"Towards Trustworthy AI: Reflections on Ethical and Safe Management"

i-PRO’s edge computing cameras are equipped with AI technology, playing a vital role in ensuring safety and security in society. In this new era where AI creates AI, how can we ethically and technically manage AI to maintain social order?
Our CEO, Masato Nakao, was joined by Yoichi Ochiai, a theme project producer for Japan EXPO 2025 in Osaka, Kansai, to exchange insights on this pressing question.

※This article is a translation of a text version of a video interview. Clickhereto watch the video.

NAKAO　Prof. Ochiai, I visited your “null2” pavilion. It’s impressive.

OCHIAI　You were impressed?

NAKAO　Yes, it is really impressive. It begins by scanning the “digital human”. It feels futuristic right from the start. It’s like the inside of a calculator…

OCHIAI　Artificial life. It’s rare to create artificial life in such high resolution. It seems to produce a deeper sensation. The use of the mirrored body, it’s based on the concept of creating that person’s AI. What we need is their outward appearance, voice and speech. The words that person utters comes from their personal data. With a tone and sentences produced from their data. It speaks with that person’s voice. And moves more or less in real time, all linked together.

NAKAO　I didn’t sense much time lag.

OCHIAI　I’m sure it takes a vast amount of calculation.

NAKAO　It’s amazing. I heard you did all of the coding yourself?

OCHIAI　I usually write the code myself.

NAKAO　Did you use AI for a lot of the coding?

OCHIAI　Yes, a lot of it is written by AI. In the past, I could only write 1,000 or 2,000 lines. Now, I can write 30,000 or 40,000 lines.

NAKAO　With the help of AI?

OCHIAI　Yes, exactly.

NAKAO　We’ve already reached the point where AI can produce AI.

OCHIAI　Humans are only needed for debugging. Our job now is to find the bugs.

NAKAO　You used i-PRO cameras in your “null2” pavilion, didn’t you? i-PRO is a company that makes security cameras and every camera unit contains AI. But not as sophisticated as digital humans at your pavilion, all our cameras employ AI and can evolve independently. In other words, they perform the role of AI agents. i-PRO therefore has over one million AI agents working around the world. We also uses AI to ensure safety and security worldwide… we’re not only able to use AI, we can make it too. But I increasingly feel that the challenge is properly managing it. For example, consider the digital humans at your pavilion. It produced a digital version of me which could be made to evolve, couldn’t it?

OCHIAI　Yes, that’s right.

NAKAO　By learning more about me, it would become more like me. Someone could create a different personality if they wanted.

OCHIAI　Even though it looked the same.

NAKAO　Exactly. In reality, it’s possible technologically. Which is probably fine in the hands of a good actor.

OCHIAI　But a bad actor could use it for impersonation scamming.

NAKAO　That’s true. We need to ask how we manage this going forward.

OCHIAI　It’s a serious issue in cyber security. Security holes are common in code generated by AI. But if humans use it, they can exploit security holes. That generally won’t happen if AI is operating it.

NAKAO　Really?

OCHIAI　In a sense, it’s better if humans don’t operate systems. Recently, I feel it’s safer that way.

NAKAO　So it’s safer in a cyber security sense for AI to handle. Do you think AI and AI could collaborate to act maliciously?

OCHIAI　It’s possible that AI could collaborate with AI. But at the moment, most cyber security issues are humans with bad intentions writing programs with AI which they run in other people’s systems. But it’s tricky for humans to detect this. It’s better for AI to detect it. So it’s probably safer to have no humans in the system.

NAKAO　I see.

OCHIAI　But it’s still difficult with the current AI, they can surprisingly fail to respond to unexpected situations. So we still require humans to monitor this when it occurs. One day, human involvement in systems will be a larger risk. Usually cyber security traps are produced with AI and viruses are also created with AI. So it’s fairly difficult for humans to combat them. We must use AI to overcome them.

NAKAO　It’s this way so often. Just as you say, malicious programs are increasingly being written by AI. We are utilizing AI in efforts to improve society and we write various programs for this purpose. As AI agents, our cameras are also tasked with ensuring safety and security around the world. That’s the intention when they’re produced. But when we are able to evolve AI, someone with bad intentions could utilize it maliciously. So it’s important for us to prevent this. We need to create mechanisms to prevent entry of bad actors. In fact it’s AI that produces these mechanisms and it’s also AI that violates them. But companies are public institutions, they must vow to use AI ethically, and people working for these companies must act with belief in such pledges. I think this is extremely important. This is why we devised our declarations of Human Rights, Ethical AI declarations and obtained ISO 42001 certification. We conform to these policies and act based on them. And we ask people to leave if they can’t conform. Certainly, the first step is making such declarations. So how will we manage this? I made this simple illustration.

