Background and OpenAI

Q: Can you introduce yourself and give a brief overview of your career journey up to this point?

A: I’m Taehoon Kim. I'm currently the head of AI at a game company called SHIFT UP. My journey began as a software engineer after completing my bachelor's degree at UNIST. However, I didn’t have an interest in AI until after I saw Alphago. I did little bit of machine learning at school, but I wasn’t confident in my knowledge, so I wanted to learn and practice. I started implementing research papers and open-sourced the code, gaining some popularity on Github. This led to introductions from large tech companies, including OpenAI. After completing my alternative military service by working as a software engineer, I joined OpenAI and participated in building GPT-3. I worked on extending the training of GPT-3 from a single GPU node to multiple. I also worked on network sparsity. Afterward, I co-founded SYMBIOTE AI, focusing on consumer product. Now, I work on generative AI products at my current company.

• In South Korea, military service is mandatory, and alternative military service offers a non-combat option for fulfilling this requirement, involving roles in public service or specialized fields.

• Network Sparsity is a technique to train large networks with fewer resources.

Q: What steps or key decisions led to the opportunity of working in the U.S., especially at a place like OpenAI?

A: Differentiating myself from other engineers was key. For me it was working on open-source projects, implementing research papers. I found that implementation of research papers was something that people needed and something that I was good at. However, the landscape is different now. Reproducing the resources (computational power, memory) of a research paper can be quite expensive. I believe it's crucial for engineers to engage in diverse side projects to find their unique path.

Q: Can you tell us about your time at OpenAI and what projects you were involved with?

A: My time at OpenAI was amazing. When I was there, the company was small, with only 80 people, so everyone knew each other. There were many different teams experimenting. There was a robot team, a game team, which worked on Dota AI, a reasoning team, and a language team. For me gaining knowledge from these teams was very educational. It was more of the research lab vibe. The most significant distinction between my experiences working in Korea and at OpenAI was the abundance of events fostering cross-team communication. In Korea, most of the time you would communicate only with your team members.

Startup Experience

Q: What were the challenges you faced when starting your own technology company?

A: The biggest challenge from the beginning to the end was hiring good people. I had the ambition to build a team composed only of engineers, but this approach had its pros and cons. I ended up doing a lot of things that engineers usually don't want to do, like marketing, community management, hiring, and business planning. It was hard for me to make the right choices, and I realized that I should have found another co-founder who was good at the things I wasn't.

Q: How did you decide on the product to develop? What was the thought process or criteria involved in making that choice?

A: I had my criteria for the market:

1. Small at the moment

2. Fast growing

3. Not very competitive

The market I chose was the Japanese anime subculture market. I thought I could build a good product that could be used by many and differentiate from existing players with technology. Our product was specifically targeting teen avatar game social experiences. At the time, competitors like Zepeto and IMVU existed, but they were not tech-oriented. I believed I could build a product that would differentiate itself by technology.

Q: Can you describe the core product your startup was focused on?

A: Our product, Virtu, was a mobile avatar social network. We built one of the fastest lightweight motion capture technologies running on mobile devices, which supports not only facial capture but also full-body motion capture. Although it's been 2-3 years since Virtu's release, no other app in the app store has yet matched the user experience we've built. Virtu provided a platform where users could communicate through their avatars, offering a way to express oneself beyond the limits of inherent physical beauty. Our edge was the immersive mirror-like experience that we provided. As you take a selfie you could see yourself as an avatar. We wanted to extend functionality by tracking emotions, but we figured that running all of these on a mobile device is too energy-consuming.

Q: Looking back, what experiences stand out the most from your time with the startup?

A: Before that, I never had a chance to build a product from scratch and deliver it to market. This was especially significant in the case of designing a consumer app with global audience. We had to communicate with users using translators, gathering user base from different countries and cultures. Only few people in Korea would’ve done the same. I'm proud of what we achieved with the team and the resources we had.

• Most tech companies in South Korea develop products for the domestic market.

Head of AI

Q: What product are you currently working on?

