A new study "compared how well top AI systems and human workers did at hundreds of real work assignments," reports the Washington Post. They add that at least one example "illustrates a disconnect three years after the release of ChatGPT that has implications for the whole economy." AI can accomplish many impressive tasks involving computer code, documents or images. That has prompted predictions that human work of many kinds could soon be done by computers alone. Bentley University and Gallup found in a survey [PDF] last year that about three-quarters of Americans expect AI to reduce the number of U.S. jobs over the next decade. But economic data shows the technology largely has not replaced workers. To understand what work AI can do on its own today, researchers collected hundreds of examples of projects posted on freelancing platforms that humans had been paid to complete. They included tasks such as making 3D product animations, transcribing music, coding web video games and formatting research papers for publication. The research team then gave each task to AI systems such as OpenAI's ChatGPT, Google's Gemini and Anthropic's Claude. The best-performing AI system successfully completed only 2.5 percent of the projects, according to the research team from Scale AI, a start-up that provides data to AI developers, and the Center for AI Safety, a nonprofit that works to understand risks from AI. "Current models are not close to being able to automate real jobs in the economy," said Jason Hausenloy, one of the researchers on the Remote Labor Index study... The results, which show how AI systems fall short, challenge predictions that the technology is poised to soon replace large portions of the workforce... The AI systems failed on nearly half of the Remote Labor Index projects by producing poor-quality work, and they left more than a third incomplete. Nearly 1 in 5 had basic technical problems such as producing corrupt files, the researchers found. One test involved creating an interactive dashboard for data from the World Happiness Report, according to the article. "At first glance, the AI results look adequate. But closer examination reveals errors, such as countries inexplicably missing data, overlapping text and legends that use the wrong colors — or no colors at all." The researchers say AI systems are hobbled by a lack of memory, and are also weak on "visual" understanding.

Read more of this story at Slashdot.

This week Meta announced several new features for "Meta Ray-Ban Display" smartglasses: - A new teleprompter feature for the smart glasses (arriving in a phased rollout) - The ability to send messages on WhatsApp and Messenger by writing with your finger on any surface. (Available for those who sign up for an "early access" program). - "Pedestrian navigation" for 32 cities. ("The 28 cities we launched Meta Ray-Ban Display with, plus Denver, Las Vegas, Portland, and Salt Lake City," and with more cities coming soon.) But they also warned Meta Ray-Ban Display "is a first-of-its-kind product with extremely limited inventory," saying they're delaying international expansion of sales due to inventory constraints — and also due to "unprecedented" demand in the U.S. CNBC reports: "Since launching last fall, we've seen an overwhelming amount of interest, and as a result, product waitlists now extend well into 2026," Meta wrote in a blog post. Due to "limited" inventory, the company said it will pause plans to launch in the U.K., France, Italy and Canada early this year and concentrate on U.S. orders as it reassesses international availability... Meta is one of several technology companies moving into the smart glasses market. Alphabet announced a $150 million partnership with Warby Parker in May and ChatGPT maker OpenAI is reportedly working on AI glasses with Apple.

Read more of this story at Slashdot.

Meta has signed long-term nuclear power deals totaling more than 6 gigawatts to fuel its data centers: "one from a startup, one from a smaller energy company, and one from a larger company that already operates several nuclear reactors in the U.S," reports TechCrunch. From the report: Oklo and TerraPower, two companies developing small modular reactors (SMR), each signed agreements with Meta to build multiple reactors, while Vistra is selling capacity from its existing power plants. [...] The deals are the result of a request for proposals that Meta issued in December 2024, in which Meta sought partners that could add between 1 to 4 gigawatts of generating capacity by the early 2030s. Much of the new power will flow through the PJM interconnection, a grid which covers 13 Mid-Atlantic and Midwestern states and has become saturated with data centers. The 20-year agreement with Vistra will have the most immediate impact on Meta's energy needs. The tech company will buy a total of 2.1 gigawatts from two existing nuclear power plants, Perry and Davis-Besse in Ohio. As part of the deal, Vistra will also add capacity to those power plants and to its Beaver Valley power plant in Pennsylvania. Together, the upgrades will generate an additional 433 MW and are scheduled to come online in the early 2030s. Meta is also buying 1.2 gigawatts from young provider Oklo. Under its deal with Meta, Oklo is hoping to start supplying power to the grid as early as 2030. The SMR company went public via SPAC in 2023, and while Oklo has landed a large deal with data center operator Switch, it has struggled to get its reactor design approved by the Nuclear Regulatory Commission. If Oklo can deliver on its timeline, the new reactors would be built in Pike County, Ohio. The startup's Aurora Powerhouse reactors each produce 75 megawatts of electricity, and it will need to build more than a dozen to fulfill Meta's order. TerraPower is a startup co-founded by Bill Gates, and it is aiming to start sending electricity to Meta as early as 2032.

