Tag Archives: Artificial Intelligence

These People Do Not Exist

The five “individuals” below have one thing in common: They are all the product of machine learning and data mining technologies brought together and introduced by computer scientists several years ago that are fully-operational today.

“Generative Adversarial Networks” provide a way for an individual or entity to create amazingly lifelike images of fictitious people upon which any identification and biography may be applied. Such graphics may then be used for any number of purposes–public relations/promotion and advertising, entertainment vehicles, or perhaps even controversial public events.

From the LyrnAI blog:

Generative Adversarial Networks (GAN) are a relatively new concept in Machine Learning, introduced for the first time in 2014. Their goal is to synthesize artificial samples, such as images, that are indistinguishable from authentic images. A common example of a GAN application is to generate artificial face images by learning from a dataset of celebrity faces. While GAN images became more realistic over time, one of their main challenges is controlling their output, i.e. changing specific features such pose, face shape and hair style in an image of a face.  

A new paper by NVIDIA, A Style-Based Generator Architecture for GANs (StyleGAN), presents a novel model which addresses this challenge. StyleGAN generates the artificial image gradually, starting from a very low resolution and continuing to a high resolution (1024×1024). By modifying the input of each level separately, it controls the visual features that are expressed in that level, from coarse features (pose, face shape) to fine details (hair color), without affecting other levels. 

This technique not only allows for a better understanding of the generated output, but also produces state-of-the-art results – high-res images that look more authentic than previously generated images.

You can demonstrate this technology and create your own cast of fictitious “persons” by visiting thispersondoesnotexist.com.

error0

Facebook Intensifies Censorship of User Content

“Hate Speech” Difficult to Police

Social media behemoth Facebook, with over two billion users worldwide, has issued a censorship report for the first quarter of 2018. The document states that its recently-deployed artificial intelligence censor-bots flag or eliminate over 85% of images posted containing “graphic violence.”

The same techniques have been successful in addressing 1.9 million in the same period posts promoting “terrorist propaganda,” the company said.

Facebook’s automated censors have much greater difficulty detecting “racist or homophobic hate speech,” because it “racist is often quoted on posts by their targets or activists,” AFP reports.

“It may take a human to understand and accurately interpret nuances like… self-referential comments or sarcasm,” the report said, noting that Facebook aims to “protect and respect both expression and personal safety”.

Facebook took action against 2.5 million pieces of hate speech content during the period, a 56 increase over October-December. But only 38 percent had been detected through Facebook’s efforts — the rest flagged up by users.

The posts that keep the Facebook reviewers the busiest are those showing adult nudity or sexual activity — quite apart from child pornography, which is not covered by the report.

Some 21 million such posts were handled in the period, a similar number to October-December 2017.

That was less than 0.1 percent of viewed content — which includes text, images, videos, links, live videos or comments on posts — Facebook said, adding it had dealt with nearly 96 percent of the cases before being alerted to them.

error0