Chinese scientists have unveiled SpikingBrain1.0, the world’s first large-scale AI language model to replicate the human brain. The model reduces energy use and runs independently of Nvidia chips, departing from conventional AI architectures.
Developed by the Chinese Academy of Sciences, SpikingBrain1.0 uses spiking neural networks to activate only the required neurons for each task, rather than processing all information simultaneously.
Instead of evaluating every word in parallel, it focuses on the most recent and relevant context, enabling faster and more efficient processing. Researchers claim the model operates 25 to 100 times faster than traditional AI systems while keeping accuracy competitive.
A significant innovation is hardware independence. SpikingBrain1.0 runs on China’s MetaX chip platform, reducing reliance on Nvidia GPUs. It also requires less than 2% of the data typically needed for pre-training large language models, making it more sustainable and accessible.
SpikingBrain1.0 could power low-energy, real-time applications such as autonomous drones, wearable devices, and edge computing. The model highlights a shift toward biologically-inspired AI prioritising efficiency and adaptability over brute-force computation.
Would you like to learn more aboutAI, tech and digital diplomacy? If so, ask our Diplo chatbot!
Broadcom has secured a $10 billion agreement to supply custom AI chips, with analysts pointing to OpenAI as the likely customer.
The US semiconductor firm announced the deal alongside better-than-expected third-quarter earnings, driven by growing demand for its ASICs. It forecast a strong fourth quarter as cloud providers seek alternatives to Nvidia, whose GPUs remain costly and supply-constrained.
Chief executive Hock Tan said Broadcom is collaborating with four potential new clients on chip development, adding to existing partnerships with major players such as Google and Meta.
The company recently introduced the Tomahawk Ultra and next-generation Jericho networking chips, further strengthening its position in the AI computing sector.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
A team of international researchers has shown how training neural networks directly with light on photonic chips could make AI faster and more sustainable.
A breakthrough study, published in Nature, involved collaboration between the Politecnico di Milano, EPFL Lausanne, Stanford University, the University of Cambridge, and the Max Planck Institute.
The research highlights how physical neural networks, which use analogue circuits that exploit the laws of physics, can process information in new ways.
Photonic chips developed at the Politecnico di Milano perform mathematical operations such as addition and multiplication through light interference on silicon microchips only a few millimetres in size.
By eliminating the need to digitise information, these chips dramatically cut both processing time and energy use. Researchers have also pioneered an ‘in-situ’ training technique that enables photonic neural networks to learn tasks entirely through light signals, instead of relying on digital models.
The result is a training process that is faster, more efficient and more robust.
Such advances could lead to more powerful AI models capable of running directly on devices instead of being dependent on energy-hungry data centres.
An approach that paves the way for technologies such as autonomous vehicles, portable intelligent sensors and real-time data processing systems that are both greener and quicker.
Would you like to learn more aboutAI, tech and digital diplomacy? If so, ask our Diplo chatbot!
Apple is rolling out Memory Integrity Enforcement on the iPhone 17 line and iPhone Air, an always-on set of protections aimed at blocking memory-safety exploits used by mercenary spyware.
MIE builds on ARM’s Enhanced Memory Tagging Extension in Apple’s A19 chips, alongside secure allocators and tag-confidentiality measures.
Older devices without the new tagging hardware also receive memory-safety upgrades. Apple says new Spectre V1 leak mitigations arrive with virtually no CPU penalty.
Comparable ideas exist elsewhere, such as Windows 11’s memory integrity (HVCI) and Android’s MTE support on Pixel 8, but Apple’s approach is enabled by default across key attack surfaces. Security reporters note the move significantly complicates spyware operations.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
At a forum in Taipei, SEMI’s Terry Tsao said Taiwan’s chip design and fabrication expertise complements Japan’s materials and manufacturing equipment strengths. He noted that TSMC’s Kumamoto expansion creates opportunities for talent development in partnership with local schools.
Tsao described Taiwanese growth as a golden semiconductor era, emphasising its success closely tied to Japanese collaboration. He argued that TSMC’s achievements in Japan also represent progress for Japan’s industry.
