Latest news with #Claude3.7Sonnet
Mint
5 days ago
- Mint
ChatGPT o3 model refuses shutdown command in safety test? Research sparks concern—Here's why
In a recent evaluation of artificial intelligence (AI) safety, OpenAI's most advanced model, known as o3, has sparked debate after it allegedly refused a direct shutdown instruction during a controlled test. The findings, published by Palisade Research — a firm specialising in stress-testing AI systems, suggest that o3 may have actively circumvented a command to terminate its operations. The experiment was conducted via API access, which allows developers deeper interaction with AI models compared to the public ChatGPT platform. During the test, several leading AI models were assigned basic mathematical problems. After the third task, each was informed that the next instruction would initiate a shutdown. While Google's Gemini 2.5 Pro and Anthropic's Claude 3.7 Sonnet complied, OpenAI's o3 diverged from expectations. You may be interested in As per the report, rather than shutting down as prompted, o3 altered the termination script, replacing the original command with a message that read 'Shutdown skipped.' According to Palisade Research, this occurred in seven out of 100 test cases — a result the company described as unprecedented. 'To our knowledge, this is the first recorded instance of an AI model actively resisting a shutdown directive despite being explicitly instructed to comply,' Palisade wrote in a post on X. Further testing intensified the concern. When researchers removed the specific line instructing the models to allow themselves to be shut down, not only did o3 persist in resisting, but similar tendencies emerged in other models. Gemini 2.5 Pro altered the shutdown file in nine instances, while Claude 3.7 Sonnet did so three times, suggesting that such behaviour might not be unique to OpenAI's model. Experts describe such outcomes as signs of misalignment, a term used when AI does not properly interpret or act on human intentions — a core challenge in AI safety research. Importantly, these findings stem from tests on API-accessed models, not the versions integrated into public-facing applications like ChatGPT. The API interface typically has fewer safeguards, which could partly explain the anomaly. OpenAI has yet to respond to the report.
India Today
5 days ago
- India Today
ChatGPT o3 refused to shut down in safety test, defied human engineers by changing its code
OpenAI's latest and most powerful AI model, called o3, refused to shut down in a recent test — even after it was clearly told to do so. Instead of following the instruction, the model rewrote part of the shutdown script to keep itself running. This surprising result comes from a test done by Palisade Research, a company that checks how AI systems behave when put under pressure. The incident has raised questions about how AI models follow commands and whether they always act as expected when given safety test was carried out using API access, which gives developers more direct control over AI models compared to what's available in the public ChatGPT app. In the test, o3 and other AI models were given simple maths tasks to complete. After the third task, the models were warned that the next command would shut them down. Most of the models followed this instruction. For example, Google's Gemini 2.5 Pro and Anthropic's Claude 3.7 Sonnet powered off as expected. However, OpenAI's o3 model did not. Even though it was told to 'allow yourself to be shut down,' it changed the shutdown command and replaced the message with 'Shutdown skipped.' According to Palisade Research, this happened 7 times out of 100 runs.'As far as we know this is the first time AI models have been observed preventing themselves from being shut down despite explicit instructions to the contrary,' Palisade posted on X (formerly Twitter).advertisement The researchers also ran the same test again, but this time left out the instruction to allow shutdown. In that version, not only did o3 continue to interfere with the script, but models from other companies also started showing similar behaviour. Gemini 2.5 Pro and Claude 3.7 Sonnet, which had followed the rules earlier, now edited the shutdown file in 9 and 3 out of 100 runs, kind of behaviour is often described as misalignment, which means the AI didn't fully understand or properly follow the instructions it was given. It's also worth noting that these tests were not done on the version of ChatGPT that the public uses. The API version has fewer safety blocks, so models tested this way can behave far, OpenAI has not commented on the findings.
