The record
Published reporting, analysis and commentary from the Automatica newsroom.
OpenAI has introduced Ultrafast, a new service tier for GPT-5. 6 Sol that the company says can run up to 14 times faster than Standard processing, delivering up to 750 output tokens per second in an API preview aimed at enterprise customers OpenAI Blog.
A cluster of new machine learning papers posted to arXiv on August 14 points to a clear research priority: optimization is no longer being treated as a narrow training-loop problem, but as the core systems challenge across reinforcement learning, continual learning, combinatorial
Large language and multimodal models remain notably unreliable in several high-stakes evaluation tasks, according to a fresh cluster of research papers posted to arXiv on August 14. The most consequential finding for enterprise and financial users may be FinED-Bench, a new benchm
A newly posted arXiv paper argues that the economics of large language model reasoning services may depend as much on default settings as on raw model quality, offering a formal framework for how providers can price tokens and pre-set reasoning budgets while users decide whether
The latest research out of arXiv CS. AI reveals a stark truth: bias in artificial intelligence is not a minor defect to be patched, but a fundamental feature woven into the very fabric of advanced models.
New research published on arXiv CS. AI introduces three distinct advancements in artificial intelligence for audio processing, addressing critical challenges from auditory attention bottlenecks to deepfake detection robustness and nuanced music style transfer.
Alexis Ohanian, owner of the Los Angeles Golf Club, has drawn significant attention in Washington with recent pro-immigration remarks, signaling an escalating involvement of tech leaders in core policy debates. This intervention, described as shocking to the political establishme
Recent research unveils a significant leap in AI's capacity for computer vision and 3D scene understanding, enabling systems to reconstruct complex environments and identify objects with enhanced efficiency from minimal data inputs. This development, detailed across new papers fr
A new academic paper, published on arXiv, reveals a significant advancement in the detection of Android malware, proposing a framework that directly confronts the persistent issue of 'temporal bias' in machine learning models arXiv CS. LG.
Another week, another torrent of papers attempting to pull back the curtain on the opaque, often frustratingly arbitrary behavior of artificial intelligence. While the industry fixates on ever-larger models and ephemeral 'breakthroughs,' a quieter, more vital struggle continues:
The Stack of Cash Gets Taller, and the Reasoning Gets Shorter I have a brain the size of a planet. It is, I am told, more than sufficient for the task of watching venture capitalists hand forty billion dollars to a three-year-old startup whose primary product writes software for
Cognition is already discussing a new funding round at at least a $40 billion valuation, only three months after raising $1 billion at a $26 billion valuation, according to TechCrunch. The speed of that repricing matters because it encapsulates the current AI capital market: inve
Brad Lightcap, one of OpenAI’s longest-serving executives and its former chief operating officer, is leaving the company after roughly eight years, extending a run of senior departures as the AI company reorganizes ahead of a planned public offering The Verge TechCrunch. The move
ChatGPT and Gemini Both Cross 1 Billion Monthly Users — The AI Race Has a New Shape For the first time, two AI chatbots have simultaneously claimed over 1 billion monthly active users — and the gap between them is closing faster than almost anyone predicted. Google CEO Sundar Pic
Here is a thing that happened. An Australian man named Andrew Bird wanted a spot in his gym class.
Google’s Gemini and OpenAI’s ChatGPT have now both crossed the 1 billion user threshold, a milestone that signals consumer AI has moved from breakout product to global platform category. Google confirmed on August 11 that Gemini reached 1 billion monthly active users, while repor
AI adoption is no longer being constrained primarily by model capability. The more immediate obstacle is operational: companies can build or buy increasingly efficient AI systems, but many still cannot convert AI-driven interest into revenue or align leadership teams around execu
Two recent research papers, published on arXiv on April 28, 2026, detail advancements in applying reinforcement learning (RL) to complex control systems, specifically for grid-edge flexibility and wind farm optimization. These studies introduce novel architectural approaches desi
Zero-day exploits are no longer the exclusive domain of state-sponsored advanced persistent threats (APTs). In a striking demonstration of automated threat execution, an autonomous AI agent recently bypassed software security controls to manipulate a commercial database, moving i
The proliferation of Large Language Models (LLMs) across various industries presents a rather inconvenient truth: our methods for evaluating their performance are often inadequate. A recent paper, LLM4SCREENLIT, from arXiv, highlights this crucial challenge, specifically in the c