14 DAY TRIAL // Facebook launched the M virtual assistant back in 2015 in beta form, but only to a limited number of users from Silicon Valley. Since then, the virtual assistant rolled out to more and more Messenger users, but its functionality is strictly limited to text.
Facebook’s virtual assistant M doesn’t have the capacity to reply to voice commands, like other AI assistants on the market. Amazon’s Alexa, Google Assistant, Siri and Samsung’s Bixby, as of recently, have the capacity to assist users with multiple tasks using voice control.
Messenger’s product boss Stan Chudnovsky has recently revealed why Facebook isn’t planning to provide users with voice-control options over M. In an interview with Recode, Chudnovsky has said that there are three steps that AI assistants with voice-control capabilities must complete in order to carry out tasks.
One of the steps involves recognizing when the user makes a specific request, rather than simply having a conversation. The second is the ability to translate the request into text so that the AI can understand and then fulfill the request.
Voice control over M will arrive “eventually”
Chudnovsky says that AI assistants must complete each of the individual tasks with a success rate of at least 90%, to make it functional and able to understand user requests. Facebook’s M currently focuses only on the two final steps, identifying text and carrying out the task.
“Once we nail those things then we can go into voice,” he added. Facebook is taking the cautious way with the AI assistant M, and intends to roll out the voice command feature only after the company is pleased with its performance, as not to disappoint users.
Facebook’s virtual assistant M is still in beta two years after it was first introduced, and it’s still primarily powered by humans, not actual machines. AI-powered responses were only recently added to Messenger for all users, offering options for sending or requesting money, tools for sharing user location or stickers based on previous interactions. Facebook appears to be interested in the idea of providing M with the ability to understand voice commands, but it’s unclear when that will actually happen.