It’s been almost a year since I joined Human Managed, and so I wanted to celebrate this milestone by writing about how I got acquainted with, and later on became deeply immersed in Conversational AI.
I received my PhD in Applied Linguistics about 6 years ago in hopes that I could use it anywhere but the academe. The reason why demands a separate and more personal blog post. I digress.
I stopped hoping after realizing that being a female forensic linguist in Manila involved too many risks and also after receiving feedback from traditional corporations that I’m “overqualified” for the positions I were applying for. It was a sad fate because in almost every aspect of my life, never did I mind being a beginner for the sake of learning.
Fast forward to July 2019 when I was introduced to Saleem, one of our founders. He told me that he had been looking for a linguist to be one of the drivers of HM’s Conversational AI campaign. I clarified that my degree didn’t cover Computational Linguistics, but he confirmed that he knew that and didn’t need another technical person in the team. I remember being intimidated after hearing the word AI but also feeling excited about the possibilities it would open for learning and growth. I decided to use that hint of self-doubt to my advantage and kick start my journey. But starting wasn’t as easy as I hoped.
As I joined HM in August 2019, I immediately brainstormed my first use case: a mission critical conversational interface for CxOs to extract real-time information on their organization’s security posture. I was just dipping my toes into Cyber Security as well so I began to write my first Happy Path (note that I didn’t know this term back then, let alone conversation design) to see if my understanding of the task was clear. Unfortunately, I don’t have a copy of this first conversation flow, but I do remember that the comment I received for it was, “Why does your bot sound like a know it all?”
It’s funny thinking about it now, but I was panicking then. Apparently, I shouldn’t be cramming too much information in the response (for reasons I promise to elaborate in a separate post). Sensing that I was somehow lost, I was advised to observe how comics are written and to research on work that’s being done in RASA. This was the turning point for me to stop creating and start learning.
The easy route isn’t always the best route
Conversational AI was a relatively new field back then and very few organizations had been exerting efforts to democratize it. RASA was one of the first companies who made it their mission to open source the standard infrastructure they have built for conversational AI, and with the help of a community that has grown exponentially within the past year, they have gone a long way to develop their technology, from introducing a UX to train chatbots to making their components ready for deployment with Docker or Kubernetes.
I also came across Dialogflow, which seemed to be the stronger contender for a starting point given it’s practically ready to use even without any coding experience. However, even though going Google was the easier route, I knew it wasn’t the best fit given the complexity of our use case. I also thought that if I were to switch to Dialogflow at some point in time, the transition would be so much easier, and the learning curve wouldn’t be as steep.
Coincidentally, I also came across a post marketing the September 2019 RASA Summit in San Francisco. The agenda seemed to cover topics I needed to know to become a “legitimate” member of the Conversational AI community. I made a pact to myself that by the time I got to the Bay Area, I should have already demonstrated understanding of how the components worked by building and training a simple FAQ chatbot using their platform. I made it my mission knowing it was the best way to get the most out of the sessions from the summit. This initiative was the start of how my knowledge expanded to Conversation Design to Machine Learning frameworks. The rest, as they say, is history.
The tweet that deserved a pair of socks. I won for being one of the participants who traveled the farthest for the conference.
Aside from knowing what’s inside the technology, learn more about what you can get out of it
It’s true that I shouldn’t have focused on the technology first. But in my attempt to uncover the value we can get out of RASA’s framework, I was able to gain an understanding of not only what constitutes good conversation design but also what it takes to learn something new in my 30s.
First is that I’m still capable of learning something new even if it’s a topic you’re intimidated by. Truth be told, I’ve spent most of my life not knowing how to launch the CMD prompt whereas now, I can manage it from a Type 2 hypervisor. I now understand the high programmers get when they are able to successfully execute commands.
Second is that learning is amplified when you are part of a community. I guess what even drew me closer to RASA is that, like us here at HM, they have high regard for their community. The web of network they have built opened the door for me to meet the key players in the industry, and the best practices they adhere to. I came across advocates of good conversation design and tools our team can use from inception to launch.
This is just the beginning! I look forward to sharing more of my chronicles with you.
Note: This blog was originally published in August 2019 and has been updated for accuracy.