Brennon Williams

View Original

Developer test of fluid conversational AI system

As I have previously stated, cognitive AI is much more complex than the existing platforms of Amazon Alexa, Siri or any number of “Agent” based AI solutions currently being used.

When I look across all the other platforms out there trying to do similar things, I think for the moment that I may be a little further down the rabbit hole then most. I don’t expect that to last for a long time and the competition is catching up fast.

I wanted to show a short demonstration of ILI – the conversational AI platform that I have been working on for a number of years now and the platform that is powering the products we are developing at my company ISR.

Here you will see ILI "pretending" to work in a hotel, which is a training activity to re-enforce her voice models.

 

Fluid conversations are pretty hard to do in artificial intelligence. And once you get to a point or cyclical input/output, you soon realise that you need a lot more flexibility in the logical structures that are the basis for the conversation direction. Hence why I am on iteration 11.

It is one thing to design a conversation – but a whole lot of magic is needed to allow that conversation to be hyper flexible.

I intend to show more of my work in the coming posts – why we need the COG to “pretend”, conversational UI and some discussions around conversation designing.

 

Do me a favour – share this post out amongst your friends and network if you like what I am doing.