A: I'm currently working on a generative AI product. I believe the game industry is a great market for this technology. Because even though game companies are more tech focused, they lack experience in developing tools. My current company is fast growing and the team has a lot of domain knowledge. In the last 6 months I learned a lot about game development and specifically genre of games my company releases. SHIFT UP releases gacha games, which is the business model used by top mobile games including Genshin Impact. As the head of AI, I aim to develop a product that can be utilized by other gaming companies as well, leveraging generative AI. Our current games in service are mobile gacha games. However, the game set to be released at the end of this year is a AAA console game, which is a 3D action RPG. We will continue to develop games for other platforms as well. I'm very interested in helping to build awesome experiences like these.

• Gacha game model is a virtual lottery where players pay for chances to win random items.

Q: What areas of AI are you most excited about and why?

A: I am interested in a field where I can make significant contributions, something that cannot be easily achieved by others. For this reason, I am not focused on language models, as most advancements in this area will likely be driven by OpenAI. Moreover, the optimal user experience will be created by individuals with domain expertise, ensuring that it is extensively utilized by users in that specific domain. As the key competitive differentiator in creating an effective chatbot is domain expertise, which I lack, I believe my contributions in this area would be limited. I’m interested in content creation AI, helping users achieve better results in their creative endeavors. Currently, I'm developing a tool to assist game designers. However, I believe that various other sectors can also utilize AI to enhance user experiences. For example, many people around the world use Instagram and take numerous pictures. Therefore, helping them take better pictures, which enhances the user experience, would be interesting.

Career Advice

Q: Could you share any resources or habits that have helped you in your career/life?

A:

1. I learned a lot from the code. However, I've realized that not many people read the code written by others. I believe that without referencing great code, you won't be able to write similar code on your own. Just as you can gain intelligence from books by great authors, I believe most computer science students should read more good code to become better programmers.

2. This advice is mostly for Korean CS students, but it's important not to limit oneself to the current community. Always aim for a larger market, a broader community where you can find more intelligent people. Instead of focusing solely on excelling in your local community, set your sights on the global stage. This approach ensures that your standards will continue to elevate as you grow.

Q: For students and young engineers aspiring to start their own startups, what skill sets do you recommend they focus on?

A: 

1. Get good cofounders

2. Execute a lot

3. Don’t overthink

4. Just build, ship, and reiterate

Personal Insights

Q: How do you balance the demanding aspects of your career with personal development and leisure?

A: I actually have achieved significant personal growth over the past six months since joining this company. This progress was made possible by having time to invest in myself, a luxury I couldn't afford while running my own startup. Focus is crucial; it's challenging to achieve growth in multiple areas simultaneously. While some may have the mental energy to divide their attention across several domains, I believe most people can only concentrate on one thing at a time. Over the past three years, my focus was on my startup, and before that, it was AI. In the last six months, I have concentrated on areas of personal growth that were previously neglected, such as enhancing my communication skills and engaging in small talk with a variety of people. I aim to build relationships beyond the tech community. Although I can easily connect with IT professionals, my goal is to expand my abilities in interacting with creators I aim to assist, like game designers, illustrators, and artists. Initially, I found it challenging to communicate with people outside my usual circles, but I have persisted in meeting individuals from various fields. Engaging with over 100 random people online has significantly boosted my confidence in conversing with diverse groups.

Q: Are there any personal philosophies or mottos that you live by which have guided you through your life?

A: I believe there are many paths to success, but having ambition and a strong desire has helped me achieve things faster than others who lack such drive. Without significant ambition, it's challenging to stand out in terms of rapid growth. Cultivating such ambition is complex; it can be influenced by friends, your environment, and the content you consume. For some, watching YouTube videos might spark immense ambition. In my case, having multiple friends who excelled in areas where I lacked skill spurred my drive. This motivated me to acquire those skills, largely through their influence. I frequently sought their advice, applied it, and then iterated based on their feedback. I have implemented over 20 research papers and met more than 100 people in the past six months. I consistently engage in actions that I believe will foster my growth along specific axes. The process is undeniably tough, time-consuming, and draining in every sense. Yet, I persist without giving up until I am satisfied with the growth I've achieved. This aspect of my personality has significantly contributed to my current self.

If you want to learn more about Taehoon, you can visit his website.