China Builds a Supercomputer Network Based in Space
In an audacious move, China has launched 12 AI-powered satellites as the first component of its planned Three-Body Computing Constellation. Eventually, this constellation will be 2,800 interconnected satellites, forming what Chinese scientists call the world’s first orbital supercomputer network.
Each of these 12 satellites carries an 8-billion-parameter AI model and boasts about 744 tera operations per second (TOPS). Together, they already deliver about 5 peta operations per second (POPS), with plans to scale up to 1,000 POPS. The goal is to process data in orbit (rather than downlink large datasets to Earth), reducing latency, bypassing bandwidth constraints, and harnessing natural cooling in space.
Why this matters:
Edge computing taken literally: Instead of the “edge” being at a nearby data center or cell tower, China is bringing computation into space. For applications like disaster response, earth observation, climate monitoring, or military/intelligence use, being able to analyze in orbit could shave off critical transmission delays.
Strategic competition: Deploying this constellation gives China technological leverage—both in civilian domains (weather prediction, mapping, scientific research) and possibly in military or surveillance uses.
Energy and environmental implications: Because the satellites use solar power and dissipate heat in the cold of space, there may be efficiencies. But the environmental cost of launches, orbital debris, and maintenance still loom.
Elton John Slams UK Government’s AI Copyright Proposals
Back in the UK, there’s been a wave of outcry from the arts world. Sir Elton John, Sir Paul McCartney, Dua Lipa and more than 400 British artists have urged Prime Minister Keir Starmer to protect creators’ works in the age of AI. Their concern: proposed copyright reforms could allow AI firms to use artists’ content for training without proper compensation or permission. The Times of India+2BBC+2
Elton John denounced the government as “absolute losers” for planning to exempt tech companies—or letting them ride roughshod over existing copyrights—if the changes go ahead. He said these proposals could be tantamount to theft of creative work. Artists are pushing for an opt-out model (instead of opt-in), where content creators must explicitly allow their work to be used for AI training, rather than having to opt out later. The Guardian+2BBC+2
NVIDIA Presents NVLink Fusion for AI Infrastructure
NVIDIA has launched a new technology called NVLink Fusion aimed at helping industries build semi-custom AI infrastructure. It’s part of their NVLink ecosystem, which already connects GPUs and CPUs tightly to accelerate model training and inference. Partners like MediaTek, Marvell, Fujitsu, Qualcomm, Synopsys, and others will be able to use NVLink Fusion to design custom silicon that is tightly integrated with NVIDIA’s GPUs — optimizing for workloads like model training or agentic AI inference. NVIDIA Newsroom
Key features and significance:
Custom CPUs integrated with NVIDIA GPUs to reduce latency and increase throughput.
Better scale-up (bigger workloads on one machine) and scale-out (systems working together) capabilities via NVLink, which already is known for low-latency, high-bandwidth connections between chips.
Allows companies to tailor hardware to their use cases (e.g. generative AI, autonomous agents) rather than always using off-the-shelf components.
Implications:
Could shift some of the hardware differentiation toward more vertically integrated AI players: those who not only use AI models but build custom chips around them.
Accelerates the trend of AI infrastructure becoming more specialized, reducing waste, increasing performance—but possibly raising costs for smaller players.
Adds another piece to the puzzle of who controls AI infrastructure, since controlling both silicon + model + deployment confers considerable power.
Microsoft Imagines Cooperative Artificial Intelligence Agents
Microsoft is pushing forward in imagining a future of cooperative AI agents, i.e. agents from different companies or ecosystems talking to each other, sharing tasks, and retaining memory of past interactions. They’re backing standards like the Model Context Protocol (MCP) (backed by Anthropic), which aims to let different agents interoperate — much like early internet protocols allowed browsers from any company to access websites. Reuters+1
One specific technical concern Microsoft is addressing is memory. Current AI agents are often transactional (they take a request, they do it, they forget it). Microsoft wants better memory by using so-called structured retrieval augmentation, where agents summarize key information from conversations and store it for future reference. Reuters
What this means in practice:
More helpful and consistent AI assistants: If you tell your agent something today, it might remember relevant details later (e.g. preferences, tasks in progress).
