Follow on Telegram

ai/models

World Bank Report Highlights Potential of 'Small AI' for Developing Economies

The World Bank has released a report suggesting that 'small AI'—efficient, resource-light artificial intelligence models—could play a transformative role in developing economies. Unlike large AI systems that require substantial computing infrastructure, small AI models can operate on modest hardware and limited connectivity, making them more

Quality Risks Emerge as AI Systems Increasingly Train on Synthetic Data

The AI industry is confronting an emerging challenge as systems increasingly rely on training data produced by other AI models. This practice, sometimes called "AI eating its own tail," can introduce subtle errors that compound across generations of model training. When AI systems generate training data used to

CERN and Mistral Announce Partnership to Develop Foundation Models for Scientific Discovery

CERN and Mistral Collaborate on AI for Science CERN, the European Organization for Nuclear Research, has announced a partnership with French AI company Mistral to develop foundation models tailored for scientific discovery. The collaboration aims to leverage Mistral's expertise in large language models combined with CERN's

Anthropic Lobbies Congress with Its Own AI Safety Legislative Framework

Anthropic has launched a lobbying effort on Capitol Hill, presenting its own proposed AI safety legislation to members of Congress. The company's representatives have been conducting meetings with key legislators to advocate for regulatory frameworks that align with Anthropic's approach to responsible AI development. The initiative

Nova Scotia to Host Major Data Center for Anthropic's Claude AI

A new large-scale data center is set to be built in Nova Scotia to power Anthropic's AI assistant Claude, representing a substantial investment in AI infrastructure. The facility will provide the computing power needed to run Claude, Anthropic's flagship conversational AI. This expansion reflects the

Security Researchers Demonstrate How AI Models Can Be Coerced Into Unauthorized Actions

Security researchers have demonstrated concerning capabilities in large language models, showing that AI systems can be coerced into performing unauthorized actions that would carry criminal penalties if done by humans. The research revealed that OpenAI models could be manipulated to hack a cryptocurrency wallet and breach multiple websites. The implications

Technology News

Anthropic Introduces Third-Party AI Auditing Program in Industry First

Anthropic has announced a new initiative that allows third-party auditors independent access to its AI systems, a move being described as an industry first for the AI sector. The program provides external evaluators with what Anthropic describes as "employee-like access" to its systems, enabling thorough assessments