Read more of this story at Slashdot.

An anonymous reader quotes a report from Wired: [P]erhaps AI can, in fact, learn in a more human way -- by figuring out interesting questions to ask itself and attempting to find the right answer. A project from Tsinghua University, the Beijing Institute for General Artificial Intelligence (BIGAI), and Pennsylvania State University shows that AI can learn to reason in this way by playing with computer code. The researchers devised a system called Absolute Zero Reasoner (AZR) that first uses a large language model to generate challenging but solvable Python coding problems. It then uses the same model to solve those problems before checking its work by trying to run the code. And finally, the AZR system uses successes and failures as a signal to refine the original model, augmenting its ability to both pose better problems and solve them. The team found that their approach significantly improved the coding and reasoning skills of both 7 billion and 14 billion parameter versions of the open source language model Qwen. Impressively, the model even outperformed some models that had received human-curated data. [...] A key challenge is that for now the system only works on problems that can easily be checked, like those that involve math or coding. As the project progresses, it might be possible to use it on agentic AI tasks like browsing the web or doing office chores. This might involve having the AI model try to judge whether an agent's actions are correct. One fascinating possibility of an approach like Absolute Zero is that it could, in theory, allow models to go beyond human teaching. "Once we have that it's kind of a way to reach superintelligence," [said Zilong Zheng, a researcher at BIGAI who worked on the project].

Read more of this story at Slashdot.

Experts interviewed by NBC News warn that the rapid spread of AI-generated images and videos is accelerating an online trust breakdown, especially during fast-moving news events where context is scarce. From the report: President Donald Trump's Venezuela operation almost immediately spurred the spread of AI-generated images, old videos and altered photos across social media. On Wednesday, after an Immigration and Customs Enforcement officer fatally shot a woman in her car, many online circulated a fake, most likely AI-edited image of the scene that appears to be based on real video. Others used AI in attempts to digitally remove the mask of the ICE officer who shot her. The confusion around AI content comes as many social media platforms, which pay creators for engagement, have given users incentives to recycle old photos and videos to ramp up emotion around viral news moments. The amalgam of misinformation, experts say, is creating a heightened erosion of trust online -- especially when it mixes with authentic evidence. "As we start to worry about AI, it will likely, at least in the short term, undermine our trust default -- that is, that we believe communication until we have some reason to disbelieve," said Jeff Hancock, founding director of the Stanford Social Media Lab. "That's going to be the big challenge, is that for a while people are really going to not trust things they see in digital spaces." Though AI is the latest technology to spark concern about surging misinformation, similar trust breakdowns have cycled through history, from election misinformation in 2016 to the mass production of propaganda after the printing press was invented in the 1400s. Before AI, there was Photoshop, and before Photoshop, there were analog image manipulation techniques. Fast-moving news events are where manipulated media have the biggest effect, because they fill in for the broad lack of information, Hancock said. "In terms of just looking at an image or a video, it will essentially become impossible to detect if it's fake. I think that we're getting close to that point, if we're not already there," said Hancock. "The old sort of AI literacy ideas of 'let's just look at the number of fingers' and things like that are likely to go away." Renee Hobbs, a professor of communication studies at the University of Rhode Island, added: "If constant doubt and anxiety about what to trust is the norm, then actually, disengagement is a logical response. It's a coping mechanism. And then when people stop caring about whether something's true or not, then the danger is not just deception, but actually it's worse than that. It's the whole collapse of even being motivated to seek truth."

Read more of this story at Slashdot.