AI Beats Humans in Turing Test, Microsoft Embeds Claude in Excel, China Launches Wind-Powered Underwater Data Center, Apple Exposes Siri’s Limits Live

**Carly Wilson** (0:05)
Welcome to Digimasters Shorts.

**Adam Nagus** (0:07)
We are your hosts, Adam Nagus and Carly Wilson, delivering the latest scoop from the digital realm.
Two advanced artificial intelligence systems have passed the Turing test, a benchmark designed to determine whether a machine can exhibit intelligence indistinguishable from a human. In the test, participants engage in text-based conversations with both a human and a machine without knowing which is which. If they cannot reliably tell the difference, the machine is considered to have passed. Researchers evaluated four AI models and found that newer large language models can convincingly imitate human behaviour in short exchanges. GPT 4.5 was judged to be human 73 percent of the time, even more often than the actual human participant. Llama 3.1-405b was deemed human 56 percent of the time, making it statistically indistinguishable from real people in the study. Older systems performed far worse, with early Chatbot Eliza and GPT 4 rarely mistaken for humans. Experts say the results show that given the right prompts, advanced models can replicate human tone, humour and even fallibility. However, researchers warn this capability raises concerns about deception and the potential rise of counterfeit digital identities.
While the systems required specific instructions to achieve these results, the findings suggest people may need to be far less certain that a conversation partner online is truly human.

**Carly Wilson** (1:27)
Microsoft's Build 2026 Conference delivered a major enterprise AI move, with the announcement that Anthropix Claude is now available in Excel Agent mode. The integration places a frontier AI model directly inside the spreadsheet software used by an estimated 750 million people worldwide. Users can ask Claude to write and explain formulas, clean and transform data, generate analysis and build automated workflows without leaving Excel. The update dramatically shortens the path between raw data and a meeting-ready explanation. The move is part of a broader expansion of Microsoft's Foundry platform, which now hosts more than 11,000 AI models accessible through a single Azure endpoint and billing system. Claude also extends into Microsoft 365 Copilot and Custom Agents, deepening its reach across enterprise tools. While the functionality is immediate and practical, it comes with usage-based pricing tied to token consumption. That means the more teams rely on the AI, the more they pay. By embedding Claude into Excel, Microsoft brings advanced AI to one of the most widely used work surfaces in the world, making it one of the most consequential deployments of AI in everyday office work this year.
The world's first wind-powered underwater data center has begun operations off the coast of Shanghai, marking a significant step in China's push to meet the soaring energy demands of artificial intelligence. The Shanghai Linggang Facility, launched in May, has a capacity of 24 megawatts and sits 10 meters below the surface, powered by a nearby offshore wind farm. Developed by High Cloud Technology and state-owned China Communications Construction, the project is located more than six miles from shore. Officials say the submerged design cuts power consumption by more than 20% compared with land-based data centers, largely due to the natural cooling effect of surrounding seawater. Traditional facilities can use up to 40% of their electricity simply to prevent servers from overheating while also consuming vast amounts of fresh water. A recent United Nations report warned that data centers global water use could reach 9.3 trillion liters a year by 2030
High Cloud previously launched a commercial underwater data center in Hainan in 2023, but the Shanghai project is the first to run on offshore wind power.
Experts note that while Microsoft tested a similar concept in Scotland in 2018, China has moved faster toward full scale commercial deployment. The 1.6 billion yuan investment reflects Beijing's broader strategy to expand AI infrastructure and significantly increase clean energy supplies by 2030 Marine scientists caution that underwater data centers could disturb ecosystems, though early assessments suggest environmental impacts are likely to be limited and manageable with monitoring.

**Adam Nagus** (4:08)
At Apple's Worldwide Developers Conference in Cupertino, one of the most striking moments came when the company chose not to edit out a delay.
During the keynote, presenters asked the new Siri to find an address in old messages, pull photos from a ranch trip, and compare three PDF quotes and the audience simply watched as the assistant worked. In past years, such pauses would have been trimmed to keep the presentation seamless. This time, Apple left the silence in. The decision stood in sharp contrast to 2024, when Apple unveiled a vision of a deeply personal, context-aware Siri that could reason across apps and data. Those earlier demos promised an assistant that could read posters, track flights from messages, and seamlessly manage daily life.
Most of those ambitious features never shipped, raising questions about how much of the presentation reflected reality. While Apple did introduce tools like Genmoji and enhanced writing features, they fell short of the broader personal intelligence it had promoted. At this year's event, presenters repeatedly emphasized that Siri was actively searching, analyzing and processing information in real time. By allowing the audience to experience the weight, Apple appeared to signal a renewed focus on authenticity over polished illusion.