Emerging tech news
CURVED SAMSUNG OLED VR SAMSUNG TELLS VR TO GET BENT.
A patient filed by Samsung in the US Patent and Trademark Office was revealed in January to include a VR headset using a curved OLED display. The design seems to allow for a full widescreen 180 degree field of view without the need to extend the screen laterally beyond the extremities of your head. The industrious design uses two types of lenses to achieve a full vertical and a partial horizontal bifocal field of view without being overly cumbersome.
AI USED TO HELP ACCELERATE NUCLEAR FUSION RESEARCH MACHINE LEARNING MAY BE THE KEY TO SUSTAINED FUSION ENERGY.
Researchers at energy startup TAE Technologies have been using hot plasma colliders to try to harness the relatively benign power of nuclear fusion for some time, but it’s turned to AI for help in recent years. TAE twirls plasma inside a straight $150 million collider using magnets, and heats it to temperatures hotter than the sun’s core to create fusion reactions. Doing this involves a considerably large number of variables and TAE has collaborated with Google to create a machine learning system that keeps track of variables and proposes new combinations for optimum results.
DEEPFAKES BUSCEMI-LAWRENCE AT LEAST NOW WE KNOW WHAT ANDROIDS REALLY DREAM OF.
We’re not exactly sure why @ vilan_guy decided to feed Steve Buscemi’s face into a deepfake machine learning algorithm and then blended it with an interview of Jennifer Lawrence revealing her favourite Real Housewives star, but we do know that Buscemi-Lawrence is something you can never un-see. From its overnight infamy as a fake celebrity porn generator to a University of California Berkeley study that transfers dance moves, we’ve covered Deepfakes a couple of times here, but this inexplicable star-crossed paring is a vivid reminder of just how advanced this visual machine learning model is.
AI CAN NOW READ YOUR MIND BRAIN IMPLANTS USED TO DECODE THE NUMBERS YOU’RE THINKING.
Combining deep learning with a speech synthesiser, researchers form Columbia University have developed a system that can convert brain signals into words. The team used existing brain electrode implants that are generally used to monitor seizure patterns in 5 epilepsy patients to recognise the brain activity of their spoken words. The test then read 40 numbers to the patients and the AI was asked to guess what numbers had been asked by analyzing brain patterns. The AI was able to accurately record numbers with 75% accuracy.