Virtual AI assistants are rapidly learning to make our daily lives easier but there is still a conversational barrier that separates them from us. But, Microsoft believes that it has cracked the code and you’ll soon not be required to talk to a dull and boring AI which doesn’t know what to do at most instances.
The researchers at Redmond have achieved an exceptional feat in the field of speech recognition, and thus, dispensed details on the technology that recognizes words in a conversation as well as a human. The trained neural networks embedded within NIST 2000 automated system were pitted against professional transcriptionists, and the results were truly unexpected. It was the first time that a machine has come so close(or beaten) a human, and produced a comparable(or lower) rate of error as compared to them.
We’ve reached human parity. This is an historic achievement.
says Xuedong Huang, the company’s chief speech scientist.
The team of engineers at Microsoft Artificial Intelligence and Research division have published a paper to report that they’ve recorded the lowest error rate against the industry standard Switchboard speech recognition task. They team has been working towards this goal for the past year, and has now achieved the lowest word error rate(WER) of 5.9 per cent, which is roughly equal to that of an average human. The team also beat their own record of 6.3 per cent WER which they’ve acheived just last month.
This research milestone has been made possible due to deep neural networks that use large amounts of data — called training sets — to teach computer systems to recognize patterns. The team adds that it used both sounds and images to train the conversational AI but there is still much work to be done to employ the service in real life situations.
The Microsoft AI team is currently focusing on ways to ensure that the speech recognition system can work in places where there is a lot of background noise, such as at a party or while driving on the highway. They’ll also work on making it more human-like by helping it learn, understand voices and assign names to it. This will most likely be the next step towards the development of Cortana as an omnipresent AI assistant.
Excited about the impact of this achievement on future technological developments at Microsoft,
The milestone will have broad implications for consumer and business products that can be significantly augmented by speech recognition. That includes consumer entertainment devices like the Xbox, accessibility tools such as instant speech-to-text transcription and personal digital assistants such as Cortana.
Yes, what you’re reading is true! And your dream of living within I-Robot world could be fulfilled in the near future.
