Elon Musk has raided Meta's collection of talented researchers, despite Mark Zuckerberg reportedly offering some a fortune to choose his company instead. The workers were part of Zuckerberg's AI team, helping Meta in the global race to build superintelligence, an almost godlike form of artificial intelligence that could think for itself and be much smarter than any human. Musk himself has gloated about the departures, posting on X that'many strong Meta engineers have and are joining xAI and without the need for insane initial [compensation].' At least 14 Meta researchers and engineers have left for their new home at Musk's AI competitor since January, while others have fled to OpenAI, the creator of ChatGPT. A spokesperson for Meta told the Daily Mail: 'Some attrition is normal for any organization of this size.'

The release of DeepSeek R1 stunned Wall Street and Silicon Valley this month, spooking investors and impressing tech leaders. But amid all the talk, many overlooked a critical detail about the way the new Chinese AI model functions--a nuance that has researchers worried about humanity's ability to control sophisticated new artificial intelligence systems. It's all down to an innovation in how DeepSeek R1 was trained--one that led to surprising behaviors in an early version of the model, which researchers described in the technical documentation accompanying its release. During testing, researchers noticed that the model would spontaneously switch between English and Chinese while it was solving problems. When they forced it to stick to one language, thus making it easier for users to follow along, they found that the system's ability to solve the same problems would diminish.

Artificial intelligence is getting better and better at generating an image in response to a handful of words, with publicly available AI image generators such as DALL-E 2 and Stable Diffusion. Now, Meta researchers are taking AI a step further: they're using it to concoct videos from a text prompt. Meta CEO Mark Zuckerberg posted on Facebook on Thursday about the research, called Make-A-Video, with a 20-second clip that compiled several text prompts that Meta researchers used and the resulting (very short) videos. The prompts include "A teddy bear painting a self portrait," "A spaceship landing on Mars," "A baby sloth with a knitted hat trying to figure out a laptop," and "A robot surfing a wave in the ocean." The videos for each prompt are just a few seconds long, and they generally show what the prompt suggests (with the exception of the baby sloth, which doesn't look much like the actual creature), in a fairly low-resolution and somewhat jerky style.

Modern Virtual Reality applications require technology that supports photo-realistic human face rendering and restoration. Due to the social nature of people and their ability to read and convey emotions from minor changes in facial expressions, minute artifacts can cause the uncanny valley, which can be detrimental to the user experience. In order to solve challenging issues like innovative view synthesis and view-dependent effects modeling, several contemporary 3D telepresence techniques now make use of deep learning models and neural rendering. These methods are typically data hungry, and the efficiency of those models is directly influenced by the architecture of the capturing device and data pipeline. In order to push the envelope in such photo-realistic human face models, a sizable dataset of high-quality, multi-view facial photos encompassing a wide range of expressions is necessary. Such a dataset was presented by Meta researchers in recent work.

Meta researcher Angela Fan is employing a novel approach to get Wikipedia to include more biographies of women: She's using AI to write the rough drafts. Why it matters: Only about 20 percent of those profiled on the online encyclopedia are women, and many other groups are underrepresented on the site. How it works: Facebook's parent company is releasing as open source software an AI model that it says can automatically create high-quality biographical articles about important real-world public figures, based on information found on the web. What they're saying: "There is more work to do, but we hope this new system will one day help Wikipedia editors create many thousands of accurate, compelling biography entries for important people who are currently not on the site," Fan said in a blog post. Flashback: Fan began her project as a computer science student at the Université de Lorraine in Inria, France.

Mark Zuckerberg has announced his social media empire is building what he claims is the world's fastest artificial intelligence supercomputer as part of plans to build a virtual metaverse. The Facebook founder said in a blogpost that the metaverse, a concept that blends the physical and digital world via virtual and augmented reality, will require "enormous" computing power. The AI supercomputer, dubbed AI Research SuperCluster (RSC) by Zuckerberg's Meta business, is already the fifth fastest in the world, the company said. "The experiences we're building for the metaverse require enormous compute [sic] power (quintillions of operations/second!) and RSC will enable new AI models that can learn from trillions of examples, understand hundreds of languages, and more," wrote Zuckerberg in a blogpost. Meta researchers added that they expected the RSC to become the fastest computer of its kind when it is completed in the summer.

Advances in the AI realm are constantly coming out, but they tend to be limited to a single domain: For instance, a cool new method for producing synthetic speech isn't also a way to recognize expressions on human faces. Meta (AKA Facebook) researchers are working on something a little more versatile: an AI that can learn capably on its own whether it does so in spoken, written or visual materials. The traditional way of training an AI model to correctly interpret something is to give it lots and lots (like millions) of labeled examples. A picture of a cat with the cat part labeled, a conversation with the speakers and words transcribed, etc. But that approach is no longer in vogue as researchers found that it was no longer feasible to manually create databases of the sizes needed to train next-gen AIs.