Jeff Herbst, vice-president of business development and head of Nvidia GPU Ventures, graphics technology producer Nvidia’s corporate venturing arm, moderated a presentation in which Jonathan Cohen, senior director of artificial intelligence (AI) software at Nvidia, and Scott Stephenson, co-founder and chief executive at artificial intelligence technology developer Deepgram, focused on speech recognition.
Cohen and Stephenson stated that there were three problems in conversational AI technology, including speech recognition, natural language processing and achieving a response that sounded like a natural voice with intonation – and that speech recognition alone was enough for a company like Deepgram to be built upon.
Stephenson said that there were two worlds developing in speech recognition. Most were aware of the consumer world which included AI on the phone to look up things like a recipe or the weather. Many were however less familiar with the business world, where companies were discovering needs for speech recognition software, including real-time conversation and simultaneous feedback.
Although Stephenson said a lot of hurdles to conversational AI had been removed and awareness of the technology had been brought to market, many difficult problems with speech recognition software were still unresolved. Cohen said that discovering the answers to all of the unknowns of speech recognition was difficult, but that there was value in each piece of the puzzle that was solved.