Microsoft is developing a new web browser around Chromium, the same open-source strain Google uses for Chrome, says WindowsCentral. According to the site’s sources, the publisher has decided to drop EdgeHTML, the engine currently used in Edge and which it has been using since 2015 in Windows 10.
This engine has never been able to convince or get Edge to take off against its competitors. According to Net MarketShare statements, Edge has a 4.2% market share on desktop PCs. It is not far ahead of Safari, which can only express itself on Mac (3.8%) and still behind Internet Explorer 11 (8.9%), Firefox (10%) and Chrome, which dominates with its head and shoulders (63.6%).
The code name of this next software that will become the default browser for Windows 10 is named after a Californian city: Anaheim.
Recently, two Microsoft developers had been seen contributing to the development of Chromium, but this had been interpreted as assistance in porting Google’s browser, Chrome, to a native version for Windows 10 on an ARM.
Microsoft is not in uncharted territory; it had already made this effort to run Edge on another engine than its own when it launched the Android and iOS versions of the browser a year ago.
This has of course nothing to do with the infamous Chromium malware that appeared lately :
Three Nvidia researchers have created an artificial intelligence capable of generating realistic photos of individuals who do not even exist. How cool/creepy is that?
Beyond technical skills, we are already questioning our ability to distinguish between the true and the fake.
Generated by neural networks trained by 3 Nvidia researchers from scratch, these perfectly realistic pictures present a portrait of people who do not exist in the real world. The three scientists published a video, accompanied by their research work, on December 12, 2018. They should put their code and database in open source in the next few days.
What is the technology behind it?
To achieve this result, they relied on a generative adversarial network (GAN), an automatic learning system used in particular by artists, including those of the painting recently sold for $400,000. This automatic learning system puts two neural networks in competition. The first, the generator, generates images from the data provided to it. The second, the discriminator, points out those that are too close to the photos he receives. The tension between these two neural networks creates a learning loop, resulting in an improved GAN.
Nvidia researchers have built their own GAN that distinguishes between different “styles.” For example, it separates “high level” or “crude” attributes such as posture or facial shape, and “medium level” attributes such as nose, eyes, and mouth. Finally, there is a
Researchers can then adjust the level of variation for each of the styles (coarse, medium, or refined) to generate more or less similar faces.
Finally, the scientists considered several aspects of a human portrait to be random (or stochastic), such as hair position, pores, or freckles. “Therefore, we can generate them randomly without affecting our perception of the image, as long as they are correctly distributed,” they write in their study. By inserting statistical noise, they generate these details randomly, realistically.
For which purposes? What are the applications?
Nvidia researchers created these neural networks, and their code, as well as the databases in which they have been trained, will be put into open source very soon. Therefore, the Californian company should not get any direct benefit from the GAN developed by its researchers but can expect to get indirect benefits if other companies appropriate it. Nvidia remains the market leader in GPUs, the graphics processors initially used by the gaming industry. They are now coveted for their ability to perform several calculations in parallel, very useful in learning the machine, and essential for operating such artificial intelligence.