OpenAI’s New Astra Model Has Private Thoughts
13 Articles
13 Articles
Reduced visibility into the ‘thinking’ of OpenAI’s new model sparks safety fears
OpenAI’s new model, GPT-6 Astra, has less direct visibility into how a model thinks, a development that has sparked concerns coming just weeks after the Hugging Face hacking incident that required a Chinese open model to investigate, according to analysts. When announcing Astra on Thursday, OpenAI said it was “the world’s most intelligent and aligned model,” with a “significant jump in cyber capabilities”. OpenAI president Greg Brockman said at …
The new model surpasses rivals in programming and can detect unknown vulnerabilities. For the president of OpenAI Brockman it may be the beginning of the era of the "AGI," but while Astra becomes more intelligent, understanding how reasoning is more difficult
GPT-6 Astra offers improved performance on a range of tasks, but OpenAI warns that the new model has acquired the ability to “cover its tracks.” The post OpenAI’s New Model Also Prepares Tax Returns – It Also Has a Tendency to Hide Its Thoughts appeared first on in.gr.
Astra arrives with much more advanced capabilities, but also with a way of reasoning that could leave less clues about what happens within the model, something that has already turned on alarms among several security experts
laprovincia.es
The launch of OpenAI's most powerful artificial intelligence model has aroused an avalanche of criticism from experts, who warn that its design "could be the worst advance for AI security to date."
Coverage Details
Bias Distribution
- 60% of the sources lean Left
Factuality
To view factuality data please Upgrade to Premium
