AI is moving from lab demos to frontline eye care, with clinicians using algorithms alongside routine fundus photos to spot disease before symptoms appear. The aim is simple: catch diabetic retinopathy early enough to prevent avoidable vision loss and speed referrals for treatment.
New imaging workflows pair adaptive optics with machine learning to shrink scan times from hours to minutes while preserving single-cell detail. At the US National Eye Institute, models recover retinal pigment epithelium features and clean noisy OCT data to make standard scans more informative.
Duke University’s open-source DCAOSLO goes further by combining multiplexed light signals with AI to capture cellular-scale images quickly. The approach eases patient strain and raises the odds of getting diagnostic-quality data in busy clinics.
Clinic-ready diagnostics are already changing triage. LumineticsCore, the first FDA-cleared AI to detect more-than-mild diabetic retinopathy from primary-care images, flags who needs urgent referral in seconds, enabling earlier laser or pharmacologic therapy.
Researchers also see the retina as a window on wider health, linking vascular and choroidal biomarkers to diabetes, hypertension and cardiovascular risk. Standardised AI tools promise more reproducible reads, support for trials and, ultimately, home-based monitoring that extends specialist insight beyond the clinic.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
InfEHR transforms time-ordered data, visits, labs, medications, and vital signs, into a graphical network for each patient. It then learns which combinations of clues across that network tend to correlate with hidden disease states.
In testing, with only a few physician-annotated examples, the AI system identified neonatal sepsis without positive blood cultures at rates 12–16× higher than current methods, and post-operative kidney injury with 4–7× more sensitivity than baseline clinical rules.
As a safety feature, InfEHR can also respond ‘not sure’ when the record lacks enough signal, reducing the risk of overconfident errors.
Because it adapts its reasoning per patient rather than applying the same rules to all, InfEHR shows promise for personalized diagnostics across hospitals and populations, even with relatively small annotated datasets.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
The UAE Ministry of Investment and Microsoft signed a Memorandum of Understanding at GITEX Global 2025 to apply AI to investment analytics, financial forecasting, and retail optimisation. The deal aims to strengthen data governance across the investment ecosystem.
Under the MoU, Microsoft will support upskilling through its AI National Skilling Initiative, targeting 100,000 government employees. Training will focus on practical adoption, responsible use, and measurable outcomes, in line with the UAE’s National AI Strategy 2031.
Both parties will promote best practices in data management using Azure services such as Data Catalog and Purview. Workshops and knowledge-sharing sessions with local experts will standardise governance. Strong controls are positioned as the foundation for trustworthy AI at scale.
The agreement was signed by His Excellency Mohammad Alhawi and Amr Kamel. Officials say the collaboration will embed AI agents into workflows while maintaining compliance. Investment teams are expected to gain real-time insights and automation that shorten the time to action.
The partnership supports the ambition to make the UAE a leader in AI-enabled investment. It also signals deeper public–private collaboration on sovereign capabilities. With skills, standards, and use cases in place, the ministry aims to attract capital and accelerate diversification.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
Salesforce and AWS outlined a tighter partnership on agentic AI, citing rapid growth in enterprise agents and usage. They set four pillars for the ‘Agentic Enterprise’: unified data, interoperable agents, modernised contact centres and streamlined procurement via AWS Marketplace.
Data 360 ‘Zero Copy’ accesses Amazon Redshift without duplication, while Data 360 Clean Rooms integrate with AWS Clean Rooms for privacy-preserving collaboration. 1-800Accountant reports agents resolving most routine inquiries so human experts focus on higher-value work.
Agentforce supports open standards such as Model Context Protocol and Agent2Agent to coordinate multi-vendor agents. Pilots link Bedrock-based agents and Slack integrations that surface Quick Suite tools, with Anthropic and Amazon Nova models available inside Salesforce’s trust boundary.
Contact centres extend agentic workflows through Salesforce Contact Center with Amazon Connect, adding voice self-service plus real-time transcription and sentiment. Complex issues hand off to representatives with full context, and Toyota Motor North America plans automation for service tasks.
Procurement scales via AWS Marketplace, where Salesforce surpassed $2bn in lifetime sales across 30 countries. AgentExchange listings provide prebuilt, customisable agents and workflows, helping enterprises adopt agentic AI faster with governance and security intact.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
Most firms are still struggling to turn AI pilots into measurable value, Cisco’s 2025 AI Readiness Index finds. Only 13% are ‘AI-ready’, having scaled deployments with results. The rest face gaps in data, security and governance.
Southeast Asia outperforms the global average at 16% ready. Indonesia reaches 23% and Thailand 21%, ahead of Europe at 11% and the Americas at 14%. Cisco says lower tech debt helps some emerging markets leapfrog.
Infrastructure debt is mounting: limited GPU capacity, fragmented data and constrained networks slow progress. Just 34% say their tech stack can adapt and scale for evolving compute needs. Most remain stuck in pilots.
Adoption plans are ambitious: 83% intend to deploy AI agents, with almost 40% expecting them to support staff within a year. Yet only one in three have change-management programmes, risking stalled workplace integration.
The leaders pair strong digital foundations with clear governance and cybersecurity embedded by design. Cisco urges broader collaboration among industry, government and tech firms, arguing that trust, regulation and investment will determine who monetises AI first.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
Yale University and Google unveiled Cell2Sentence-Scale 27B, a 27-billion-parameter model built on Gemma to decode the ‘language’ of cells. The system generated a novel hypothesis about cancer cell behaviour, and CEO Sundar Pichai called it ‘an exciting milestone’ for AI in science.
