Singapore AI News & Daily Briefing

Singaporean-Malaysian comedian Ronny Chieng told Harvard graduates their generation's mission is to 'destroy AI,' to loud cheers. In his Class Day keynote, he argued AI robs people of the satisfaction of creative work, saying 'the creating is the fun part.' He acknowledged AI's potential for medicine and physics breakthroughs but railed against using it for mundane tasks like drafting speeches or podcasts. 'Mastery versus faking it' will define his generation, he said.

Tech billionaires like Jeff Bezos, Sam Altman, and Elon Musk are proposing various solutions to wealth inequality they helped create through the AI boom. Proposals range from Bezos suggesting the bottom 50% pay zero income tax, to Altman's 'universal basic compute' idea of giving people AI access instead of cash, to a 5% California billionaire wealth tax on the ballot. Anthropic CEO Dario Amodei has warned fellow billionaires that if they don't support good tax policies, they'll 'get a bad version designed by a mob.' The AI revolution could wipe out millions of jobs while minting the world's first trillionaires, and politicians are sensing voter anger.

A franchise owner operating 110 Pizza Hut locations is suing the chain for $100 million, claiming a botched AI tool implementation destroyed their business. This case highlights how rushed AI adoption in everyday industries can backfire badly — costing jobs and livelihoods when the technology doesn't work as promised.

Nvidia is teasing its first Arm-powered laptop chips, with Microsoft and Arm joining the hype ahead of Computex. All three companies posted "A new era of PC" on social media, hinting at processors that could rival Apple Silicon. If real, these chips could bring serious AI performance to Windows laptops and challenge Qualcomm's dominance in ARM-based PCs.

An editor shares how to spot AI-generated writing: it tends to be perfectly grammatical, evenly paced, and oddly uniform — lacking the quirky imperfections of real human prose. The author recounts a car crash where the at-fault driver sent a grammatically elegant AI-written explanation, and a mechanic responded in the same 'distinctive voice of AI.' The piece explores how AI text is now creeping into books, newspapers, and literary magazines.

Microsoft and NVIDIA are teasing what experts believe is the fabled N1X chip — a major new processor that could redefine PC performance. The announcement, described as 'a new era of PC,' suggests significant hardware improvements that could benefit everything from gaming to AI workloads.

Pope Leo XIV's new encyclical warns humanity must safeguard itself from AI's most disruptive effects. Anthropic advised on the document. The SF Standard explores what this means for Silicon Valley—essentially a top-down moral framework calling for human dignity over AI efficiency, and restrictions on AI in warfare and decision-making affecting lives.

Anthropic has overtaken OpenAI to become the world's most valuable AI startup. The shift reflects intensifying competition between major AI labs and growing investor confidence in Anthropic's approach to safe, interpretable AI systems. Singapore's sovereign wealth funds GIC and Temasek recently anchored a significant investment in Anthropic's latest funding round.

Silicon Motion launched the SM2524XT, a PCIe Gen5 SSD controller built for AI inference tasks. It delivers up to 14GB/s read and 12GB/s write speeds with 2.5 million IOPS, while consuming under 5W. The controller is designed to handle the random, fragmented data patterns that AI inference generates — think of it as a storage upgrade specifically engineered for running local AI models and agents without choking on data.

SoftBank plans to invest up to €75 billion to build massive data centers in France, adding 5 gigawatts of capacity to support AI infrastructure. This shows the scale of global investment needed for AI, and it could affect energy markets and job growth in data center hubs.

GitHub Copilot is switching up how it charges developers—moving to a token-based billing system—and programmers are not happy about it. The change means users now pay based on how many "tokens" (chunks of text the AI processes) they consume, which can make costs unpredictable. Many developers are venting frustration, calling the new model a step backward from the simpler flat-rate subscription they had before.

Researchers at the University of Washington planned to have preschool teachers wear body cameras recording everything from their first-person perspective — including the children they teach — to use as training data for AI models. Parents were given consent forms describing the recording setup. One parent understood it as opt-out rather than opt-in. The university says participation required parental permission for all children. This raises serious privacy concerns about using footage of young children to train AI systems.

Traffic to DuckDuckGo's 'No AI' search page has tripled after Google's latest AI search update, indicating growing user demand for AI-free search. This suggests many people are tired of AI-generated summaries and want simpler, text-based results.

Scientists have developed a new device that could make processors 1,000 times faster without generating extra heat, potentially slashing data center energy demands. If scaled, this could dramatically reduce the environmental cost of AI and make electronic devices much more efficient.

University of Illinois researchers have demonstrated a new method for stacking multiple layers of silicon electronics directly on top of one another using ultrathin nanomembranes. The breakthrough achieves 98-100% device yields while staying below 400°C, solving the thermal problem that has blocked true monolithic 3D chip integration for years. This could dramatically increase computing density and bandwidth for AI chips while reducing energy consumption, potentially extending Moore's Law. The work was published in Nature with industry partners including IBM, Intel, and TSMC.

Rivian's software chief wants to replace physical buttons and phone mirroring (CarPlay/Android Auto) with AI voice agents you talk to constantly. His vision: tell your car 'I have a bag in front of the car' to open the trunk, or have the car read your emails while driving. Critics say this is solving a problem that doesn't exist — your phone already does all this. The article argues that adding unnecessary voice commands to simple tasks (like opening a trunk) makes things worse, not better, and locks car owners into proprietary AI systems that could be discontinued or turned into subscription services.

Pope Leo XIV has called on Europe to ban lethal autonomous weapons, warning that outsourcing life-and-death decisions to algorithms is morally unacceptable. The EU currently excludes military AI from its main AI Act, creating a gap that could erode ethical standards.

Nvidia researchers demonstrated that robots trained entirely in computer simulations can successfully perform real-world tasks without additional physical training. This could dramatically speed up and cheapen robot development — instead of spending months programming robots through trial-and-error in factories, developers can simulate millions of scenarios virtually. Practical robots at home or work could arrive sooner if this approach scales.

SoftBank plans to invest up to €75 billion to build massive data centers in France, adding 5 gigawatts of capacity to support AI infrastructure. This shows the scale of global investment needed for AI, and it could affect energy markets and job growth in data center hubs.