AI-based transcription services have been around for a while now, but the shift to online meetings in recent times has almost made them less useful. However, more and more businesses are returning to face-to-face meetings, but with safeguards in place, like masks and physical distances. This can make automatic transcription more difficult, causing some to take notes and lose the flow of the conversation. However, Microsoft Garage’s latest project attempts to fix this with seemingly magical AI, as long as everyone in the meeting has an iPhone and the Group Transcribe app installed.
To be clear, Group Transcribe is designed for in-person meetings, not online meetings. While it can still be used for remote or virtual meetings, its effectiveness could decrease dramatically. Indeed, the application uses the collective audio input of all the phones connected during a meeting to create a “very precise transcript” which also includes who said what.
This “live” requirement also fuels Group Transcribe’s real-time translation capabilities. This means that participants can speak comfortably in their own language, and others can follow along with a live translated transcript. Group Transcribe supports over 80 languages, Microsoft Garage prides itself on it, but its seemingly magical power doesn’t come without a cost.
Like any AI-based transcription and translation system, Group Transcribe improves and grows based on the data provided to it. While Microsoft promises not to store audio or transcribed text recordings on its servers, Garage’s research team is calling on users to donate some of this data to help improve the system. This is a condition of acceptance, fortunately, and requires that all participants actually accept the donation.
Recordings and transcripts will be “de-identified” and split into extracts which will then be distributed to examiners. That said, Microsoft Garage is making it clear that humans will actually be involved in the processing of these snippets, possibly bringing back the nightmares around smart voice assistants and third-party contractors a few years ago.