Author: Carter

AI could cost you your job. Perhaps it already has. Maybe you’re finding it harder to break into a field that’s hiring fewer entry-level workers. And even if you manage to keep your position, the landscape is shifting: there are fewer roles available, the bar is higher, and you have less room to negotiate. That said, AI by itself isn’t the real threat. The real issue is AI being rolled out in an economy where employees have almost no say in how the benefits of increased productivity get divided up. Worrying About AI Makes Perfect Sense Every major technological shift…

Tincopa, M. A., Anstee, Q. M. & Loomba, R. New and emerging treatments for metabolic dysfunction-associated steatohepatitis. Cell Metab. 36, 912–926. (2024).Article  CAS  PubMed  Google Scholar  Lazarus, J. V. et al. Advancing the global public health agenda for NAFLD: A consensus statement. Nat. Rev. Gastroenterol. Hepatol. 19, 60–78. (2022).Article  CAS  PubMed  Google Scholar  Sumida, Y., Nakajima, A. & Itoh, Y. Limitations of liver biopsy and non-invasive diagnostic tests for the diagnosis of nonalcoholic fatty liver disease/nonalcoholic steatohepatitis. World J. Gastroenterol. 20, 475–485. (2014).Article  PubMed  PubMed Central  Google Scholar  Castera, L., Friedrich-Rust, M. & Loomba, R. Noninvasive assessment of liver disease…

Conventional cybersecurity defenses are no longer sufficient. AI-driven autonomous systems can create zero-day exploits instantly, and defense organizations struggle to patch vulnerabilities quickly enough to keep up. Bureaucratic procurement processes, institutional inertia, and a lack of deep technical expertise have created a major imbalance, with adversaries facing no practical limits on how aggressively they can escalate. AI-fueled attacks routed through space-based platforms or software-based channels bypass traditional attribution methods and overwhelm standard response protocols. The U.S. cannot simply overpower this threat through force. Success in today’s agentic warfare will be measured not by the strength of a counterattack, but by…

By Marc Kavinsky, Lead Editor at IoT Business News.The LoRa Alliance has released a three-year technical roadmap that focuses on simplifying how LoRaWAN is integrated and managed. The plan covers application integrations, device onboarding, network interfaces, coverage expansion, and certification. It signals a move beyond basic connectivity toward streamlined deployment and lifecycle management across the entire ecosystem.In many large-scale IoT projects, the wireless connection itself is rarely the biggest challenge. The real difficulties tend to lie in other areas: getting devices onboarded, decoding data payloads, linking servers together, migrating between networks, and filling coverage gaps at the edges of infrastructure.…

Nous Research has launched Hermes Desktop in public preview. This native app works on macOS, Windows, and Linux, giving the open-source Hermes Agent a proper visual interface. Previously, users could only interact with Hermes through a command-line interface or messaging platforms. The current version running behind the scenes is Hermes Agent v0.15.2. According to Nous Research’s official documentation, the desktop version taps into the exact same agent engine. Your settings, API keys, conversation history, learned skills, and stored memory all sync between the desktop app and the CLI or messaging gateway. Think of the desktop as just another window into…

AI-driven robots today can do a lot of things, but they’re still a long way from replacing people because they struggle to adapt to new conditions, according to experts.Last month, robotics firm Figure demonstrated humanoid robots doing simple chores like cleaning, while nine straight days of its robots sorting packages fueled debate over how quickly robots might take over jobs.Oliver Obst, an associate professor of robotics at Australia’s University of New South Wales, told Cointelegraph that highly routine roles, especially physical work in structured settings, are currently the most at risk. He added that office-based administrative and document-related work may…

Two recent reports present contrasting perspectives. One argues that security tools are failing to meet the actual needs of security teams. The other contends that the necessary tools already exist but are being poorly managed. Cybercrime has become increasingly industrialized, posing a growing threat to cyber defenses. This trend began before ChatGPT emerged, was dramatically accelerated by the rise of artificial intelligence, and now defines what’s known as the post-Mythos era. In this landscape, defenders must significantly enhance their capabilities—or risk losing ground to adversaries. Applications have become the primary battleground. AI-powered attacks are faster, more widespread, and more sophisticated…

This guide walks you through fine-tuning Liquid AI’s LFM2 using a fully open-source pipeline. We begin by loading the base LFM2 model with QLoRA, then prepare a supervised fine-tuning dataset in a conversational chat format. Next, we train a lightweight LoRA adapter using TRL and PEFT, and merge the adapter weights back into the main model. Additionally, the workflow includes an optional DPO stage to refine output preferences using chosen and rejected responses. By the end, you’ll have an end-to-end pipeline that transforms the original LFM2 model into an SFT-trained, preference-optimized version—ready for evaluation or deployment. Copy CodeCopiedUse a different…

The White House is rolling out fresh federal cybersecurity protocols and launching a new artificial intelligence security “clearinghouse,” as outlined in a long-awaited executive order on AI. Published on Tuesday, the executive order on AI innovation and security instructs government agencies to collaborate with the private sector to upgrade systems and “strengthen them against outside threats.” A key element of the order is its direction to multiple agencies to create an optional system where the government can assess cutting-edge, frontier AI models for cybersecurity threats before they are made available to the public. The order, which has been in development…

David Gewirtz/ZDNETFollow ZDNET: Add us as a preferred source on Google. ZDNET’s Essential Findings Claude Opus 4.8 manages uncertainty more effectively than 4.7.Several AI systems were used to verify the test outcomes.Even truthful AIs may still rationalize incorrect assumptions.Recently, Anthropic introduced its newest advanced large language model, Claude Opus 4.8. A standout aspect of this update is its increased honesty and “markedly improved judgment” compared to earlier versions.Also: Anthropic introduces Opus 4.8, featuring honesty as its headline capabilityBut does this claim hold up? In this piece, we put that assertion to the test.Before walking you through the entire evaluation process…