deepneuralnetworks
Latest
Microsoft says its AI can translate Chinese as well as a human
Run Chinese text through a translation website and the results tend to be messy, to put it mildly. You might get the gist of what's being said, but the sheer differences between languages usually lead to mangled sentences without any trace of fluency or subtlety. Microsoft might have just conquered that problem, however: it has developed an AI said to translate Chinese to English with the same quality as a human. You can even try it yourself. The trick, Microsoft said, was to change how it trained AI.
AliveCor wearables may detect unsafe potassium levels in the future
AliveCor is working on a new application for the technology behind its KardiaBand for Apple Watch. Last year, the FDA approved KardiaBand as a medical device, and it can record your heart rhythm and report on any rhythmic abnormalities that could be linked to dangerous health issues. Now, The Verge reports, AliveCor is developing a way for its technology to be used to detect high levels of potassium in the blood.
Alibaba and Microsoft AI beat human scores on Stanford reading test
20 years ago, IBM's Deep Blue beat Garry Kasparov in a game of chess and the practice of pitting human against computer continues to this day. Google's AI was the first to beat a Go champion a couple of years back (and it continues to teach itself how to play other games, too). Now Bloomberg reports that Alibaba and Microsoft have both developed AI that scores better than humans on a Stanford University reading test.
IBM wants to accelerate AI learning with new processor tech
Deep neural networks (DNNs) can be taught nearly anything, including how to beat us at our own games. The problem is that training AI systems ties up big-ticket supercomputers or data centers for days at a time. Scientists from IBM's T.J. Watson Research Center think they can cut the horsepower and learning times drastically using "resistive processing units," theoretical chips that combine CPU and non-volatile memory. Those could accelerate data speeds exponentially, resulting in systems that can do tasks like "natural speech recognition and translation between all world languages," according to the team.
YouTube trains its thumbnailer to generate better images
Since a YouTube video's thumbnail could convince or deter viewers from hitting play, we're sure a lot of creators would love the website's new and improved automatic thumbnailer. According to the Google Research Blog's announcement post, the website analyzes videos you upload at one frame per second and scores each frame based on their quality. The generator will then display images that scored high enough during analysis. But, how exactly does YouTube's algorithm know how to score images? Apparently, by training it to recognize between good and bad photos.
Google voice search can now hear what you've got to say
Need a better reason to say "OK Google" in public? Well, the search engine just announced a refinement to voice search on its research blog. Now, it can more accurately pick out your voice even in noisy surroundings, and perform the necessary analysis in real-time. This is all made possible by software that can make faster phoneme predictions by listening further ahead in a user's speech.
