Read on Twitter
So inspired by the @voicebotai podcast with @alanmnichol - https://voicebot.ai/2020/07/05/alan-nichol-cto-and-co-founder-of-rasa-voicebot-podcast-ep-157/, I spent time transcribing this morning.
"The key piece is learning from real conversations, looking at what people are doing and then bringing in best practices from software development."
"Because I think this is another key piece and issue I have with a lot of tools out there in market, is that they don't have any support for tests, right? And I think that is inexcusable."
"If you are working with a conversational AI platform and it doesn't support creating and writing tests as part of your deployment cycle, then it's just a prototyping tool. It's not something that is fit for production. Professional teams do not ship applications without tests."
"And just because we are building something which has a machine-learning component, doesn't absolve us of that responsibility to write those tests and run those tests."
My only revision - I would not say "just" - I would say "especially because we are building something which has a machine-learning." The dynamism of machine learning means automated testing is even MORE important.
Overall, I love this conversation and really appreciate the focus that @RasaHQ and @alanmnichol are bringing to raising the bar on Conversational AI dev practices. It is much needed, long overdue, and enthusiastically welcomed by us @bespokenio
