Google's AI-focused DeepMind division has taken a minority stake in the developer of popular sci-fi simulation EVE Online , saying it will use the game to study "intelligence in complex, dynamic, player-driven systems."

The research partnership comes as the management behind EVE Onlin e developer CCP Games announced that they have spent $120 million to buy themselves out from their former owners at South Korean publisher Pearl Abyss (Crimson Desert ) . The newly independent entity is being rebranded as Fenris Creations, which will continue to operate as normal without any restructuring or layoffs, the company said.

"Something that already behaves like a living world"

In today's announcement, Fenris and DeepMind said that EVE Online presents "a uniquely rich environment for study," especially when it comes to developing AI systems that use "long-horizon planning, memory, and continual learning." DeepMind says it will conduct controlled experiments on its models in a specially designed offline version of the game running on a local server, without directly impacting the experience for online players. The two companies "will also explore new gameplay experiences enabled by these technologies," they wrote.

Read full article

Comments

The astronauts flying aboard the Artemis II mission to the Moon last month took a lot of pictures, and a few dozen of the best ones were released during and shortly afterward the flight.

But it wasn't until last weekend that NASA released the whole trove of more than 12,000 images, dumping them onto the Gateway to Astronaut Photography . The astronauts used three different cameras on the mission: a Nikon D5, a Nikon Z9, and an iPhone 17s. There are some hits and misses in the archive, plus some new gems.

One of the early highlights during the mission was the "Hello, world" image captured by Artemis II Commander Reid Wiseman as the Orion spacecraft left Earth on its outbound journey toward the Moon.

Read full article

Comments

Despite Health Secretary Robert F. Kennedy's pledge to provide "radical transparency," the agencies under his control continue to suppress scientific research that conflicts with his anti-vaccine agenda.

On Tuesday, The New York Times reported confirmation from the Department of Health and Human Services that the Food and Drug Administration had blocked the publication of studies showing the safety and efficacy of vaccines against COVID-19 and shingles . The revelation follows a report from The Washington Post last month that the Centers for Disease Control and Prevention scrapped a scientifically vetted study previously scheduled for publication that found COVID-19 vaccines sharply cut the risk of emergency care and hospitalization among healthy adults. The study was ultimately rejected by Kennedy's acting CDC director, who claimed to have concerns about the study's methodology.

Similarly at the FDA, two studies on COVID-19 vaccines by agency scientists were accepted for publication at medical journals, according to the Times. But unnamed FDA officials directed the agency scientists to withdraw the studies. While a preliminary abstract of one of the studies presented at a conference last fall remains online, the Times obtained a copy of the full manuscript, the conclusion of which reads, "Given the available evidence, FDA continues to conclude the benefits of vaccination outweigh the risks."

Read full article

Comments

SAN FRANCISCO—At its Code with Claude developers' conference, Anthropic has introduced what it calls "dreaming" to Claude Managed Agents. Dreaming, in this case, is a process of going over recent events and identifying specific things that are worth storing in "memory" to inform future tasks and interactions.

Dreaming is a feature that is currently in research preview and limited to Managed Agents on the Claude Platform. Managed Agents are a higher-level alternative to building directly on the Messages API that Anthropic describes as a "pre-built, configurable agent harness that runs in managed infrastructure." It's intended for situations where you want multiple agents working on a task or project to some end point over several minutes or hours.

Anthropic describes dreaming as a scheduled process, in which sessions and memory stores are reviewed, and specific memories are curated. This is important because context windows are limited for LLMs, and important information can be lost over lengthy projects. On the chat side of things, many models use a process called compaction, whereby lengthy conversations are periodically analyzed, and the models attempt to remove irrelevant information from the context window while keeping what's actually important for the ongoing conversation, project, or task.

Read full article

Comments

Google launched its Gemma 4 open models this spring, promising a new level of power and performance for local AI. Google's take on edge AI could be getting even faster already with the release of Multi-Token Prediction (MTP) drafters for Gemma. Google says these experimental models leverage a form of speculative decoding to take a guess at future tokens, which can speed up generation compared to the way models generate tokens on their own.

