Archive: March 2018

One of the most vexing public concerns regarding powerful AI systems such as large language models (LLMs) is their lack of interpretability: How can people trust a model’s output when they cannot understand how it was formed? In the new paper Finding Neurons in a Haystack: Case Studies with Sparse Probing, a research team from theMassachusetts Institute of Technology, Harvard University and Northeastern University proposes sparse probing, a technique desig

Goats are comedy gold. They headbutt confused cats! They faint! They make hilarious noises! Really hilarious noises! They do unspeakable things to the sheriff! And sometimes they’re complete and utter… well, let’s just say that as an Australian, I feel compelled to shout out the immortal Kevin, whose epithet shall not be spoken in polite American company but whose YouTube career should be rejoiced in all its foul-mouthed glory. The point is that if you wan

Getty / Futurism Investors are increasingly becoming concerned that AI companies may be grossly overvalued — and could struggle to justify their enormous hype by successfully ratcheting up profits. So it’s naturally alarming that shares of AI industry stalwarts, including Oracle, Nvidia, AMD, and Amazon, fell sharply on Tuesday, following concerns over enormous valuations that have propped up the entire US economy for months now. AI chipmaker Nvidia’s shar

Getty / Futurism Google’s forays into exposing users to dubious generative AI features have an incredibly poor track record. From error-ridden AI Overviews to AI slop dominating Google’s image search results, users have had to put up with a lot of needless and often easily avoidable nonsense. Now, Google Discover, the firm’s personalized content feed that’s heavily featured on Android phones, is showing users misleading and seemingly AI-generated headlines

How attackers turned a Teams meeting booking into a device code phishing trap Attackers used a fake Microsoft Teams meeting scheduler to trick an employee at a construction company into approving a device code sign-in. They got a valid session token before Huntress’s security operations center identified the incident. Device code phishing abuses a real Microsoft sign-in flow. The attacker starts a sign-in, gets a short code back, and persuades the victim t

Illustration by Tag Hartman-Simkins / Futurism. Source: Getty Images An indie game developer experienced a moment of profound moral clarity, and is now deleting his partially-AI-made game from the digital storefront Steam. The game, dubbed “Hardest,” is a free-to-play “rock-paper-scissors” roguelike card game released last July to mixed-to-negative reviews (of which there are barely thirty total.) Some users called it “soulless” and disparaged it for using

The 5-minute AI newsletter. The stories that matter, in your inbox every weekday morning. Subscribe Now

Google Gemini has a new feature: Letting users generate AI slop without visible watermarks to identify it. The official Twitter account for the LLM announced on Friday (news dump, everyone!) that users will soon gain the ability to “decide if your image, video, and music creations made with Gemini will have visible watermarks.” Now it’s a toggle. A few months ago, per Android Authority, Google VP Josh Woodward polled Gemini users on their most-requested fe

Research teams from the Massachusetts Institute of Technology (MIT) and Northeastern University have developed a new metamaterial incorporating network structures found in nature, such as leaf veins and dragonfly wings. Compared to traditional metamaterials with regular internal structures, these materials exhibit superior impact absorption performance and are expected to be applied to helmets, armor, packaging materials, and other uses. Scale-rich network

Credit: Pixabay/CC0 Public Domain From sourdough starters at home to loaves in supermarkets—artificial intelligence (AI) could be key to standardizing and upscaling unpredictable sourdough, according to new research. In a recent paper published in Trends in Food Science & Technology, scientists from Cardiff University, Shanxi University and Jiangnan University analyzed how combining AI, a biological analysis approach called multi-omics, and computational m

tirimid via YouTube Last year, OpenAI cofounder and former exec Andrej Karpathy coined the term “vibe coding,” a new approach involving the rapid development of software by feeding an AI model a series of natural language prompts. But the popular approach comes with some glaring shortcomings, as evidenced in an entirely vibe-coded operating system, dubbed Vib-OS. The project, which is being hosted on GitHub, can be run on ARM64 and x86-based computers, fea

The evolution of a single spin under coherent driving and spontaneous decay. TWA (solid lines) shows strong agreement with the exact solution (circles) for weak to moderate loss rates. Credit: PRX Quantum (2025). DOI: 10.1103/1wwv-k7hg Imagine zooming into matter at the quantum scale, where tiny particles can interact in more than a trillion configurations at once. If that sounds complicated, it is: Physicists often rely on supercomputers or even artificia

New智元 reports Just now, Anthropic launched the most severe cleansing in history. The official also introduced a new risk control policy, explicitly banning unsupported regions. This new regulation will officially take effect on November 12. Just yesterday, Anthropic released Fable 5.2 and officially started its IPO roadshow. However, for developers, the real shocker is this third move. November 12, three red lines Why is this considered the most severe in

Illustration by Tag Hartman-Simkins / Futurism. Source: DNY59 / Getty Images Google’s AI Overviews are peddling misinformation on a scale that may be virtually unprecedented in human history. A recent analysis conducted by the AI startup Oumi at the behest of The New York Times found that the AI-generated summaries, which appear above Google search results, are accurate around 91 percent of the time. In a sense, that may sound like an impressive figure. Bu