Cross-company agent collaboration: Imagine an AI agent from one company helping or coordinating with an agent from another, if privacy and security allow.
Potential for misuse: Agents with memory raise issues of privacy, data retention, and surveillance. Who controls what’s remembered, how, and who has access?
Klarna, a Swedish Fintech, Reverses AI-Driven Layoffs
AI hasn’t always gotten it right at the operational level. The Swedish “buy now, pay later” firm Klarna provides a cautionary tale. A few years ago, Klarna replaced roughly 700 customer service employees with AI bots, and planned deeper cuts as “efficiency gains” were expected from automation. UKTN+1
However, over time Klarna found that “customer service AI agents” could not fully deliver in terms of quality, reliability, or customer satisfaction. As issues accumulated, the company began to rehire human workers to restore service standards. The reversal underlines the limits of automation, especially in tasks with human interaction and nuance. The Economic Times
UK Government Launches AI Tool for Public Consultations
Across the pond, the UK is beginning to put AI to work in governance. The government has developed a tool called “Consult”, part of a larger suite dubbed “Humphrey”, to help staff analyze public consultation responses more efficiently. The Guardian+1
One of the first trials was with the Scottish Government, reviewing responses (over 2,000 submissions) about how non-surgical cosmetic procedures should be regulated. The AI tool was able to pick out key themes, sort responses into categories, and produce analysis that matched human expert reviews. Using “Consult” could save the government £20 million annually and 75,000 staff days in analysis time across all its consultations. GOV.UK+1
Pros and risks:
Pros: Faster turnaround for policy feedback, less reliance on contractors or outsourcing, more consistent thematic extraction. Could free civil servants for higher-level analysis.
Risks: Bias in what gets surfaced, possible manipulation of public responses, or over-simplification of nuanced feedback. Also concerns about transparency: how much does the public know what the AI is doing?
Trump’s AI Agreements with Saudi Arabia and UAE Generate Controversy
During a recent Middle East tour, former US President Donald Trump mediated a set of agreements or deals with Saudi Arabia and the UAE tied to artificial intelligence — particularly around exporting advanced AI chips, building AI infrastructure, and establishing campuses. TechSpot+3The Guardian+3www.ndtv.com+3
Highlights:
UAE is to build the largest AI campus outside the US.
Saudi Arabia’s sovereign wealth fund–backed AI firm “Humain” is to receive tens of thousands of advanced AI chips.
The US has committed to easing export restrictions or enabling large shipments of Nvidia / AMD chips under certain conditions. Breitbart+2The Guardian+2
Controversies & criticism:
Security concerns: There’s alarm among “China hawks” in US politics that these deals may enable indirect access by China to restricted AI technologies. The concern is that once chips are in the Gulf, supply chains or partnerships could lead to repatriation or use by parties allied or with economic exchange with China. www.ndtv.com+1
Safeguards not always clear: The agreements reportedly include high-level language barring Chinese firms, but critics say legally binding, enforceable terms are vague.
Strategic balancing: For the UAE and Saudi Arabia, these deals represent an opportunity to accelerate their ambitions as regional/global AI hubs. For the US (or US companies), they represent economic opportunity but also diplomatic risk.
OpenAI CEO Sam Altman publicly defended the agreements, calling the critics “naïve,” arguing that these moves shift the global technological balance advantageously. TechSpot
Final thoughts: Crossing the AI Frontier
The events on May 18, 2025, highlight the fast evolution and worldwide influence of artificial intelligence. AI keeps changing businesses and society from space-based computing networks to policy discussions and corporate strategies. Balancing innovation with ethical issues is top priority for countries and businesses negotiating this frontier.