Developers around the world are striving to build software that leverages conversational AI. It is helpful to split these efforts to 5 levels of capability, in analogy with research on autonomous vehicles. There is no straightforward dichotomy between dumb chatbots and AGI, but a series of gradations which can help us understand progress. I will talk about the key characteristics of level 3 conversational AI, which is where the most advanced teams are today. I will discuss the difficult machine learning and engineering problems that this poses, and how we see solutions developing within the Rasa community, leveraging research in the field to advance the state of the art.
The 5 Levels of AI Assistants
Nathaniel (Natty) Kohn is the Data Science Lead at Lemonade, an insurance company powered by AI and behavioral economics, and driven by social good. A seasoned engineer, Natty was previously VP Smart Interactions at Citi Bank, and prior to that, worked on artificial intelligence and machine learning at Intel and Checkpoint.
Dr. Alan Nichol is the co-founder and CTO of Rasa, the standard open source infrastructure for conversational AI. He holds a PhD in machine learning from the University of Cambridge and has years of experience building AI products in industry.