AI lip-readers are being used on babysitters with parents getting alerts whenever foul language is used.
The better than human tech could soon also scan CCTV footage in the UK following tests on workers and hospital patients, VICE reports.
Skylark Labs chief Amarjot Singh said his company's AI tech can detect about 50 words associated with swearing, abuse and violence.
He said it was being used in private homes in India and is being piloted by the Indian government to detect employee harassment at Punjab State Power Corporation Limited.
Meanwhile, a project funded by the British Defence and Security Accelerator is at least a year away from being able to lip-read keywords from silent CCTV footage.
Liopa CEO Liam McQuillan admitted privacy concerns could result in it never being used, and is working on eliminating racial or gender bias in the tech's algorithms.
Its app SRAVI (Speech Recognition App for the Voice Impaired) has been used on hospital patients.
It can detect a few dozen phrases such as "I need suction" at 90% accuracy and the company hopes to get European regulatory approval for it to start selling to healthcare providers by August.
Get latest news headlines delivered free
Want all the latest shocking news and views from all over the world straight into your inbox?
We've got the best royal scoops, crime dramas and breaking stories – all delivered in that Daily Star style you love.
Our great newsletters will give you all you need to know, from hard news to that bit of glamour you need every day. They'll drop straight into your inbox and you can unsubscribe whenever you like.
You can sign up here – you won't regret it…
Surveillance company Motorola Solutions also has a patent for a lip-reading system designed to aid police.
Mr Singh said: “We’re doing it in the wild and trying to solve use cases that have a direct implication for the safety of people.
“I think there is merit since the designer can control the words the system should flag, so I think it’s still kind of OK.
"The risk over here is that once you start to calibrate the systems to pick up everyday speech in the wild, that is when it becomes very hairy [ethically].”
Experts say it will be years before lip-reading AI is advanced enough to interpret full conversations, if it ever happens at all.
But Imperial College researchers have now created programs that can lip read words with a record 88.5% accuracy.
A 2018 study found human versions only got 7% of the words right, going up to 14% when they were told the context.
Source: Read Full Article