work to build “Leapnet”.
Mitani: The image I personally have in mind is the cyberspace in the “GHOST IN THE SHELL” comic- you can blend yourself into information without exchanging it through the internet. I hope we will be able to deliver AI in that kind of sense, and make AI something that supports human thoughts.
- That idea sounds familiar… is it similar to digital twins? That is something many of our members are interested in.
Tanaka: Based on my understanding, a digital twin is there to foresee the future; such as using digital reproduction to replicate accidents and help create a safe environment without causing real accidents.
Iwasaki: I agree with Tanaka-san’s idea. It might sound too cool to say, “it makes real space computable.” Still, if we could digitize the world so that computers can recognize it, it would be easier for AI to be trained. If this is realized, more things could be done without human beings. The conversion of analog data into digital data, which AI inside is working on, should also contribute to this.
AI inside’s development environment that fosters free thinking and vigorous discussion
- Can you tell us about the development environment you are using and what made you decide to use it?
Mitani: We use “Kubernetes” across all units, with the Technology Layer matching the work for each unit. We’re also planning to use it for the infrastructure technology of the virtual data center too. On the other hand, I feel that particularity is something that can not be produced by the tools we use. There are various SaaS services available now which are designed to be easily used, so the degree of freedom in the selection and the speed at which the tools are utilized are more important than what kind of tools we use.
Iwasaki: Most of the time, I agree with what kind of tools my unit members want to use. It has been easy for me to make these decisions since our company is a unit-based organization with a clear transfer of authority to the VPs.
Mitani: When I select tools, I choose those that “make my job easier.” That is why I adopt many tools commonly used in development processes, and I prioritize introducing tools that could eliminate inefficient human operations.
Tanaka: As Iwasaki-san mentioned, we are a company with vertical integration and a unit system, and we adopt a non-hierarchical structure. The decision-making is fast, but there is still a sense of understanding that we can move forward with consensus. However, it is also important to think “will our past operations improve with this change?” together as a set.. I hope our unit members will find and use these patterns to suggest new tools or ideas to their VPs.
Mitani: Logical explanations are essential when suggesting something, although there are less tedious processes to apply.
- How do you communicate internally and externally?
Iwasaki: Since we work remotely, we use “Discord” to communicate between engineers. We have a space for each product, and the members will be there so that we can voice chat any time.
Mitani: Chatting on Discord is sometimes more comfortable than talking with each other in person!
Iwasaki: Sometimes we visit our customers, too. It is not just the sales team but also the engineers who contact our customers to hear their opinions on the products. Also, we often communicate with members of the AI research team and UX team.
AI inside specializes in “the ability to turn AI into products”
- What do you think AI inside specializes in?
Tanaka: I believe it is the ability to turn AI into products. “DX Suite” is a good example now that it is developed as a service that lets anyone use AI for their work.
Iwasaki: We also specialize in collecting high-quality data. This is important because a balance of quality is necessary for AI training. If you have it read only the same kind of data, it will only be able to recognize that. If it reads only outliers, it will not recognize the volume zones. Balance is the key here.
- What makes it possible to collect high-quality data?
Iwasaki: Probably because our “cyclical synergy” is firmly in place. The key is that the starting point of the cycle is the application. For example, there is a workflow required for a BPO company to convert application forms and other documents into data, and “DX Suite” has been developed as an application that can respond to that workflow based on an understanding of the processes before and after the data checking and other processes. When I joined the company in 2018, I remember there was not much difference in accuracy between “DX Suite” and other companies’ services. However, while many of the AI-OCR systems were only APIs, what AI inside created was an application that could be used for BPO operations, and as the number of users increased and actual business data was collected, the accuracy improved rapidly.
Mitani: It is essential to deeply understand the user’s business and incorporate it into the product, which is exactly what AI inside is working hard on. That is how we succeeded in turning AI into products, and it’s a pleasure as an AI inside engineer to see many users using our products.
Iwasaki: In fact, this is one of the reasons why I decided to join this company. Before I joined, I had a chance to see “DX Suite” and thought that it was amazing. In my university days, when I was researching AI, I used to think that a black terminal screen with numbers on it was the norm, but at AI inside, they were implementing AI in society by making it into a cool application.
Taking on unprecedented challenges seriously- that is what the AI inside engineers do
- Can you share some of the challenges you are currently working on?
Tanaka: The difficult part of our job is that there is no precedent since we are trying to do something completely new, such as creating “Leapnet.” What we want to do is an initiative that dares to move away from the Internet in a society that is so dependent on it. It is a challenge, but it’s also what makes it interesting.
Mitani: With the Internet becoming more common and 5G becoming more widespread, it may be rare for people to dare to challenge themselves outside the Internet.
Tanaka: However, it is still a fact that people do not have access to the Internet in some environments. If we don’t have this in mind, we will not achieve our mission “to bring AI to every being in the world”. Besides, if we need the Internet everywhere, no one will choose to go to space, won’t they?
Mitani: Exactly. We must make it work online or offline to develop an AI that can move autonomously on a space station or lunar rover.
Tanaka: It makes me realize that engineers should be people who continue to question the common sense that is implicitly intellectualized as an assumption in the world. I do not want to reinvent the wheel, but would like to create something new with my own hands.
Mitani: I couldn’t agree more.
Iwasaki: Same here!
- What kind of person do you think is suited to be an engineer at AI inside?
Iwasaki: It is important to imagine beyond your task, such as how it will link to the company’s vision and benefit society, rather than staying in your shell and working on what you are given as you are given. Since we have a lot of discretion here at AI inside, this is an environment where we can achieve anything we want as long as the direction the company is heading and the vector of what we want to do are aligned.
Mitani: There should still be more “things we want to do” than “things we have done” as a company, so those who have the determination to start up on their own can gain a lot of experience.
Tanaka: Also, I think it’s important to accept other people’s opinions and enjoy the decision-making with the team.
Iwasaki: Sometimes, we need to learn to compromise, too. Overthinking will never make the goal.
Mitani: We are challenging ourselves to do something with no right answer, so it’s natural for us to have all kinds of opinions. It is important to be able to discuss and get to a solution. I am also excited when someone is thinking seriously about something on a large scale, such as being the best in the world or the universe.
Tanaka: One more important thing- it should be someone thrilled to hear Toguchi-san’s ideas!
At AI inside, we are still looking for people serious about becoming the global №1. If you have any questions, please feel free to contact us!