Bitcoin-centric AI language model aims to drive BTC education and adoption.

“It's good to answer questions related to Bitcoin and economics – at least better than GPT-4,” Alexander Svetsky told Cointelegraph at a busy Bitcoin Amsterdam.

Entrepreneur, author and founder of Spirit of Satoshi, a fictional artificial intelligence (AI) major with a large language model (LLM), the small startup begins to unravel the arduous journey it took to create a Bitcoin-centric AI chatbot.

The model is the result of a time-consuming training process to generate responses based on popular Bitcoin resources, the Austrian School of Economics, and libertarian ideals. In its infancy, the spirit of Satoshi reflects ideas from a “well-produced bitcoin corpus,” including resources such as Saifedean Ammous' bestseller The Bitcoin Standard.

Svetsky explained that the main challenge in building the model was not only gathering relevant data sources from books, research papers, and podcasts, but also directing the model to generate responses through a complete training process. A common misconception about LLMs is that they're getting information like a search engine, he adds.

“You're stringing words together to represent the patterns in the model. So it's not even a source.”

That's one reason why AI chatbots are increasingly “cheating,” Svetsky explains, and why developing an LML requires a focus on training on response style. Satoshi's spirit is by no means perfect, at least not in its current iteration.

“Our model evolves as well. It's also a rumor, but it's going to say more than a bitcoiner would say.”

Svetski's team, after acquiring a large but targeted Bitcoin central data and information base, programmed tens of thousands of question-and-answer pairs to feed the model. However, a human element is still needed to help Satoshi's spirit generate responses from his name.

Related: Bitcoin Amsterdam: BTC shines in the depths of the crypto bear market

The continuous development of the model therefore depends on the wider Bitcoin community. Satoshi Spirit uses an incentive process that allows the public to verify, create, and validate the model's data.

A “proof of knowledge” method allows users to pay in satoshis to train the model using credentials from the Lightning Network, Nostr or email addresses.

The process uses a consensus model that automatically penalizes users if they are creating “junk data.” Svetsky described Satoshi's spirit as a vital “human” element in improving it:

He is producing incredible content. It's the last part of taking your content from 80% good to 95% good. And this has a big impact on the quality of the model.

According to Svetsky, the difference between the responses generated by Spirit of Satoshi and ChatGPT is clear. The latter is trained on the main ideas of Bitcoin (BTC) and concepts such as inflation:

“If you ask ChatGPT about inflation, it will tell you that it is a sign of a healthy economy. Well, no; Inflation is a sign of systemic problems, such as declining purchasing power.

Svetsky says this situation is part of the raison d'être behind Satoshi's spirit of retraining LL.M., which reflects the type of thinking behind the Bitcoin movement.

“If you ask about inflation, our model says, ‘No, inflation is bad for the economy because it erodes savings,' or ‘Savings affect people's time choices.'

According to the founder, the future of the platform is fairly open. Satoshi Spirit can be a learning tool or an online tutor embedded in educational platforms or online universities. It could also be the basis for the “ultimate bitcoin influencer” with its BTC central results:

I want to see it become a destination for the next 100 million or 500 million people who want to learn about Bitcoin – their first step to understanding.

Satoshi Spirit is built based on an open source model with English behavioral competence and a “Wikipedia-style bias”. The latter is presented by fitting the model responses to the Bitcoin and Austrian Principles of Economics datasets.

Magazine: ‘Stylish and ass-back': James Lopp's Bitcoin First Impressions