[Editor's Note] There are two axes that are driving the development of artificial intelligence. The one that is currently receiving the most spotlight in the market is definitely generative AI. The small ball that Chat GPT launched has become a snowball, and there is currently a fierce competition for leadership in the development of ultra-large language models centered around big tech. And the remaining axis is edge AI or on-device AI. Edge AI is relatively less popular than the previous generative AI, but its practical use in industry and technology is rapidly expanding.

Here, we explored the cognitive solution, computer vision, which is the hottest area in edge AI and the starting point of all AI-based services. We met with Kim Soo-hoon, CEO of DeltaX, who has expertise in computer vision and edge AI, and talked.
■ Introduction to Delta X

I am the CEO of Delta X. We are researching artificial intelligence, and among them, we are researching computer vision. Computer vision is a field that analyzes images or videos and develops related solutions. Among them, we are focusing on camera computer vision.

It is a technology that analyzes and understands images or videos coming from a camera. It is called cognitive technology, and in English it is called perception technology. These technologies are also widely used in autonomous driving.

These analysis technologies are also needed for various mobile devices equipped with camera sensors and other sensors, so they are necessary and already being utilized across various industries.

DeltaX is an AI startup with about 70 researchers. It has been about 3 years since its establishment, and it mainly develops camera recognition models. Our technologies are mainly used in the automotive sector, and we mainly work on projects with automobile OEMs.

In Korea, we are conducting technology development PoC with Hyundai Motor Company, Kia Motors, Mobis, and other Tier 1 companies in the automotive industry. In addition, there are projects that we are discussing with overseas companies and projects that are being discussed and researched in the early stages.

■ How are research personnel organized?

The proportion of foreign researchers is a bit high. Currently, there are researchers from 14 different nationalities, and most of them are software and computer majors who completed their master's and doctoral degrees in Korea. There are also cases where they joined after requesting interviews with companies while living abroad. Of course, the majority are Korean.

Since there are employees of various nationalities, the company atmosphere is very horizontal. As a company that focuses on research, artificial intelligence in particular is an area where discussion is extremely important from the planning stage before development, and open discussion is an important area, so there is a need for such discussion.

So, I prefer an atmosphere where we can comfortably discuss and negotiate in an open space. That's why the office atmosphere seems to be structured in an open manner.

■ Are you also developing algorithms and core technologies?

The characteristic of our company is that we study the algorithm itself. Rather than developing applications using algorithms created by others, I am much more interested in developing the algorithm itself and the original technology itself, and I have a strong tendency to do so. Our researchers also like that more.

Our company is quite unique in that it not only develops technology and solutions, but also publishes papers. Last year, two of my papers in the field of artificial intelligence were already approved as papers. The company operates in a slightly different way from companies that only sell general solutions.

■ Do you feel that the need for computer vision is gradually increasing in the market?

DeltaX has chosen computer vision so far, which is the hottest field in the artificial intelligence field, and looking at the current trend, autonomous driving, various entertainment that requires cameras, industry, and even smart factories are closely related to the technology we are developing, so we are receiving a lot of needs, requests, and proposals from the industry.

DeltaX is fortunate to have many PoC and research joint projects with large companies, so it seems to be well prepared as it is able to read industry trends well.

■ As a computer vision expert, what is your outlook on the market?

It is known that the market size is about 17 billion dollars (about 23 trillion won in Korean currency) in 2023. It is expected that the market will grow by more than 20% per year for the time being. Based on this prediction, we can roughly predict the vision market for 2030-40.

Now, new autonomous vehicles, now Level 3 autonomous vehicles, are being introduced to the market, but ultimately, the ultimate autonomous vehicles must be equipped with, and are equipped with, many more cameras.

When you look at this area, it's only natural that the market will grow. I think it will inevitably become the most important sensor beyond the automotive field, and will be used in home applications, safety, and smart factories.

As the market grows due to this trend, (in terms of the situation where vision development companies are increasing), the demand for such things naturally increases, and I think that among those who majored in artificial intelligence, those who understand the appeal of computer vision and the explosive growth potential of the industry are starting businesses and conducting research.

■ What about patent registration in artificial intelligence algorithms?

We are also diligently and aggressively applying for patents, and if you look at our IP strategy roadmap, there are many targets that we plan to apply for or need to apply for. We are taking a quantitative and qualitative approach to IP and working hard to produce high-quality IP.

However, IP is a bit difficult in terms of software. Rather than an intuitive product, when you open the software, it is made up of a very long code, so it is difficult to intuitively define what should be accepted as an idea or what idea should be patented.

Although it is difficult, we are developing our own algorithms and are continuously releasing IPs to ensure that both our approach and new problem-solving methods are well expressed.

■ DeltaX's corporate vision is

I think I'll focus on cars for the time being. We will also be working hard on smart factories and home applications, but overall, we will still focus on automobiles in terms of sales and manpower.

The automobile industry is a huge market that produces 90 million passenger vehicles every year, and automobiles are changing from being a simple means of transportation to a new culture and a new space where people spend a lot of time.

As the definition of a car changes from the past when it was mechanical or electronic to an era where software defines the purpose of a car and its raison d'être, cognitive and computer solutions are increasingly required to do more.

The direction that automakers want to take with smart cars is also a solution based on cognition.
We believe that only then can we create smart cars, so we will focus on research and development related to cars for the time being.

■ Computer vision and artificial intelligence companies have a common interest in the need to optimize and automate workloads. What direction do they take in terms of management and human and material resource efficiency?

We have also been struggling with the same issue for a long time. We are still struggling with the same issue.

Every time, we analyze the customer's needs and start new coding from there. Of course, from planning ideas to coding, creating solutions, preparing data, training, testing and verifying the final solution, fine-tuning, and confirming that there are no problems, this is what typical companies do.

If this continues, the amount of money that a startup can receive will be limited, and because of these problems, it will be difficult to actually make a profit. In the end, it will be an environment where you will inevitably be forced into financial pressure by a series of deficits.

We are solving this problem by modularizing technology. By organizing the team well and deploying key researchers who are developing core technologies by modularizing them within the team, we can imagine in advance which modules would be well integrated to satisfy the client's requirements for the solution needs of Company A and define the technology.

Our first strategy is to define the area and scope of core technologies, place key researchers there, develop source technologies in modular units, and then assemble and bundle these to deliver solutions that customers want.

Even so, we cannot avoid doing fine-tuning and such. I think that customization and fine-tuning are always essential for solutions to suit the customer.

Since we cannot develop the technologies needed for that level again and again, we are approaching it in a way that modularizes and combines them.

Even that can only be said to be ready once the company has accumulated some experience and the project has been going on for a while. Of course, there is no company that can start a business by doing that from the beginning, and we certainly couldn't do that.

I think that as this goes on and 2, 3, or 5 years pass, the system will become more robust, and from then on, I think things like efficiency and solution costs will gradually improve.

(Continued in Part 2)