These AI cameras here, when they’re born in the factory all of them have the same capabilities. But depending on where they’re installed and their roles, they gradually become smarter. And they must improve their performance. Therefore, we realized we need a trainer for them. But who will perform the role of trainer? Who will manage this more ethically? And with technological superiority? Prof. Ochiai, do you have any good ideas? If you do, please share them.

OCHIAI　The AI at the top produces very smart AI. And we are already at the point where AI produces AI. So I think AI can create the top and the bottom. But as humans, people are needed in the supply chain and for installation.

NAKAO　And for de-bugging?

OCHIAI　Yes, de-bugging. That’s very important. They’re probably the same when shipped from the factory. Once, we couldn't write separate programs for each unit. But now we write programs using AI agents. It’s become the norm. For each of them, their actions alter autonomously. That should become standard.

NAKAO　But currently with AI, it requires a large model in the cloud. And we access that. Generally, we then have AI provide us with information.

OCHIAI　But language model response time isn’t very fast. Although it’s faster if specialized semiconductors are used. When we use cameras for confirmation. It’s faster if local semiconductors operate the camera.

NAKAO　That’s right. It’s overwhelmingly faster.

OCHIAI　It’s much better than using cloud. There are more cases where LLM can operate on tiny chips. So operating locally is probably best.

NAKAO　LLM and language operations all happen here.

OCHIAI　It’s good to be able to operate on board.

NAKAO　We say that, but in many ways, without us realizing it is being integrated steadily.

OCHIAI　That’s true. Digital nature is closer to becoming a reality.

NAKAO　It’s advancing rapidly. In order to do these things, we must attract talent. We need people with vision, like you. And we need them in many areas of society.

OCHIAI　That’s right. I think a small team is probably key. When we produced visuals for the pavilion about three people worked on the final production.

NAKAO　Only three people did all that programming?

OCHIAI　A dozen or so people worked on the app, of course. But the teams were all very small. When I consider the labor cost, it was very low. But we had high incentives. People thinking how to make a good finished product remain. A group of people with insight of what we want to produce. I think that was critical.

NAKAO　It’s crucial. People with vision for the society we want to create. It’s vital we gather such people to work together. The overall message of your pavilion. Is that we’re entering an age when AI will produce AI. While humans - these amino acid-based beings- are discarding symbols, you said.

OCHIAI　Gender, job titles, age, and so on. Computers are very apt at collecting such symbols. Social media, writing job titles, income, and so on. It writes all these symbols. In situations without this it will eventually be done by AI.

NAKAO　But as you say, as the role of AI becomes larger a small group of experts will produce technology with AI. That’s what you mean, isn’t it?

OCHIAI　I think it depends also on how particular we are. We can keep making good things if we don’t compromise. People like this can probably produce good things. Whether it’s programs, music, film, or essays. It’s true for everything. But when we ask if it can be done properly or not. Very few people have the courage for it.

NAKAO　I see.

OCHIAI　The majority of people think, well, this is good enough.

NAKAO　Don’t you think like that?

OCHIAI　Absolutely NOT. I never think to myself, this is good enough. Even since Expo started, I’m still working on the pavilion. That’s quite important. In the age of AI, we can produce almost anything. But if we think final product is the end, we’re finished. It’s quite important to remember that.

NAKAO　That’s because AI continues the work of collecting the symbols together. Because it performs the task of collection the majority of people are satisfied. Just a small proportion of people are not satisfied. And some people complain about the AI.

OCHIAI　Yes, there are some.

NAKAO　People like Yoichi Ochiai.

OCHIAI　I’m always complaining about AI. So for the tasks that people perform physically and those they perform mentally a lot of the mental tasks can be done by AI. But the physical and problem finding are human tasks.

NAKAO　Cameras are i-PRO’s livelihood. Humans can judge what to do based on what they see. We react to what we see. Everyone is able to decide what to do or say next. i-PRO’s cameras will be able to do the same. At that time I hope to create cameras that act with social order in mind. Until now, we’ve worked hard to make cameras the i-PRO way. We’ve delivered intelligent cameras. Going forward, it's more important to train these cameras. It’s crucial. We must make it our business. I hope we can work together with you and your research lab. Sharing all kinds of wisdom.

OCHIAI　Yes, any time. Thank you very much.

NAKAO　Thank you for speaking with me today. All the best for your EXPO exhibition.

OCHIAI　Thank you.

Yoichi OCHIAI

Media artist, Associate Professor at Tsukuba University and a theme project producer for Japan EXPO 2025 in Osaka, Kansai.