Nvidia has created an editing tool turning landscape sketches into genuine pictures while the Shenzhen government is trialing a facial recognition system for fare payment. What’s new in the field of artificial intelligence? Read on to find out more in our weekly digest.
Nvidia’s tool for landscape creation
This week, Nvidia has presented GauGAN, a neural network turning non-detailed sketches into realistic landscapes. It was trained by 1 million Flickr’s images. A usual blue line can be turned into a waterfall, gray figures – mountains, green clouds – leaves on the trees, etc. The neural network’s paintings are always unique even if the sketches fed into it are the same.
As reported by Nvidia’s representatives, GauGAN would not pursue commercial purposes, but the demo version of the tool will be available.
China to test facial recognition systems to pay for a fare
The Shenzhen government (China) is trialing facial recognition systems for subway fare payment. If the experiment is a success, such a technology will be introduced all over the country.
The artificial intelligence would allow Chinese citizens to do away with traditional subway payment fare. It would be automatically withdrawn from a personal card.
US researchers embed AI in a polygraph
Researches from Florida State University have developed artificial intelligence that detects a lie on the Internet by the sentences’ structure and typing speed. The demo testing has shown that the AI correctly points out the truth and a lie in 85% of cases while genuine lie detectors are right in 60-70%.
It turned out that both truth-tellers and liars use certain word combinations in their speech and type at different speed. For example, the words ‘always’ and ‘never’ were frequently used in untruthful sentences, and the response was quick. Those who were telling the truth often used the word ‘maybe’ and were typing the text in a more pensive manner.
AI to help children with foster care after parents’ divorce
coParenter has developed artificial intelligence to help solve the problem of children’s foster care.
coParenter allows ex-spouses to exchange documents, discuss divorce related issues, get qualified legal assistance, schedule time of visiting children, calculate daily expenses, etc. The neural network would also search for insulting words and delete them to prevent disputes.
The app was launched in January 2019 and has already helped resolve about 3 thousand conflicts. A yearly subscription to coParenter costs $200.
DeepMind’s mysterious Ethics Board to control general AI
DeepMind’s mysterious Ethics Board was closely guarded against the media since 2014 after Google bought the company. However, in 2019, Hal Hodson at The Economist came out with the report shedding light on the DeepMind and Google’s document called the ‘Ethics and Safety Review Agreement" in the year leading up to the acquisition.
The document lays down that if DeepMind builds artificial general intelligence (AGI) that completes any intellectual task that a human can, then the control of that machine will lie with the representatives of the Ethics Board. The participants’ names remain a secret.
You'll know more than your colleagues and business rivals.