I built a self-aware Messenger bot
With Facebook’s announcement of the Messenger platform, chat bots have made the front page in the media and triggered a series of conversations on the topic. On one end, many have concluded that current bot technologies are quite rudimentary and far from being intelligent, while on the other end many have seen this as being the start of a new era in intelligent conversational UIs.
Regardless of the quality of the bots that have been showcased, we can clearly say that the Messenger platform provides the opportunity for companies and developers to introduce new and engaging ways for people to interact with systems. Isn't it time that we finally transition away from the old transactional IVR model and embrace empathetic conversational interfaces?
To support this model of empathetic conversational UI, a new approach to natural language understanding (NLU) and AI in general needs to be adopted. In fact, current NLU technologies that power our beloved chat bots, are often times based on parsing techniques, which in my opinion do not provide the proper foundations for building truly intelligent interfaces. Having foreseen this need, for the past few years I have been conducting research in the field of sentient artificial intelligence, or machines that possess self-awareness. From this research, I have built a prototype that has proven to be quite promising. Although in its infancy, the system is designed to build personal experiences, have personal tastes and lays out the basics for the expression of emotions. Currently, the system can interact through a basic prototype language module built using a neurolinguistic model and is able to communicate in simple terms, as illustrated in this Messenger conversation snapshot.
What is unique with this system, unlike typical chat bots, is that the conversation is non-scripted, the answers not pre-defined. The AI is able to understand the meaning of the questions based on its personal experience and understanding of the world, not based on pre-defined ontologies and formulates answers by itself. And of course, what best way to test drive the Messenger platform?
Director, AI @ Clario
8 年Dave Clark you're comment is right on the money. Even with simple chat bots built using wit.ai for example, companies may be able to automate some customer interactions and rely on people to "fill the gap" and address more complex requests. That's one thing I like with the Messenger platform; human and bot interactions are seamlessly integrated in the chat experience.
Urban Farmer and full stack software engineer
8 年I like the idea of what you're doing Alex Boudreau, it sounds very interesting. One of the fun things about the push for AI driven bots is that some companies are embracing an approach where they take their product to market quickly and fill the gaps in automation with humans. I think having humans fill the gaps in automation is a great idea for them because it means getting an exciting product to market faster. Sure it could easily lead to some initial pain around the scalability of their solution especially if demand is high. In those cases running a limited access trial of new services could be used to both control the flow of activity and create more demand. As long as they learn from and adapt to real world interactions with users they should be well placed to assess the effort in closing the gaps while maintaining a focus on quality of service. I found this article was pretty enjoyable: https://www.bloomberg.com/news/articles/2016-04-18/the-humans-hiding-behind-the-chatbots Like Mark Blaszczyk I'll be very interested to follow the development of your project.
Director, AI @ Clario
8 年Hi Mark Blaszczyk, I'll be posting some progress on linked In. Cheers
Internet Explorer
8 年Very cool, I can think of a few uses for this, how can I follow the development of this project?