AI Weekly Roundup
Is tricking AI into ignoring your face the best way to protect your privacy? The cat-and-mouse game between commercial facial recognition systems and AI researchers took another turn in the guise of a new tool that hampers the former’s ability to train algorithms using personal data. It means that when AI sees an image it not only mis-identifies it, it completely ignores it.
By making tiny changes to the facial images, this new system can throw an AI off the scent and encourage it to ignore the face altogether, rendering the images ‘unlearnable examples’. Meanwhile another of our stories explains how IBM wants to teach AI to write its own code, and don’t miss DeepMind’s team-up with Liverpool to explore AI in football!
P.S. Time to get your free virtual pass to CogX Festival 2021! Apply our code: cogx21news and use: cogx21newssvp for 50% off any festival pass. Sign up now.
P.P.S. ‘What happens when superhuman vision meets superhuman strength?’ Delve into the latest innovations in AI at CogX Festival 2021.
P.P.P.S. Need some weekend reading? Get all the works of our CogX Festival 2021 speakers, including, ‘The Deficit Myth’ by Stephanie Kelton.
The more adversarial examples an AI sees, the better it gets at recognizing them. But because Erfani and her colleagues stop an AI from training on images in the first place, they claim this won’t happen with unlearnable examples. Find out more about the technology that might allow you to keep your selfie to yourself.
Read more (MIT Technology Review)
IBM has announced Project CodeNet, a large dataset that aims to help teach AI how to understand and even write code. CodeNet will lead to enhanced tools that help to speed up the writing and checking of code by humans by improving an AI’s own understanding of how to do such tasks.
Read more (AI News)