On Monday, we got a glimpse of Talla, the new virtual assistant startup from Rob May. In a video posted to YouTube and circulated among Talla’s investors, Talla shows off what it can do.
The video isn’t likely to go viral: It’s a screen share of an instant-messaging exchange between a user and Talla’s chatbot, in which the user gets help managing a to-do list, catching up on messages from a prospective hire and scheduling a meeting to discuss the hire.
The AI startup Viv wowed the world with a virtual assistant that handles speech requests (and Siri continues to wow exactly nobody). Talla’s demo video is not likely to go viral.
Boston-based Talla uses natural language processing and is focused on text. In an Xconomy interview in January, May imagined a future in which “every knowledge worker will have a virtual assistant to help them do their job.” Text, he said, will be “the primary way we communicate with them.”
So, Talla is a lot like virtual assistants you may have already encountered in email–Amy, for example, from New York-based x.ai, or Boston-based Vesper. Talla, based in chat, is just a little more conversational.
Monday’s demo makes it easy to grok how Talla works. If you had an assistant, you wouldn’t walk over to his desk and ask him to help you schedule a meeting. You’d hit him up on Slack. Talla is like a smart Slack bot that does a little more than, say, the ones we’ve built here at BostInno HQ, which are trained to respond to certain keywords with ridiculous gifs.
Talla, which raised about $4 million in January from investors including Avalon Ventures and Jason Calacanis, has goals that are similar to Viv, Siri or Amazon’s Echo–just based in text and limited to the workplace. I asked May a few questions about how those limitations impact the startup’s work and its product:
GM: How does text impact the user experience in a work setting?
RM: Voice is a different problem – it’s a signal processing problem about picking out words given different accents and background noise, etc. NLP is about translating words to intent. So the technology is very very different. But ultimately, I can’t see an office full of people all talking to their assistants for quick personal tasks. It would be too chaotic. The experience we are trying to replicate is the one where you and I work together, and I need to ask you to do something simple, or pull a simple piece of information. I wouldn’t walk over to your desk to do that, I would just hit you up on chat. But these are tasks an A.I. can do well.
GM: How does it impact the technical challenges?
RM: The technical challenges are incredibly difficult, which is why we have hired a very senior team.
GM: How has focusing on a narrower context (work) given you an advantage?
RM: By constraining the problems, we can make a lot of assumptions that make the product more powerful for that specific domain. I think companies like x.ai that are going very horizontal are making a mistake.
GM: Does your virtual assistant have a name, other than Talla?
RM: No. We debate this as part of our product strategy but, for now it’s just Talla.