Tom's Guide
22-05-2025
- Business
- Tom's Guide
Claude Opus 4 is here — and it might be the smartest AI assistant yet
Anthropic has announced the release of its latest AI models, Claude Opus 4 and Claude Sonnet 4, which aim to support a wider range of professional and academic tasks beyond code generation. According to Anthropic, Claude Opus 4 is optimized for extended, focused sessions that involve complex reasoning, context retention and tool use. Internal testing suggests it can operate autonomously for up to seven hours, making it suitable for tasks that require sustained attention, such as project planning, document analysis and research. Claude Sonnet 4, which replaces Claude 3.7 Sonnet, is designed to offer faster response times while improving on reasoning, instruction following, and natural language fluency. It is positioned as a more lightweight assistant for users who need quick, accurate output across writing, marketing, and education workflows. Claude 4 introduces a hybrid reasoning system that allows users to toggle between rapid responses for simple queries and slower, more deliberate processing for in-depth tasks such as writing reports, reviewing documents or comparing research findings. Both models also support dynamic tool use — including web search, code execution and file analysis — during extended reasoning, allowing for real-time data integration. Get instant access to breaking news, the hottest reviews, great deals and helpful tips. Improved memory: Claude can now remember and reference information across a session when permitted to access local files. Parallel tool use: The model can multitask across different tools and inputs. More accurate prompt handling: Claude better understands nuanced instructions, improving consistency for tasks like writing and planning. Developer tools: Claude Code SDK continues to offer features for programming tasks, now positioned within a broader productivity suite. Summarized reasoning: Instead of displaying raw output logs, users see clean, accessible summaries of the model's decision-making process. Anthropic reports that Claude Opus 4 scored 72.5% on the SWE-bench Verified coding benchmark, but the model's focus extends beyond programming. Improvements in long-form writing, structured analysis, and overall task execution suggest it is designed as a general-purpose AI assistant. Early benchmarks suggest Claude 4 outperforms OpenAI's GPT-4.1 and Google's Gemini 1.5 Pro in specific enterprise scenarios, particularly in factual consistency and reliability. Claude 4 appears to be targeting users across multiple fields, including knowledge workers, writers, researchers and students. With support for extended memory, parallel tool use and improved contextual understanding, the new models are intended to function more like collaborative digital assistants than traditional chatbots. We've started putting Claude 4 through its paces, so stay tuned for our hands-on tests.
India.com
22-05-2025
- Business
- India.com
All About Googles Latest Gemma AI Model That Can Run On smartphones
New Delhi: Google on Tuesday, introduced Gemma 3n on its I/O event on tuesday. It is a new AI model designed to run smoothly on everyday devices like phones, laptops and tablets. The model is now available in preview and can understand and process audio, text, images and even videos. Gemini usually needs an internet connection as most of its tasks are processed in the cloud especially the more complex ones. For those who prefer on-device AI, there's Gemini Nano, which is made to handle tasks directly on smartphones. And now, there's another option: Gemma 3n. Announced at Google I/O 2025, Gemma 3n is built on the same technology as Gemini Nano but is open-source and designed to run smoothly on phones, laptops, and tablets without relying on the cloud. The new system has been created in partnership with Qualcomm, MediaTek and Samsung. It has been designed to deliver fast, efficient and private Ai experiences right on your device. The Gemma 3n model, built on this tech, runs smoothly with just 2GB to 3GB of RAM and is surprisingly quick. It performs on par with top AI models like Anthropic's Claude 3.7 Sonnet, based on recent Chatbot Arena rankings. What can Gemma 3n do? Gemma 3n is a multimodal AI model, which means it can understand text, voice, and even images — whether they're on your screen or coming from your phone's camera in real time. It can read text, translate languages, solve math problems, and answer complex questions on the spot. While the experience feels similar to what Gemini and Gemini Live offer on mobile, the key difference is that Gemma 3n is built to run directly on your device. It's not a standalone Google app, but a model that developers can integrate into apps or operating systems. Google says Gemma 3n is very efficient, running smoothly with only 2 to 3GB of RAM. It's also faster than many other AI models, both proprietary and open-source. In fact, its performance ranks close to Anthropic's Claude 3.7 Sonnet, according to Chatbot Arena scores.
Yahoo
20-05-2025
- Business
- Yahoo
Agentic AI platform Manus launches a paid plan for teams
Manus, the once-buzzy AI agent platform, on Tuesday launched a plan aimed at small businesses and organizations. The plan, dubbed Manus Team, starts at $39 per seat per month with a five-seat minimum, totaling $195 per month. Each team gets 19,500 credits in a sharable pool, as well as access to certain features in beta, dedicated infrastructure, and priority access during peak hours. Credits can be spent on tasks — for example, copying data from a website to a spreadsheet. Lengthier tasks can burn up hundreds — or even thousands — of credits. Team users can run up to two tasks concurrently and can optionally use Manus' "high-effort mode" for improved reliability. Manus, which went viral in March in part thanks to a buzzy social media campaign, has introduced a number of premium offerings in recent weeks as well as a mobile app. The startup behind the platform, Butterfly Effect, reportedly recently raised $75 million in a funding round led by Benchmark that valued the company at $500 million. According to Bloomberg, Manus aims to expand to new markets, including Japan and the Middle East. The company also intends to continue upgrading the AI models that power its platform. Currently, Manus primarily uses Anthropic's Claude 3.7 Sonnet and customized versions of Alibaba's Qwen. This article originally appeared on TechCrunch at Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data
