by Valentine Baldassari
Google’s new artificial intelligence assistant Duplex faces criticism over allegations the company faked or exaggerated its demonstration.
The company presented the tool on 8 May, demonstrating it with two phone calls booking a hairdressing appointment and a reservation at a restaurant. In the audio recording, Duplex sounds like a real person and uses pauses and signs of hesitation such as “um.”
Professor Artur Garcez, the director of the Research Centre for Machine Learning at City, University of London, says it is impressive but thinks Duplex’s capacities have been overstated: “What is impressive is the level of language processing. This may be useful in a number of situations, but specific, practical situations.
“I can see many conversations where you have unforeseen questions coming up and then the system will get confused.”
There were also some issues with the recordings themselves. When you call a business, the person usually immediately identifies the establishment, but this didn’t happen in the two sample calls.
To prove this, news site Axios called restaurants and hairdressers in Google’s hometown of Mountain View in California, and said: “Every one immediately gave the business name.”
In addition, the employee did not ask for a phone number or name and there is no background noise in the recordings.
None of this proves Google had edited the calls before the demonstration, but if true, it could mean the system was not reacting to an unscripted, real-world scenario.
It is not the first time that Google Duplex has drawn criticism, despite how recently it has been announced. Reactions to the demonstration were mixed, particularly because the assistant never identified itself as a robot in the recording, potentially leading the person on the other end to think they were talking to another human being.
Zeynep Tufekci, a technology expert and professor, tweeted that Duplex was horrifying. She added: “Silicon Valley is ethically lost, rudderless and has not learned a thing.”
A Google spokesperson clarified that the system would identify itself as an artificial intelligence: “We are designing this feature with disclosure built-in, and we’ll make sure the system is appropriately identified. What we showed at I/O was an early technology demo, and we look forward to incorporating feedback as we develop this into a product.”
Featured Image by ergoneon on Pixabay