https://www.facebook.com/mannyyhl/posts/1752224735189982

In this post, we’re still stuck in the old mindset of using AI in the same old ways, which is a typical pattern during the early stages of new technologies. Initially, those who excel at using these new technologies are often the experts in their respective fields. It’s akin to when smartphones first introduced advanced camera capabilities; the first people to take stunning photos with their phones were professional photographers. As a result, the cost of staying competitive in the photography field rose (similar to the article’s mention of AI-driven game voiceovers becoming the norm, potentially leading to increased costs due to bidding wars).

However, smartphones also opened up new possibilities. Anyone could become a content creator and a YouTuber, and many YouTubers started their journey with just a smartphone. The reason AI is now used for voiceovers is that it has become affordable for everyone. Today, anyone with a CPU can use AI to add voiceovers to their games, leading to cost reductions under the new rules.

For instance, AI can assist in drawing, so professional artists are integrating AI into their production processes to find references more quickly. For artists, using AI as a reference-finding tool has become a competitive race to have better and faster computers.

But for conceptual designers, they don’t need high-end computers to get started; they can input their concepts without the need for exceptional drawing skills. Therefore, AI reduces costs in this area.

I’ve been thinking about how to obtain cost reductions in other fields without falling into a competitive race. Lately, in my trading endeavors, I’ve started to grasp this concept. In the past, quantitative trading was about competing for the length of network cables and bidding for them. Later on, the cost became too high, so institutions switched to using protocols with standardized cable lengths.

When AI first emerged, large institutions used it to learn patterns for trading, and they had to continue the competitive race. (It’s the classic scenario where the first users of new technology are the experts in the existing industry.)

But for someone like me, who is outside of the industry, writing trading strategies with AI comes at a low cost. Even if I only make a 1% profit in a month, it’s still profitable. (Large institutions, due to their substantial investments, need higher profit margins to cover their costs.)

One recent insight I’ve gathered is this: when Master AlphaGo (the Go AI) plays against humans, it often starts by keeping the game open, maintaining flexibility, and avoiding early determinations. LuckyJ (jp Mahjong AI) frequently adjusts its flexibility during the card game, striving to keep the later branching possibilities open and broad.

Language models like ChatGPT follow a similar approach. ChatGPT starts with a high degree of generality and then offers customization options through fine-tuning, preserving flexibility first and adapting as needed.

Conversely, in the context of the rat race in capitalism, professionals who specialize in a single field find it increasingly difficult to switch to other professions the longer they stay specialized. Applying AI’s logic, it’s essential not to narrow our focus too early and to maintain flexibility when making decisions to keep future possibilities wide open.

Looking at AI from this perspective, many professionals in various fields have made the mistake of pigeonholing AI into fixed roles, restricting their own potential by narrowing their paths.