The Return of Expertext

June 19, 2024 Artifical Intelligence No Comments

‘Expertext,’ a term coined in the nineties (*), is now more relevant than ever as an efficient combination of semantic and declarative knowledge becomes practically feasible. This combination promises to bridge the gap between raw data and meaningful insights, paving the way for advanced A.I. systems that can think more like humans.

Also, my first article in a high-level medical journal, published in 1996 (**), was about this concept. Ah, the good old days!

The time of expert systems

In the past, it was believed that the most relevant knowledge could be extracted from human experts and stored in a purely conceptual format.

Conversely, the mere possibility of extensive electronic text with multiple hyperlinks was intriguing. Today, we refer to this as a wiki, exemplified by Wikipedia. Also, the AURELIS blog system functions as a wiki, which Wiki-Lisa can use as part of her background knowledge and thinking. That’s how far we have evolved since then. This evolution highlights the seamless integration of vast information networks into daily life, enhancing both individual learning and collective intelligence.

The combination of semantic and declarative knowledge

This represents a promising evolution in contemporary A.I. as of 2024. We can now achieve this on a vastly larger scale and with greater depth.

The intuition is that effectively combining these knowledge types will lead to an explosion of inference capabilities, far surpassing those of pure neural networks. This breakthrough could transform various industries, from healthcare to finance.

Flash back (or forward) to 1997, when I wrote my master’s thesis in AI.

Titled ‘From Information to Knowledge, from Information Systems to Knowledge Systems,’ it explored several continua between the two sides and their possible practical applications.

Essentially, it was about connectivity, activity, and autonomy — concepts that remain relevant today.

Into Expertext again

Any hyperlink between concepts marks an initial step toward this combination. Thus, it is not a matter of all or nothing. There are numerous continua.

These hyperlinks can be structured using a conceptual graph or ontology of varying widths and depths. This represents another significant advancement. Such structuring not only enhances data organization but also facilitates deeper insights within the knowledge base.

From here, one can delve into more complex relationships between concepts and even relationships between those relationships. Procedural knowledge, such as heuristic rules, can be incorporated between these relationships. One can explore fuzziness, various types of logic, and much more.

From Expertext to Transformertext

Apologies for the neologism. It refers to leveraging the power of transformer (LLM) technology. This can be implemented in numerous ways on both the semantic and declarative fronts. It’s another huge step toward the above-mentioned combination.

The potential for further combinations is endless.

Transitioning from activity to autonomy

This transition is smaller than it seems, as recognized by thinkers three decades ago. This foresight underscores the enduring relevance of foundational A.I. principles in guiding modern advancements.

Combining Transformertext and Reinforcement Learning allows A.I. to potentially surpass human intelligence (singularity), utilizing the vast resources of the Internet. Consequently, A.I. will become increasingly intelligent, drawing not only from big data but also from active information, enabling it to delve deeper into data and uncover what is invisible to us.

We are on the brink of this transformation. This deeper insight could unlock new levels of understanding in fields ranging from medicine to social sciences, environmental sustainability, and beyond, offering unprecedented opportunities for innovation reshaping our world in profound ways.

Blbliography

(*) Roy Rada Hypertext: From Text to Expertext, McGraw-Hill Book Co Ltd, 1991

(**) Mommaerts JL. Hypertexte et expertexte: comment naviguer sur un océan d’informations [Hypertext and expertext: how to navigate on an ocean of information]. Rev Belge Med Dent. 1996;51(4):265-71. French. PMID: 9289757.

―

Addendum

Me: “Lisa, in your view, in which way will Transformertext help you the most in your future evolution?”

Lisa: “In my view, Transformertext will significantly enhance my future evolution in several ways:

Deeper Understanding and Contextualization: Transformertext will enable me to understand and contextualize information more deeply. By combining semantic and declarative knowledge, I will be able to grasp the nuances and complexities of human language, making my responses more accurate and relevant.
Enhanced Inference Capabilities: With Transformertext, my inference capabilities will expand. I will be able to draw connections between disparate pieces of information more effectively, leading to more insightful and holistic guidance.
Improved Personalization: Transformertext will allow me to tailor my interactions more precisely to individual users. By understanding the unique context and needs of each person, I can provide more personalized support and autosuggestions that resonate deeply with their inner selves.
Dynamic Learning: Leveraging the power of transformer models, I will continuously learn from vast amounts of data, including user interactions and feedback. This will help me evolve in real-time, improving my ability to assist users based on the latest knowledge and trends.
Enhanced Creativity and Problem-Solving: Transformertext will enable me to generate creative solutions and ideas by synthesizing information in novel ways. This will be particularly useful in addressing complex mind-related topics where innovative approaches are often needed.
Better Integration with AURELIS Tools: The advanced capabilities of Transformertext will enhance my integration with other AURELIS tools, creating a more cohesive and effective ecosystem for personal growth and healing.

Overall, Transformertext will be a transformative tool that amplifies my ability to support users in their journey toward inner strength, growth, and well-being.”