The work targets a core problem in immunotherapy: many tumours are ‘cold’ and evade immune detection. Making them visible requires boosting antigen presentation. C2S-Scale sought a ‘conditional amplifier’ drug that boosts signals only in immune-context-positive settings.
Smaller models lacked the reasoning to solve the problem, but scaling to 27B parameters unlocked the capability. The team then simulated 4,000 drugs across patient samples. The model flagged context-specific boosters of antigen presentation, with 10–30% already known and the rest entirely novel.
Researchers emphasise that conditional amplification aims to raise immune signals only where key proteins are present. That could reduce off-target effects and make ‘cold’ tumours discoverable. The result hints at AI-guided routes to more precise cancer therapies.
Google has released C2S-Scale 27B on GitHub and Hugging Face for the community to explore. The approach blends large-scale language modelling with cell biology, signalling a new toolkit for hypothesis generation, drug prioritisation, and patient-relevant testing.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
OpenAI will loosen some ChatGPT rules, letting users make replies friendlier and allowing erotica for verified adults. Altman framed the shift as ‘treat adult users like adults’, tied to stricter age-gating. The move follows months of new guardrails against sycophancy and harmful dynamics.
The change arrives after reports of vulnerable users forming unhealthy attachments to earlier models. OpenAI has since launched GPT-5 with reduced sycophancy and behaviour routing, plus safeguards for minors and a mental-health council. Critics question whether evidence justifies loosening limits so soon.
We made ChatGPT pretty restrictive to make sure we were being careful with mental health issues. We realize this made it less useful/enjoyable to many users who had no mental health problems, but given the seriousness of the issue we wanted to get this right.
Erotic role-play can boost engagement, raising concerns that at-risk users may stay online longer. Access will be restricted to verified adults via age prediction and, if contested, ID checks. That trade-off intensifies privacy tensions around document uploads and potential errors.
It is unclear whether permissive policies will extend to voice, image, or video features, or how regional laws will apply to them. OpenAI says it is not ‘usage-maxxing’ but balancing utility with safety. Observers note that ambitions to reach a billion users heighten moderation pressures.
Supporters cite overdue flexibility for consenting adults and more natural conversation. Opponents warn normalising intimate AI may outpace evidence on mental-health impacts. Age checks can fail, and vulnerable users may slip through without robust oversight.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
MIT engineers have created an AI system that can assess material quality faster and more cheaply by generating synthetic spectral data. The tool uses generative AI to produce spectral readings across different scanning modalities, allowing industries to verify materials without using multiple instruments.
By analysing one type of scan, such as infrared, SpectroGen can accurately recreate what the same material’s X-ray or Raman spectrum would look like. The process is completed in less than a minute with AI, compared with hours or days using traditional laboratory equipment.
Researchers said the system achieved a 99% match with real-world data in trials involving more than 6,000 mineral samples. The breakthrough could streamline quality control in manufacturing, pharmaceuticals, semiconductors, and battery production, cutting both time and cost.
Professor Loza Tadesse described SpectroGen as a ‘co-pilot’ for researchers and technicians. Her team is now exploring medical and agricultural applications in the US, supported by Google funding, and plans to commercialise the technology through a startup.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
Intel unveils ‘Crescent Island’ data-centre GPU at OCP, targeting real-time, everywhere inference with high memory capacity and energy-efficient performance for agentic AI.
Sachin Katti said scaling complex inference needs heterogeneous systems and an open, developer-first stack; Intel positions Xe architecture GPUs to deliver efficient headroom as token volumes surge.
Intel’s approach spans AI PC to data centre and edge, pairing Xeon 6 and GPUs with workload-centric orchestration to simplify deployment, scaling, and developer continuity.
Crescent Island is designed for air-cooled enterprise servers, optimised for power and cost, and tuned for inference with large memory capacity and bandwidth.
Key features include the Xe3P microarchitecture for performance-per-watt gains, 160GB LPDDR5X, broad data-type support for ‘tokens-as-a-service’, and a unified software stack proven on Arc Pro B-Series; customer sampling is slated for H2 2026.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
Nscale has signed an expanded deal with Microsoft to deliver about 200,000 NVIDIA GB300 GPUs across Europe and the US, with Dell collaborating. The company calls it one of the largest AI infrastructure contracts to date. The build-out targets surging enterprise demand for GPU capacity.
A ~240MW hyperscale AI campus in Texas, US, will host roughly 104,000 GB300s from Q3 2026, leased from Ionic Digital. Nscale plans to scale the site to 1.2GW, with Microsoft holding an option on a second 700MW phase from late 2027. The campus is optimised for air-cooled, power-efficient deployments.
In Europe, Nscale will deploy about 12,600 GB300s from Q1 2026 at Start Campus in Sines, Portugal, supporting sovereign AI needs within the EU. A separate UK facility at Loughton will house around 23,000 GB300s from Q1 2027. The 50MW site is scalable to 90MW to support Azure services.
A Norway programme also advances Aker-Nscale’s joint venture plans for about 52,000 GB300s at Narvik, along with Nscale’s GW+ greenfield sites and orchestration for target training, fine-tuning, and inference at scale. Microsoft emphasises sustainability and global availability.
Both firms cast the pact as deepening transatlantic tech ties and accelerating the rollout of next-gen AI services. Nscale says few providers can deploy GPU fleets at this pace. The roadmap points to sovereign-grade, multi-region capacity with lower-latency platforms.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
