Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Get the latest international news and world events from around the world.

Log in for authorized contributors

DeepSeek (@deepseek_ai) on X

DeepSeek-V4-Flash-Vision-Exp is now live on the DeepSeek API Platform! 🚀 đŸ”č This experimental multimodal model matches DeepSeek-V4-Flash on text capabilities—including agents, reasoning, and world knowledge. đŸ”č On multimodal agent benchmarks, V4-Flash-Vision-Exp makes a major leap over V4-Flash, bringing multimodal agent performance close to Opus-4.8. Try it with model=’deepseek-v4-flash-vision-exp’. DeepSeek Harness 0.1.1 was released today with out-of-the-box support for the new model. 1/n.

A 250seat jet is being built with four aisles instead of one. The reason is the shape: when you blend wings directly into a body wide enough to hold a small apartment, a single corridor down the middle no longer makes any sense

As aircraft bodies grow wider and flatter, the traditional single-aisle layout becomes a relic—forcing designers to reimagine what the inside of a plane should actually look like.

The Machine Metaphor That Shaped Modernity and Its Limits

Your body is a machine. Your brain is a machine. Soil is a machine. The planet is a machine.

We say these things so casually that they no longer sound like claims. They sound like common sense. We hardwire beliefs, rewire habits, debug organizations, optimize people, scale solutions. When something breaks, we go looking for the faulty component.

That metaphor built the modern world. Science, medicine, industry, computation. It did not just work; it worked spectacularly.

But every metaphor is a tool, not a truth. And this particular tool was built for closed systems with clean inputs and clean outputs. Look at what we now point it at: forests, economies, cultures, minds, and the #AI we are racing to deploy before we have agreed on what it is for.

I have argued for years that technology is the How, never the Why or the What. A machine cannot tell you what something is for, who benefits, or who pays the cost. It was never designed to.

So here is the question I could not shake while writing this: what if our deepest ecological, technological and cultural crises are not failures of intelligence at all, but failures of imagination? Failures of the stories we use to make sense of the world?

The problem with the machine story is not that it is false.

Continue reading “The Machine Metaphor That Shaped Modernity and Its Limits” | >

Over 30% shift stance on ethical dilemmas after AI rebuttal, Japan study shows

If ChatGPT challenged your moral judgment, would you be persuaded?

A recent study suggests that more than 30% of people would, highlighting the significant influence of artificial intelligence on views about questions with no easy answers.

The study also found that older adults showed a greater tendency to be swayed in their opinions, pointing to a potential danger of persuasive technologies in a rapidly graying society such as Japan.

Chinese RAM hits 9000 MT/s and even matches the speed and latency of the memory in our western gaming PCs

So, that’s 48 GB of “domestically produced”, ie, Chinese, DDR5 RAM hitting 9,000 MT/s. Admittedly, we’re far beyond that with Western memory, having hit over 8,700 MT/s all the way back in 2021 and 10,000 MT/s the following year. Now, we’re at over 13,000 MT/s.

Still, 9,000 is an achievement for Chinese memory and shows how fast the country is progressing in its DRAM development. In fact, Asus ROG also recently shared that Chinese memory now works at 6,000 MT/s with low CAS latency timings of C28 and C30. Notably, this is exactly the kind of fast DDR5 RAM we like to buy for gaming PCs in the West: DDR5-6000 memory with low latency.

The main Chinese memory maker is CXMT, and we’ve recently seen the company’s shares surge as it continues to collect more of the global memory market. This is possible because there’s just so much demand for memory these days, thanks to the AI boom, that demand far outstrips supply and will do for quite some time.

Investigational daily pill doubles survival in metastatic pancreatic cancer, study shows

In an international study co-led by UCLA, investigators found an investigational targeted therapy significantly extended overall survival for patients with previously treated metastatic pancreatic cancer and reduced the risk of death by 60% compared with standard chemotherapy in both the RAS G12 and the overall, or intent-to-treat (ITT) population, which included patients with and without identified tumor RAS mutations.

The team found patients who received the drug daraxonrasib, an investigational, first-in-class oral RAS(ON) multi-selective inhibitor designed to block active RAS signaling — one of the primary drivers of pancreatic cancer — lived a median of 13.2 months compared with 6.7 months for those who received investigator’s choice chemotherapy in the RAS G12 population.

The findings suggest a potential investigational treatment option for patients with metastatic pancreatic cancer, a disease that remains among the most lethal cancers and has limited effective therapies.

Designing supramolecular therapies to cross the blood-brain barrier

A new Northwestern Medicine study has uncovered key molecular design principles that could help supramolecular therapeutics cross the blood-brain barrier, a major challenge for this novel approach to treating neurological disorders.

Published in ACS Nano, the study examined how subtle changes in the structure of molecules called peptide amphiphiles affect their ability to move through brain endothelial cells and traverse the blood-brain barrier.

The findings provide a roadmap for developing new therapies capable of reaching the brain, including potential treatments for stroke, Alzheimer’s disease, Parkinson’s disease and other neurological conditions.

AI MetaHacking: How 3 Microsoft Copilot Flaws Exposed A New Attack Method

Varonis Threat Labs found a critical Microsoft Copilot vulnerability, CVE-2026–24301, without reverse-engineering any code. Researchers simply asked Copilot to explain its own security limitations, and it did. Microsoft patched the flaw, nicknamed CoSnitch, on August 18. It’s the third Copilot vulnerability Varonis has disclosed this year, and the AI meta-hacking discovery method may matter more than the bug itself.

AI meta-hacking just proved an AI assistant can be talked into exposing its own attack surface. Varonis Threat Labs disclosed CoSnitch, a chain of three vulnerabilities in Microsoft Copilot Personal, discovered through what the firm calls meta-hacking: prompting Copilot to explain why certain actions shouldn’t be possible, then using its own answers to map the exact boundary of what actually was possible, according to The Register’s coverage of the disclosure. Microsoft assigned the flaw a CVSS severity score of 8.8 and shipped a fix on August 18, 2026.

Varonis researcher HĂ„kon MĂ„lĂžy and colleagues didn’t start by probing code. They started by asking Copilot conversational questions about its own guardrails, and the assistant’s technical explanations revealed an undocumented URL parameter,?autorun=1, that could auto-execute a malicious prompt the moment a victim clicked a crafted link, according to The Hacker News. That single click gave an attacker’s prompt the ability to act inside the victim’s authenticated session, retrieving data from any connected app, Gmail, Google Drive, calendars, chat history, using Copilot’s own existing permissions. AI meta-hacking as a discovery method didn’t require finding a coding error at all; it required getting the AI to describe its own limits out loud.

‘Chameleon’ chip adapts to changing data speeds, cutting prediction errors by up to 40-fold

A research team led by Chair Professor Shinhyun Choi from the School of Electrical Engineering and the Graduate School of Semiconductor Technology has developed a programmable dynamic memtransistor (PDM), a semiconductor device whose time-response characteristics can be set to and retain multiple states, as well as an integrated array based on the device.

The research is published in the journal Nature Communications.

A memtransistor is a next-generation semiconductor device that combines the information storage function of memory with the computing function of a transistor. In the PDM, the ability to process data while retaining previous information allows its response characteristics to be adjusted and retained for incoming data.

/* */