Skip to main content

Scale AI says 'tanks a lot' to Pentagon for data-classifying deal

1 month 2 weeks ago
First up: $41M to use human annotators to label all that unstructured military data. What could go wrong?

Data curation firm Scale AI has partnered with the Pentagon to deploy its AI on Top Secret networks - a move its interim CEO says is necessary if the US wants AI to be useful for national security.…

Brandon Vigliarolo

Flying Cars Crash Into Each Other At Air Show In China

1 month 2 weeks ago
Two Xpeng AeroHT flying cars collided during a rehearsal for the Changchun Air Show in China, with one vehicle catching fire upon landing. While the company reported no serious injuries, CNN reported one person was injured in the crash. The BBC reports: Footage on Chinese social media site Weibo appeared to show a flaming vehicle on the ground which was being attended to by fire engines. One vehicle "sustained fuselage damage and caught fire upon landing," Xpeng AeroHT said in a statement to CNN. "All personnel at the scene are safe, and local authorities have completed on-site emergency measures in an orderly manner," it added. The electric flying cars take off and land vertically, and the company is hoping to sell them for around $300,000 each. In January, Xpeng claimed to have around 3,000 orders for the vehicle. [...] It has said it wants to lead the world in the "low-altitude economy."

Read more of this story at Slashdot.

BeauHD

AMD tries to catch CUDA with performance-boosting ROCm 7 software

1 month 2 weeks ago
House of Zen promises 3.5x improvement in inference and 3x uplift in training perf over last-gen software

AMD closed the performance gap with Nvidia's Blackwell accelerators with the launch of the MI355X this spring. Now the company just needs to overcome Nvidia's CUDA software advantage and make that perf more accessible to developers. …

Tobias Mann

Microsoft Favors Anthropic Over OpenAI For Visual Studio Code

1 month 2 weeks ago
Microsoft is now prioritizing Anthropic's Claude 4 over OpenAI's GPT-5 in Visual Studio Code's auto model feature, signaling a quiet but clear shift in preference. The Verge reports: "Based on internal benchmarks, Claude Sonnet 4 is our recommended model for GitHub Copilot," said Julia Liuson, head of Microsoft's developer division, in an internal email in June. While that guidance was issued ahead of the GPT-5 release, I understand Microsoft's model guidance hasn't changed. Microsoft is also making "significant investments" in training its own AI models. "We're also going to be making significant investments in our own cluster. So today, MAI-1-preview was only trained on 15,000 H100s, a tiny cluster in the grand scheme of things," said Microsoft AI chief Mustafa Suleyman, in an employee-only town hall last week. Microsoft is also reportedly planning to use Anthropic's AI models for some features in its Microsoft 365 apps soon. The Information reports that the Microsoft 365 Copilot will be "partly powered by Anthropic models," after Microsoft found that some of these models outperformed OpenAI in Excel and PowerPoint.

Read more of this story at Slashdot.

BeauHD

Gemini AI Solves Coding Problem That Stumped 139 Human Teams At ICPC World Finals

1 month 2 weeks ago
An anonymous reader quotes a report from Ars Technica: Like the rest of its Big Tech cadre, Google has spent lavishly on developing generative AI models. Google's AI can clean up your text messages and summarize the web, but the company is constantly looking to prove that its generative AI has true intelligence. The International Collegiate Programming Contest (ICPC) helps make the point. Google says Gemini 2.5 participated in the 2025 ICPC World Finals, turning in a gold medal performance. According to Google this marks "a significant step on our path toward artificial general intelligence." Every year, thousands of college-level coders participate in the ICPC event, facing a dozen deviously complex coding and algorithmic puzzles over five grueling hours. This is the largest and longest-running competition of its type. To compete in the ICPC, Google connected Gemini 2.5 Deep Think to a remote online environment approved by the ICPC. The human competitors were given a head start of 10 minutes before Gemini began "thinking." According to Google, it did not create a freshly trained model for the ICPC like it did for the similar International Mathematical Olympiad (IMO) earlier this year. The Gemini 2.5 AI that participated in the ICPC is the same general model that we see in other Gemini applications. However, it was "enhanced" to churn through thinking tokens for the five-hour duration of the competition in search of solutions. At the end of the time limit, Gemini managed to get correct answers for 10 of the 12 problems, which earned it a gold medal. Only four of 139 human teams managed the same feat. "The ICPC has always been about setting the highest standards in problem-solving," said ICPC director Bill Poucher. "Gemini successfully joining this arena, and achieving gold-level results, marks a key moment in defining the AI tools and academic standards needed for the next generation." Gemini's solutions are available on GitHub.

Read more of this story at Slashdot.

BeauHD