2023년 3월 25일 일요일 – Mountain View 생활 76주차

tl;dr

I attended the GDC expo on Friday, where major companies like Meta, Adobe, Unity, and Unreal Engine showcased their VR/AR products. I tried out various devices, but AR technology still seemed immature. AR gloves were impressive but lacked rotation support. A Korean company presented a chair game controller, but users may experience dizziness with rapid scene rotation.

Generative AI is predicted to dominate next year’s GDC, but hardware remains the key to successful products. Drawing parallels between life and machine learning, I discussed risk-taking and avoiding local minima. Although I haven’t taken significant risks, I desire to reach global minima without risk, acknowledging that luck is 90% of success.

---

GDC

On Friday, I went GDC (Game Developers Conference) expo. This event is held annually in San Francisco and has 20 years of history. Since it was the last day of the conference, the conference center wasn’t crowded as we expected.

There were popular companies like Meta, Adobe, Unity, Unreal Engine, and others at the expo. Since the metaverse was the hottest topic last year, most products or games were about VR/AR.

We visited experience booths and tried VR devices like oculus and some AR devices. My friends and I agreed that the device for AR is still immature to replace the current gaming experience.

For example, there were AR glasses that could replace the monitor. However, the real experience was like Google cardboard VR that attached a mobile phone to the cardboard. When I wore the glasses, I could see the screen in the middle of the black background. The only thing better than cardboard was that the glasses were lighter than a mobile phone.

AR glove was nice, but it doesn’t support rotation. The glove nicely and accurately detects finger movement. Also, it can distinguish the custom posture of the hands. So, developers can define the set of postures. When the users make the posture in the game, it can be transformed into some commands. Suppose that the player can cast a ninja spell by making a finger posture. It would look real and fun. However, to make the experience closer to real anime, it should support rotation too.

There were some Korean companies. One interesting company displayed the chair as a game controller. The CTO of the company seemed really passionate. My friend tried the chair wearing VR and said the experience was good. One limitation of the technology is that a user feels dizzy when the scene rotates quickly. So, they said they adjusted the game speed to slow to avoid dizziness.

The next year GDC will be full of generative AI. However, regardless of the topic and trends, only hardware technology is the key to making a real-life product. Think about why PDA failed, and Smart Phone was successful. They have the same concept, but insufficient hardware spec and tech-infra strictly limited the PDA. Therefore, hardware is the future.

Local minima

In ML, following a gradient descent can be stuck in the local minima. I also think that the company’s growth or people’s life can be in the same situation.

Avoiding a loss and pursuing a profit seems reasonable, and everyone follows this strategy. However, our life is uncertain, and sometimes we need to take 1 step back to go forward two steps. But we never know if the 1 step back will turn into a two-step forward or an extra three-step back. That is risk-taking.

In my life, I usually didn’t take huge risks and hardly failed. So, I haven’t lost anything but haven’t made a tremendous achievement. I know myself well. I’ll never be satisfied until all uncertainty in my life is gone. Maybe even after the uncertainty is gone, I’ll still carve for what I haven’t achieved.

However, I want to jump from local minima to global minima without taking risks. That’s greedy, but it might be possible. Luck is 90% of the recipe for success, with the other 10% being hard work and skill.