According to the UK-based AI company that was bought out by Google in 2014, it trained its AI by using a total of 91000 “de-identified” data of women residing both in the US and UK. Upon completion, the AI’s algorithm then tested 25000 scans in the UK and an additional 3000 scans from the US. Four images of each test were then put into its neural network with scores between zero to one. The latter being a subject of higher risk. Ultimately, DeepMind proved that it was able to reduce false positive by 5.7% and false negatives by 9.4% in the US. While the UK readings saw a reduction of 1.2% for false positives and 2.7% for false negatives, suggesting that the US detection system is less accurate than the UK’s. Beyond the scope of healthcare, DeepMind is more famously known among the AI and tech circles as the AI that defeated a reigning Korean Go player back in 2017. Additionally, DeepMind has also trained its AI in popular game titles, including Starcraft II and Quake III Arena. (Source: Wired via Engadget, Hot Hardware // Image: BSIP via Getty Images)
