Google now says controversial AI voice calling system will identify itself to humans

Following a widespread protest over the ethical dilemmas posed by Google's new Duplex system, which allows AI to imitate a human voice to make appointments, Google now says that the experimental system will have "built-in revelation". That seems to mean what eventually Duplex takes as a consumer product will involve some kind of verbal announcement to the person on the other side who is really talking to an AI. The news of the change in Google's approach to Duplex transparency was reported for the first time this afternoon at CNET .

Duplex is not yet a running product, and Google CEO Sundar Pichai introduced him on stage at the company's I / O developer conference on Tuesday with just a pre-recorded phone call. That demo showed how the Google Assistant could sound much more real when using DeepMind's new WaveNet audio generation technique and other advances in natural language processing, all of which helps the software to more realistically replicate the patterns of the speaks human For example, Duplex is so convincingly human because Google includes tics like "uh" and "um" and other more colloquial phrases in the Assistant's verbal library.

However, a piece of software deliberately deceives a human being, in this case a hairdressing receptionist, caused massive alarm among technology critics and those who fear that artificial intelligence technologies will develop without supervision or regulation adequate Technological critic Zeynep Tufekci described the demo as "horrible" and the positive initial public reaction in I / O as evidence that "Silicon Valley is ethically lost, rudderless and has learned nothing".

"We understand and value the discussion on Google Duplex: as we said from the beginning, transparency in technology is important , "A Google spokesman said The Verge in a statement. "We are designing this feature with the built-in disclosure, and we will make sure that the system is properly identified." What we showed at I / O was an early technology demonstration, and we hope to incorporate comments as we develop this into a product. "

Google had originally said in a blog post written by engineers Yaniv Leviathan and Yossi Matias that "For us it is important that users and businesses have a good experience with this service, and transparency is a key part of that " The duo added that "we want to be clear about the intent of the call for companies to understand the context" and that Google will experience how to get ahead in the months before Duplex is expected to enter the tests within the Assistant platform. summer.

But the company did not explicitly state that disclosure would be a mandatory feature. Later, after the presentation, Google representatives said The Verge that the company felt responsible for informing people that they might be talking about software, and that the assistant team was looking to safeguard the protection against improper uses such as spam. Now, it seems that Google takes a tougher stance on transparency after the online protest. That includes making sure that Duplex will be "properly identified" in the future, for the benefit of all parties involved.