Why Blockchain and LLMs Are More Similar Than You Think

A few weeks ago, something became clear to me. I was thinking about large language models, about how they work and about its limitations. And suddenly, I remembered how blockchains work. And I realized And I realized that these two technologies are much more similar than most people think. This is not a fact written in stone. This is just a thought, my thought. But I think it explains something important about both their power and their limitations. Blockchains and AI, they both change how people think about technology. Blockchain had a lot of hype, it also had a lot of scams, bad projects, bad coins and noise.

But technology is not the same as application. Even if some applications fail, the core idea can still be very important. So here is my thought on this. Both blockchains and large language models are closed systems. A closed system means that it cannot see the real world by itself. Let's start with the blockchains first. A blockchain does not know what time it is now, for example. It doesn't know what is the price of Bitcoin. It doesn't know what happens in the real world. Everything inside the blockchain is self-contained.

And this is not a bug. This is how blockchains are designed to be. And now let's look at the large language models. After training, LLM is also a closed system. It doesn't know what day is it today. It doesn't know what just happened in the real world. Its knowledge is Its knowledge is frozen at the moment in time, which is called a cut-off date. Any real world information should be additionally communicated to large language model in the form of user prompt, system prompt, or in general, as other parts of context.

So here is the similarity between these two systems. They are both closed systems. They are both powerful systems. And these systems do not have access to reality. They need something extra to get this piece of knowledge. And this is where things are getting interesting. Because the real problem is not really intelligence. The real problem is how to connect these systems to reality. And even more important, how to trust that connection. Blockchain already faced this problem some years ago. And they gave it a name, the Oracle problem.

In AI, we're now facing the same problem. But this we'll discuss next time. This was the thought on AI for today. I hope you liked it. And I'll see you in the next one. Bye.