Behind the scenes: what makes Dasha AI tick
Arthur Grishkevich2 minute read
Voice is the native interface for human communication. So why is it that we still communicate with our technology using the tips of our fingers?
True human-machine voice communication requires that the machine be at least as capable as the human of deriving meaning and expressing “thoughts”. Without this capacity, any human-machine voice interaction will deteriorate into a command-action relationship. If you’ve ever had a significant other, you know that kind of relationship is no relationship at all.
In a nutshell, our technological limit is the reason we are still using our fingers (non-native interface) to communicate with machines.
Our goal is to break through that limit.
Dasha AI platform lets you build voice interfaces that communicate at the human level already today. With time, our models will enable humans to communicate with machines on any topic at a fully conversational, natural language level. The same way you communicate with your parents in law.
Since language is, in a lot of ways, how our intelligence is framed, the challenge of talking machines is one of the more important ones to be solved on the rocky path to general artificial intelligence.
Dasha AI is a product born out of a real world pain. In 2017 Alex Zaytcev was running a family manufacturing business and needed to expand his reseller network. The business was small, so he hired a team of three operator agents to call on the resellers. Expectedly, the agents were undermotivated, required training and performed less than optimally. Alex worked with his cousin Vlad Chernyshov to build the first prototype of Dasha AI.
The prototype performed its function more effectively than the three operators and cost significantly less to run. At the point the two friends realized the potential of this tech to positively impact other businesses. As the product progressed, so did the thinking around the importance of human-likeness in conversations with AI. This led the team to progress beyond thinking of itself as a product for call centers and into tackling full human-indistinguishable voice AI.
Right now we are taking baby steps by focusing our AI models on automating call center communications. With time and data we expect to progress to the level of general conversational AI.