The latest Gemma models are built on the same underlying technology that powers Google's frontier Gemini AI, but they're tuned to run locally. Gemini is optimized to run on Google's custom TPU chips , which operate in enormous clusters with super-fast interconnects and memory. A single high-power AI accelerator can run the largest Gemma 4 model at full precision, and quantizing will let it run on a consumer GPU.

Gemma allows users to tinker with AI on their hardware rather than sharing all their data with a cloud AI system from Google or someone else. Google also changed the license for Gemma 4 to Apache 2.0, which is much more permissive than the custom Gemma license Google employed for previous releases. However, there are inherent limitations in the hardware most people have to run local AI models. That's where MTP comes in.

Read full article

Comments

NASA's goal of reaching the Moon's surface as many as 21 times over the next two and a half years will require an overhaul of the agency's approach to buying lunar landers and success in rectifying the myriad problems that have, so far, caused three of the last four US landing attempts to falter.

It will also require improved oversight of NASA's industrial base and better management of a supply chain that has often failed to deliver on time.

These landers are separate from NASA's Human Landing System program , which has contracts with SpaceX and Blue Origin to develop and deliver human-rated landers to ferry crews to and from the lunar surface for the agency's Artemis program. Alongside the crew landers, dozens of robotic and cargo landings will deliver payloads to scout for a future Moon base and demonstrate technologies for larger vehicles, mining and resource utilization, and sustained operations during the two-week-long lunar night.

Read full article

Comments

They entered the world the way babies should, with piercing cries announcing their arrival. They passed their newborn screening tests. Some made it to their 2-week wellness visits without concern.

Then, without warning, their systems began to shut down. A 7-week-old boy in Maryland developed sudden seizures. An 11-pound girl in Alabama stopped breathing for 20 seconds at a time. A baby boy in Kentucky vomited before becoming lethargic. A brown-haired girl in Texas, not yet 2 weeks old, bled around her belly button.

Desperate to save them, records show, doctors inserted tubes into their airways and hooked them up to IVs. They ordered blood transfusions. They spent half an hour trying to resuscitate one boy until his parents told them they could stop. They shaved another boy’s soft locks to embed a needle directly into his skull to reduce the pressure in his brain.

Read full article

Comments

I spend more time today than ever before interacting with terminal windows, which is something I don't think Past Me would have believed in the early '90s. Back then, poor MS-DOS was the staid whipping boy of the industry, and at least on the consumer side, graphical environments like Windows (and maybe even odder creatures like AmigaOS ) seemed poised to stamp the command line into oblivion, leaving text interfaces behind as we all blasted into the ooey-GUI future.

As it turns out, though, the command line is still the best tool for some jobs—many jobs, in fact. I read a wise post some years ago (probably on Slashdot) arguing that a mouse-driven point-and-click interface essentially reduces the user to pointing at something on the screen and grunting, "DO! DO THAT!" at the computer. (The rise of right-click context menus adds the ability for the user to also grunt "MORE THINGS!" but doesn't otherwise add vocabulary.)

The command line, by contrast, gives the user the opportunity to precisely tell the computer what they want done, using words instead of one or two gestalts that the computer must interpret based on context.

Read full article

Comments

We don't have too much longer to wait for Rivian's hotly anticipated R2 electric SUV. After cutting its teeth with the ground-breaking R1T electric pickup truck, plus the three-row SUV version (the R1S) and all those Amazon delivery vans, its next step is something smaller and more affordable—launch models of the midsize R2 are competitively priced against rivals like the BMW iX3 , and next year if all goes to plan, Rivian will add a $45,000 R2 with a smaller battery.

But according to Rivian CEO RJ Scaringe, who was interviewed by Reuters , there's more R2 in the works. "So clearly there could be an R2X," Scaringe told Reuters. "There's ⁠going to be combinations... I want to be careful not to announce the program," he said. Scaringe also told the news organization that Rivian was considering making its own lidar sensors, in collaboration with a Chinese company.

Initial R2s are being built at Rivian's factory in Normal, Illinois. But the R2, together with an even smaller R3 (and R3X) that will follow it, will also be built at Rivian's new factory in Georgia. That plant is due to come online in 2028, funded in part by a $4.5 billion loan from the Department of Energy, which Rivian will begin to access next year.

Read full article

Comments