57 articles
AI agents perform better with human-curated skills, gaining up to 16% in task completion. A study found self-generated skills by AI are ineffective or harmful, showing human expertise is still essential for guiding AI.
Meta funds US politicians to protect AI interests amid datacenter expansion concerns and regulatory pushback. Other tech giants also invest heavily in AI-friendly political campaigns.
AI's climate benefits are overstated, with most claims citing traditional AI, not energy-hungry generative AI. Evidence is weak, often corporate-backed, and may greenwash AI's true fossil fuel-driven emissions impact.
Google released Gemini 3.1 Pro, a new AI model it claims excels at complex reasoning, citing benchmark scores that often surpass rivals like OpenAI and Anthropic. It's available now across various platforms.
Microsoft proposes a blueprint to verify online content authenticity using digital provenance, watermarks, and fingerprints. While promising, adoption faces hurdles from tech platforms and public skepticism, even as regulations like California's AI Transparency Act push for change.
Reddit is testing an AI search tool that shows product carousels based on user recommendations, linking directly to retailers. This blends community discussions with e-commerce.
Google Chrome adds new features like Split View, PDF annotations, and direct saving to Google Drive to compete with AI browsers and keep users.
AI may replace some tasks, but experts argue humans will remain essential for decision-making, with AI freeing them to focus on higher-value work.
AI summaries can be dangerously biased, especially in multilingual contexts. Hidden instructions can steer outputs to hide human rights abuses or give unsafe advice, as shown in refugee aid tests. Evaluation must lead to better safeguards.
Fomi is an AI-powered Mac app that blocks distractions by analyzing your screen to tell if you're working or procrastinating. It uses cloud AI, raising privacy concerns, but redacts personal data first.
AI-generated PRs are flooding open source, overwhelming maintainers and degrading code quality. A "guild system" (web of trust) is proposed to verify contributors and filter out bot-generated spam.
Gentoo is migrating from GitHub to Codeberg, citing "continuous attempts to force Copilot usage" and concerns over AI-generated code quality and ethics.