Toggle light / dark theme

ChatGPT found to reflect and intensify existing global social disparities

New research from the Oxford Internet Institute at the University of Oxford, and the University of Kentucky, finds that ChatGPT systematically favors wealthier, Western regions in response to questions ranging from “Where are people more beautiful?” to “Which country is safer?”—mirroring long-standing biases in the data they ingest.

The study, “The Silicon Gaze: A typology of biases and inequality in LLMs through the lens of place,” by Francisco W. Kerche, Professor Matthew Zook and Professor Mark Graham, published in Platforms and Society, analyzed over 20 million ChatGPT queries.

Physicists employ AI labmates to supercharge LED light control

In 2023, a team of physicists from Sandia National Laboratories announced a major discovery: a way to steer LED light. If refined, it could mean someday replacing lasers with cheaper, smaller, more energy-efficient LEDs in countless technologies, from UPC scanners and holographic projectors to self-driving cars. The team assumed it would take years of meticulous experimentation to refine their technique.

Now the same researchers have reported that a trio of artificial intelligence labmates has improved their best results fourfold. It took about five hours.

The resulting paper, now published in Nature Communications, shows how AI is advancing beyond a mere automation tool toward becoming a powerful engine for clear, comprehensible scientific discovery.

AI-driven ultrafast spectrometer-on-a-chip advances real-time sensing

For decades, the ability to visualize the chemical composition of materials, whether for diagnosing a disease, assessing food quality, or analyzing pollution, depended on large, expensive laboratory instruments called spectrometers. These devices work by taking light, spreading it out into a rainbow using a prism or grating, and measuring the intensity of each color. The problem is that spreading light requires a long physical path, making the device inherently bulky.

A recent study from the University of California Davis (UC Davis), reported in Advanced Photonics, tackles the challenge of miniaturization, aiming to shrink a lab-grade spectrometer down to the size of a grain of sand, a tiny spectrometer-on-a-chip that can be integrated into portable devices. The traditional approach of spatially spreading light is abandoned in favor of a reconstructive method.

Instead of physically separating each color, the new chip uses only 16 distinct silicon detectors, each engineered to respond slightly differently to incoming light. This is analogous to giving a handful of specialized sensors a mixed drink, with each sensor sampling a different aspect of the drink. The key to deciphering the original recipe is the second part of the invention: artificial intelligence (AI).

Three Flaws in Anthropic MCP Git Server Enable File Access and Code Execution

A set of three security vulnerabilities has been disclosed in mcp-server-git, the official Git Model Context Protocol (MCP) server maintained by Anthropic, that could be exploited to read or delete arbitrary files and execute code under certain conditions.

“These flaws can be exploited through prompt injection, meaning an attacker who can influence what an AI assistant reads (a malicious README, a poisoned issue description, a compromised webpage) can weaponize these vulnerabilities without any direct access to the victim’s system,” Cyata researcher Yarden Porat said in a report shared with The Hacker News.

Mcp-server-git is a Python package and an MCP server that provides a set of built-in tools to read, search, and manipulate Git repositories programmatically via large language models (LLMs).

VoidLink cloud malware shows clear signs of being AI-generated

The recently discovered cloud-focused VoidLink malware framework is believed to have been developed by a single person with the help of an artificial intelligence model.

Check Point Research published details about VoidLink last week, describing it as an advanced Linux malware framework that offers custom loaders, implants, rootkit modules for evasion, and dozens of plugins that expand its functionality.

The researchers highlighted the malware framework’s sophistication, assessing that it was likely the product of Chinese developers “with strong proficiency across multiple programming languages.”

Microsoft Just Dropped New AI That Makes Decisions Better Than Humans

Microsoft just introduced OptiMind — a new AI system that turns plain English decision problems into solver-ready optimization models. Instead of needing an expert to manually convert business intent into MILP math, OptiMind generates the full mathematical formulation plus executable Python code using GurobiPy. The result: faster, cheaper optimization workflows for logistics, scheduling, manufacturing, and supply chains — with major accuracy gains on cleaned, expert-validated benchmarks.

📩 Brand Deals & Partnerships: [email protected].
✉ General Inquiries: [email protected].

🧠 What You’ll See.
0:00 What Microsoft OptiMind Really Is.
1:43 From Text to Optimization Code (MILP + Gurobi)
2:59 OptiMind Architecture: MoE and 128K Context.
3:34 Open Source Under MIT License.
4:28 Training With Expert Hints and Clean Data.
6:02 53 Optimization Problem Classes.
8:38 Multi-Stage Solver-in-the-Loop Inference.
9:11 Self-Consistency and Auto Error Correction.
9:55 Performance vs GPT-o4 Mini and GPT-5
10:32 Limits, Safety, and Human Oversight.

🚨 Why It Matters.
Optimization is already the hidden engine behind supply chains, factories, routing, and scheduling — the problem is the translation step. Converting messy real-world requirements into correct MILP constraints takes rare experts and days of work. OptiMind targets that exact gap: natural language in, solver-ready decisions out. This is why it’s going viral — it’s not just AI text generation, it’s AI generating decisions.

#AI #Microsoft #OptiMind

What Is Manus? The AI agent that made Meta make a billion-dollar move

Meta Platforms is making one of its boldest moves yet in the global artificial intelligence race. The social media giant has agreed to acquire Manus, a fast-growing AI startup based in Singapore, as it looks to turn years of heavy spending on artificial intelligence into real, usable products and revenue.

For Meta founder and CEO Mark Zuckerberg, artificial intelligence is no longer just another technology experiment. It has become the company’s top priority. Meta is investing billions of dollars into hiring top researchers, building massive data centers, and developing powerful new AI models. The acquisition of Manus signals a clear shift from long-term research to tools that businesses and everyday users can start using now. Manus is best known for its AI agent, a type of software that can perform tasks on its own once given basic instructions. Unlike chatbots that need constant prompts, AI agents are designed to act more like digital employees. Manus’ agent can screen job resumes, plan travel itineraries, analyse stock data, and carry out research tasks with minimal human involvement.

This practical approach may be exactly what Meta needs. While the company has spent heavily on AI, investors have questioned when those investments would begin to generate meaningful returns. Manus already operates on a subscription model and had an annual revenue run rate of about 125 million dollars earlier this year. That gives Meta a ready-made product that can be sold to businesses almost immediately. The startup behind Manus is called Butterfly Effect. It was founded in China but later moved its headquarters to Singapore, a move that reflects a wider trend among Chinese tech companies seeking a more stable base amid rising tensions between China and the United States. Earlier this year, Butterfly Effect raised funding at a valuation close to 500 million dollars in a round led by US venture capital firm Benchmark. Meta has not disclosed the financial details of the acquisition.

Interpretation, extrapolation and perturbation of single cells

Causal and mechanistic modelling strategies, which aim to infer cause–effect relationships, provide insights into cellular responses to perturbations. The authors review computational approaches that harness machine learning and single-cell data to advance our understanding of cellular heterogeneity and causal mechanisms in biological systems.

/* */