Is Crypto+AI Web3 the last hope?

The Meme market seems to be on the verge of a collapse, and anxiety is starting to spread throughout the market. What’s the reason? It’s the lack of new narratives; the funds are quite clever, they’ve all flocked to blue-chip NFTs.

Regarding the new anxiety, I saw several KOLs discussing AI on Twitter today. They believe that Crypto+AI might be the next narrative direction and a hope for the future. As someone engaged in related business, I want to casually talk about this topic with everyone today. Let me first share two of my own experiences:

1. Last year, ChatGPT attracted widespread attention, and soon after, a game called “Stanford Town” was launched on the market. In this virtual town Smallville, researchers deployed 25 AI agents powered by GPT-4, which lived, worked, and socialized in the town, exhibiting complex behavior patterns such as handling emergencies, waiting to use public facilities, and communicating with others. Although this research was not open source at the time, the source code has now been made available. My friends and I also deployed this game locally and integrated it with GPT. After playing for less than five minutes, I found that we had already paid more than ten yuan to OpenAI. After repeated checks and confirmations, we concluded that the cost of using AI is indeed not low.

2. This year, we transformed an AI Computing Power factory inside a hydropower station in Yunnan. This was once a Bitcoin Mining Farm, which was abandoned in 2018 for various reasons. This year, we have renovated and put it back into use, utilizing Meta’s Llama3. Our client is a big data company, for whom we provide Computing Power support. Although it is our own Computing Power, the usage cost is still not low. It’s manageable during the summer flood season, but when we reach the dry season after October and the power supply is insufficient, we will have to Reverse negotiate with the local State Grid for power purchase, which will increase costs at that time.

Through the above two cases, I want to tell everyone that the cost of using AI is indeed very high. Of course, some frens may refute this, saying that using ChatGPT for chatting, generating images from text, creating videos from images, composing songs, and other AI tools is not expensive or even free. Indeed, these services are not considered expensive, but they are limited to these simple tasks. The cost for more complex tasks remains a significant issue, which is why we saw a large number of shell projects claiming to be AI based on GPT in the market last year. After all, they can’t afford the expensive development costs, and there won’t be many users.

The combination of Crypto+AI certainly has the potential to create many innovative products, but the biggest problem currently is that the cost of AI utilization is too high, limiting the possibility of large-scale applications. In this regard, I believe that both VC and the project party should be aware of this.

For the AI+Crypto sector, I believe we may have to wait until the next cycle or even further down the line to truly achieve large-scale implementation. Before that, we need to address two key issues: controllable nuclear fusion (low-cost electricity) and dedicated ASIC chips for AI (high efficiency and low energy consumption). These will be the foundation for the large-scale application of AI+Crypto.

At the end of the article, I will drop an Easter egg; there may be huge opportunities hidden inside.

Controlled nuclear fusion may be a future unattainable for us in the crypto community, basically hopeless. However, in the field of Asic chips, we do have a chance to see breakthroughs. Currently, the strongest in the domestic AI-specific Asic chip research and development should be Bitmain. That’s right, the Bitmain that makes mining machines, and they have been working on it for many years. Isn’t it a bit unexpected and a bit delightful? If that day really comes, you can each buy a little and host it at our Computing Power factory, just like mining BTC before.