Knowledge Revolution at Zero Cost: How Sovereign AI Reveals Hidden Patterns and Transfer Knowledge

With low-cost hardware and local AI, new insights can be generated from existing knowledge. Transfer knowledge can be generated too: concepts and paradigms "hidden" in the middle of a text can be applied to unrelated fields of knowledge. With sovereign AI, this comes at zero cost.

Introduction

Find out how to apply insights from baking recipes to software development, materials management or knowledge field X. Sovereign AI gets better every week. We are close to the point where ChatGPT will soon be obsolete.

A small taste of what is possible with local AI:

• Input text 1: news snippets on AI bias
• Input text 2: news snippets on AI ethics
• The AI delivers the essence: If the quality of the input data is unreliable, stricter monitoring rules must be applied.
• The AI transfers the essence to legal science (topic chosen at random, any other field would be possible) and finds an insight: To avoid legal uncertainty in highly complex matters, preventive relief rules for the weaker party to the proceedings must be anchored in law in order to guarantee the normative stability of the legal system.

The essence and the insight mentioned were not present in the input data. They were obtained through abstraction and synthesis of the terse statements in the input texts!

To unearth a treasure of knowledge with AI, several prerequisites must be met:

1. The AI used must be powerful enough
2. The use of the AI must be independent of the volume of data (reason: the recovery work is lengthy, and the volume of data is large to huge)
3. The AI must work fast enough to finish in a finite amount of time

These three prerequisites can recently be met with sovereign AI. "Recently" means: within the last few weeks.

Today's reality looks like this:

1. High-performance AI models are freely available
2. These models can be run sovereignly on your own hardware at fixed costs (= the operating costs of the AI system)
3. The hardware required to run them is available for little money

Sovereign AI is AI that runs on your own infrastructure. It is of particular interest to companies, because every company owns data worth protecting. We are not even talking about personal data here, but about trade secrets, confidential or sensitive data, or simply valuable data.

A sovereign AI does not send data to the USA or to China. At best, the NSA cannot listen in either, because a computer sitting on or under your desk is beyond the remote access of any intelligence service.

What makes sovereign AI so interesting?

As mentioned at the start, a treasure of knowledge can be unearthed with AI. What does that mean?

In almost every company there are thousands of documents containing company knowledge, patents, insider knowledge or analysis data. No human can absorb all of this knowledge at once, perhaps with the exception of a few long-serving employees who, by definition, will soon be gone (retirement). That is, if such employees exist at all, and if they can share their knowledge comprehensively (time, recipient of the message) and want to (strategic considerations, motivation?).

How about automatically having your documents

• have concepts extracted and explained in newly generated documents based on your document knowledge,
• have related information from different documents merged into one document,
• have abstract truths (universal statements) generated from your company knowledge,
• have knowledge from your documents transferred to other fields of knowledge (transfer knowledge)?

All of this is possible today. Forget the timescales of the bureaucratic state of Germany, or the timelines that are common in teams of five or more people. We are talking about days and a few weeks, not months. Anyone who takes months to decide for AI might as well retire right away.

Here is a concrete example of the astonishing possibilities that AI automation on low-cost hardware offers with a sovereign AI system.

Impressive Possibilities – Without Agents

Forget agents. Agents are only for people who do not have a professional solution at hand but want a solution anyway. This applies to agentic coding just as much as to other areas where agents are used: agents are a mediocre (dangerous) solution instead of no solution at all.

Better are dedicated AI systems that carry out pre-designed, complex tasks in a highly automated way.

We want to generate knowledge that is hidden in company documents, so to speak between the lines.

For the sake of clarity, baking recipes serve as input here. They could also be contracts, negotiation agreements, meeting minutes or technical specifications.

The first recipe is a recipe for Mutzenblätter. These are delicious deep-fried pastries.

The second recipe is a crêpe recipe.

The recipes were searched for on the internet and extracted as text from the respective web pages via copy & paste. The copied text is somewhat jumbled, because ingredient lists and recipe texts are often not structured linearly. But that does not bother our AI.

Our sovereign AI then analyzed these recipes and extracted the essence of each one. The result:

• Mutzenblätter recipe: Thermal processing of a viscous dough by deep-frying produces a firm, crispy structure.
• Crêpe recipe: Thermal processing of a liquid batter by spreading it on a hot surface produces a thin, flexible structure.