Anthropic has developed new security assessments for AI models that test their ability to sabotage. In a blog post, the company describes four types of tests: “human decision sabotage,” “code sabotage,” “sandbagging,” and “undermining oversight.” In human decision sabotage, the models try to trick people into making the wrong decisions without arousing suspicion. Code sabotage …
Meta has launched Spirit LM, its first Spirit LM is Meta’s first freely available multimodal language model that integrates text and speech inputs and outputs, positioning it as a competitor to models like OpenAI’s GPT-4o. Developed by Meta’s Fundamental AI Research (FAIR) team, Spirit LM aims to enhance AI voice experiences by improving speech generation’s …
AI startup Perplexity is in funding talks to double its valuation to $8 billion or more. The company is looking to raise around $500 million in its fourth round of funding within a year, according to the Wall Street Journal. The move comes on the heels of major investments in OpenAI and is intended to …
US companies are investing heavily in computers, data centers, and other infrastructure to fuel the AI boom. According to Joseph Politano, investment in fixed assets has increased rapidly to meet the growing demand for computing power. Hundreds of billions of dollars are being poured into high-end computers, data centers, power plants and more. In the …
Schools and universities are increasingly relying on AI detectors to check student work for AI use. According to one survey, about two-thirds of teachers use such tools regularly. As Jackie Davalos and Leon Yin report in Bloomberg Businessweek, at that scale, even low error rates can lead to a large number of false positives. The …
Mira Murati, the former chief technology officer at OpenAI, is reportedly seeking funding from venture capitalists for her new AI startup, Reuters reports. Her company aims to develop AI products using proprietary models, although it remains uncertain whether Murati will take on the CEO role. Initial discussions suggest that the startup could raise over $100 …
Researchers have developed a new method called “linear-complexity multiplication” (ℒ-Mul) to make calculations in artificial intelligence more efficient. The method replaces complex multiplications with simpler additions, according to Jason Hickey and his team at the Google AI Research Center in Accra. The researchers showed that ℒ-Mul achieves the same accuracy as traditional methods for language …
H2O.ai has unveiled two new vision-language models, H2OVL Mississippi-2B and H2OVL Mississippi-0.8B, aimed at enhancing document analysis and optical character recognition (OCR) tasks. Despite their smaller size, these models demonstrate competitive performance against larger models from major tech companies, with the 0.8B model excelling in the OCRBench Text Recognition task. CEO Sri Ambati emphasized that …
Perplexity has launched an Internal Knowledge Search feature that allows Pro and Enterprise Pro users to search both their internal files and the web from a single platform. Users can upload important files, limited to 500 for Enterprise Pro, in various formats like Excel, Word, and PDFs, VentureBeat reports. The aim is to streamline the …
OpenAI has launched a Windows desktop application for ChatGPT, enhancing its integration into daily workflows and productivity. This release follows a similar app for macOS and is currently available to select subscribers, VentureBeat reports. The app allows users to access ChatGPT easily via a keyboard shortcut, aiming to streamline user experience without the need for …