What Is Knowledge Engineering?
Let me explain knowledge engineering to you directly: it's a field in artificial intelligence where we create rules to apply to data, mimicking how a human expert thinks. We examine the structure of tasks or decisions to figure out how conclusions are reached.
From there, I can build a library of problem-solving methods and related knowledge, which the system uses to diagnose problems. The software that results can handle diagnosis, troubleshooting, and issue resolution on its own or alongside a human.
Key Takeaways
Knowledge engineering is part of AI that builds rules applied to data to copy the thought process of an expert in a particular area. Originally, it focused on transferring human expertise directly into a program to process the same data and reach the same conclusions.
We found limitations in this transfer process because it didn't capture human intuition or gut feelings, which involve analogous reasoning and nonlinear thinking that aren't always logical. Now, knowledge engineering uses modeling to create systems that achieve the same results as experts without following the exact same paths or using the same information.
The aim is to integrate this into software for expert-like decisions, such as those made by financial advisors. It's already in decision support tools, and eventually, it could make better decisions than humans.
Understanding Knowledge Engineering
Knowledge engineering started by trying to transfer problem-solving expertise from humans into programs that could take the same data and arrive at the same conclusions—this was called the transfer process, and it was the main approach early on.
It lost favor when we realized human decision-making knowledge isn't always explicit. Humans base decisions on past experiences of what worked, but they also draw from unrelated knowledge pools that don't seem logically connected.
What some call gut feelings or intuitive leaps is really analogous reasoning and nonlinear thinking—these don't fit into straightforward decision trees and might involve data that seems inefficient to process.
We've moved to a modeling process instead of copying steps exactly. Now, we focus on systems that reach the same results as experts without the same paths or sources, avoiding the challenge of pinpointing unconscious knowledge.
As long as conclusions match, the model is valid. We can refine it by debugging bad outcomes and promoting processes that match or improve on expert results.
Knowledge Engineering to Exceed Human Experts
Knowledge engineering is already built into decision support software. We employ specialized knowledge engineers in fields advancing human-like machine functions, like facial recognition or understanding spoken language for meaning.
As models get more complex, even engineers might not fully grasp how conclusions are reached. The field will evolve from matching human problem-solving to outperforming humans quantitatively.
By combining these models with natural language processing and facial recognition, AI could become the ultimate server, financial adviser, or travel agent you've ever encountered.