Kazuhito Hashimoto, head of the Japan Science and Technology Agency, said joint projects with Taiwan’s National Science and Technology Council are underway. He pledged support for expanded research exchanges between the two partners.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
Meta has plans to spend at least $600 billion on US data centres and AI infrastructure by 2028. The forecast, reported by The Information, was shared by CEO Mark Zuckerberg during a dinner with President Donald Trump and other technology leaders.
Capital expenditure is set to rise sharply over the next three years. Meta projects spending of $66–72 billion in 2025, nearly 70% higher than 2024, with another significant increase expected in 2026.
The company said the surge in investment will be driven primarily by the need to expand AI computing power.
Zuckerberg confirmed that Meta aims to deploy more than one million GPUs to train its next generation of AI models.
The company is also investing heavily in talent and infrastructure as it builds a dedicated team focused on developing artificial super intelligence, a concept referring to AI systems with capabilities beyond those of humans.
The spending commitment highlights how major US technology companies are racing to secure computing capacity for AI. Meta is pledging ‘hundreds of billions of dollars’ towards expanding its data centre footprint in the years ahead.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
Quantum computers could become more efficient with a new quantum router that directs data more quickly within machines. Researchers at Stanford have built the component, which could eventually form the backbone of quantum random access memory (QRAM).
The router utilises superconducting qubits, controlled by electromagnetic pulses, to transmit information to quantum addresses. Unlike classical routers, it can encode addresses in superposition, allowing data to be stored in two places simultaneously.
In tests with three qubits, the router achieved a fidelity of around 95%. If integrated into QRAM, it could unlock new algorithms by placing information into quantum states where locations remain indeterminate.
Experts say the advance could benefit areas such as quantum machine learning and database searches. It may also support future ideas, such as quantum IP addresses, although more reliable designs with larger qubit counts are still required.
The Stanford team acknowledges the device needs refinement to reduce errors. But with further development, the quantum router could be a vital step toward practical QRAM and more powerful quantum computing applications.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
For decades, quantum mechanics has powered tools like semiconductors, GPS and fibre optics, a foundation often described as Quantum 1.0. The UN has declared 2025 the International Year of Quantum Science and Technology to mark its impact.
Researchers are now advancing Quantum 2.0, which manipulates atoms, ions and photons to exploit superposition and entanglement. Emerging tools include quantum encryption systems, distributed atomic clocks to secure networks against GPS failures, and sensing devices with unprecedented precision.
Experts warn that disruptions to satellite navigation could cost billions, but quantum clocks may keep economies and critical infrastructure synchronised. With quantum computing and AI developing in parallel, future breakthroughs could transform medicine, energy, and security.
Achieving this vision will require global collaboration across governments, academia and industry to scale up technologies, ensure supply chain resilience and secure international standards.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
Aimed at promoting the adoption of SBOM practices, the report highlights their role in improving transparency and addressing risks within the software supply chain.
By integrating SBOM generation, analysis, and sharing into existing security processes, organisations can better manage vulnerabilities and strengthen cyber resilience.
Practical risk management strategies and real-world examples outlined in the CSI support the broader Secure by Design initiative.
Authors urge a unified SBOM approach across the cybersecurity community to prevent fragmentation, lower implementation costs, and enhance long-term effectiveness.
Inconsistent or siloed adoption, they caution, could limit the sustainability and impact of SBOM as a core cybersecurity tool.
Would you like to learn more aboutAI, tech and digital diplomacy? If so, ask our Diplo chatbot!
Quantum computers may soon grow more powerful through 3D printing, with researchers building miniaturised ion traps to improve scalability and performance.
Ion traps, which confine ions and control their quantum states, play a central role in ion-based qubits. Researchers at UC Berkeley created 3D-printed traps just a few hundred microns wide, which captured ions up to ten times more efficiently than conventional versions.
The new traps also reduced waiting times, allowing ions to be usable more quickly once the system is activated. Hartmut Häffner, who led the study, said the approach could enable scaling to far larger numbers of qubits while boosting speed.
3D printing offers flexibility not possible with chip-style manufacturing, allowing for more complex shapes and designs. Team members say they are already working on new iterations, with future versions expected to integrate optical components such as miniaturised lasers.
Experts argue that this method could address the challenges of low yield, high costs, and poor reproducibility in current ion-trap manufacturing, paving the way for scalable quantum computing and applications in other fields, including mass spectrometry.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
