Do you know the basics underlying every AI system? - Access Innovation

I was recently asked “what’s going on in the world of discovery and search that’s being influenced or disrupted by AI and ML systems?” Well, everything! Everything to do with discovery and search is and always has been based on AI systems, an ML component included or not.

AI systems are baked into our everyday lives at this point, but do you know the basics underlying every AI system?

One thing required for any AI: an expert system. An expert system has two components: a knowledge or rules base and an inference engine. The knowledge or rules base is made up of rules based on Boolean syntax: IF/ THEN statements: IF a condition is met (likely a text string), THEN take the action in the rule.

Large tech companies tout their natural language processing and machine learning algorithms. Those algorithms are creating, ultimately, IF/ THEN statements. Furthermore, the rules are created using either forward chaining or backward chaining. Forward chaining starts with the known facts and combines them to reach a conclusion. From forward chaining, we get AI fails along the lines of Microsoft’s ill-fated chatbot that was trained on transcriptions that mimicked the conversation of teenage girls. Backward chaining starts with a goal, makes logical decisions about how to reach that goal, and comes to a conclusion. The challenge with backward chaining is that it ignores data that doesn’t fit getting to the goal, so results are easily skewed.

It’s prudent to make sure you know how rules are being created in your AI.

---

Access Innovations, Inc.
Web: www.accessinn.com