Machines that Understand have arrived. But there’s a catch.

Harpal Parmar - Jul 20, 2020 - 3 min read

Artificial Intelligence and Machine Learning (ML) are terms that abound in this age of digital transformation. ML is based on statistical pattern matching – it’s great for finding you a song amongst thousands or identifying an anomaly within a document - after being trained on a large data set of similar documents. Does this system understand the meaning and context of the language and information in the documents? No, and we agree with Melanie Mitchell on this topic, but it’s effective enough to solve some practical problems.

At the other end of the AI spectrum is AGI (Artificial General Intelligence), of the kind embodied by the character “Data” of Star Trek fame. This android robot functioned at mental (and physical) capacities close to human – with forgivable challenges in understanding puns, jokes and complex human emotions. Achieving this level of AI has some time to go, as some experts opine.

In the meantime, is it possible for AI to deliver results beyond the pattern matching ML or of the Deep Learning kind? An AI that truly understands context and meaning in language? And can apply logic, reasoning, and math to information? And thereby can deliver value as a digital partner to humans – by making contextual sense of complex notes and transcripts to extract embedded insights? Or by truly understanding contact center conversations to identify risk and issues that may be escalated and lead to financial damages?

Yes, but there’s a catch. Machines can truly understand and extract meaningful information from complex language – when they mimic human cognition. We know that humans derive context and meaning from what they read or hear by referring to their very own database – the knowledge existent in their brains!

We’ve created such an Understanding machine at AUI Systems. With its own brain – a semantic knowledge base. One that’s flexible, transparent and scalable. A teachable brain. Not data-driven, but knowledge-based.

The result is true understanding of complex earnings call transcripts, contact center conversations, and myriad other knowledge that can be taught to this “brain” and is then accessible via natural language interaction. So, while the Star Trek android will become a reality one day, practical industry problems beyond the reach of machine learning are being solved by our Understanding system. We’ll be sharing more in subsequent articles here and on social media.