Google’s Gemini model made headlines last year when it produced pictures of racially diverse people in Nazi uniforms, among other historically incoherent imagery.

While the episode fueled culture wars around “wokeness” online, Stanford University researcher Angelina Wang and her team saw it as an example of a bigger problem with how generative AI researchers treat algorithmic bias. Many of the metrics developers use to make models more fair strive for racial colorblindness or other identical treatment, which can sometimes ignore contexts where group differences do matter.

That’s why Wang and her co-authors at Stanford’s Institute for Human-Centered AI (HAI) have developed a new benchmark suite aimed to test difference awareness in AI models, consisting of eight scenarios and 16,000 multiple choice questions.

“One-size-fits-all definitions of fairness don’t work very well—we really have to have a more contextualized understanding,” Wang told Tech Brew. “The predominant way of going about making these AI systems more fair can sometimes be misguided and lead to wrong answers.”

Keep reading here.—PK

The Mozilla Foundation might be best known as the org behind Firefox, but these days, it’s recasting the same stance it brought to the Browser Wars for a new generation-defining technology.

The nonprofit is currently shifting its strategy and rebranding to put AI front and center as it pushes for more open-source ideals in the technology—and helps define what that even means, invests in startups that align with that goal, and builds its own AI tools.

The shakeup comes as open-source AI is having a moment. Upstart Chinese lab DeepSeek shocked Silicon Valley with its purportedly ultra-efficient models released under an open-source license, prompting reconsideration of a debate that’s been brewing over the topic for years.

The Mozilla Foundation laid off around 30% of its staff last year in a restructuring of its operations, but Mozilla President Mark Surman said the org’s advocacy goals haven’t changed. “The short answer is that this restructuring hasn’t affected Mozilla’s mission at all. Our mission remains the same: pushing for a positive and transparent future that puts people first,” Surman said in an email.

We talked with Surman about how Mozilla is changing, whether advocacy efforts might look different under a new administration, and what DeepSeek means for open-source.

Keep reading here.—PK

Between the rise of AI-powered competitors and a landmark antitrust ruling, Google has faced unprecedented threats to its search dominance in recent months. And yet people are still googling, for now—the tech giant remains the No. 1 player in the search space.

As the courts are set to decide on how to break Google’s search stranglehold, a new study from the National Bureau of Economic Research takes a look at some of the factors that have sustained Google’s market power for the past couple decades.

The researchers recruited more than 2,000 people and studied their search habits for a period of two weeks, offering certain segments cash rewards to switch to Microsoft Bing or asking others to make an active choice rather than sticking with a default engine.

The authors found that users who were paid to switch to Bing had a more positive opinion of its quality after the two weeks were up, and a third even chose to continue using it. Removing Google as a default option and forcing users to choose only increased Bing’s share by 1.1%, however.

When Bing was made the default search option for a certain segment of the group, many didn’t bother to switch to Google, suggesting that a certain portion of search market share is driven by consumers not paying much attention to which search engine they use, the authors wrote.

Overall, the researchers concluded that if people had a fully accurate view of each search engine’s quality and if the inconvenience of switching between them were removed, Bing would have a 15% larger market share than it does now.

Keep reading here.—PK

Google’s new AI Gems promise personalized assistants for any task, but are they actually worth it? We put them to the test.