In tests, some AI systems resisted being shut down. But an instinct to stay “alive” isn’t the reason they behaved that way.
Articles on Large language models
Displaying 1 - 20 of 142 articles
None of the rapid development in AI exists in a vacuum. It is built on decades of blue-skies work by academics.
There’s no single giveaway. But there are patterns.
AI-powered breakthroughs are raising big questions about how to do mathematics – and why we do it in the first place.
A standardized framework for regulating the safety and efficacy of AI in healthcare has yet to be established.
AI models that are open source allow developers and users to see and modify the underlying program code.
Yet another longstanding mathematical question has been resolved by AI – and it feels a little different from the ones that have come before
A Commerce Department order requiring Anthropic to wall off its most advanced AI models from foreign nationals could be the government’s attempt to keep the best AI technology for itself.
Anthropic says there is something lurking inside Claude that looks a lot like the ‘global workspace’ one influential theory of consciousness depends on.
AI systems can now improve themselves by themselves. Is this the dawn of machine superintelligence?
‘Silicon sampling’ promises easy answers to tough questions about human behaviour – but the reality is more complicated.
AI-powered translation tools are certainly impressive. But there is an important frontier for translation technology, one AI might never be able to breach: the poem.
An “export control directive” for Anthropic’s Fable and Mythos models highlights the chaotic, fast-changing state of AI regulation.
The executive order is voluntary for AI companies but aligns with AI safety experts on the potential for harm.
AI models can simulate the answers thousands of people would provide to a survey, but the results aren’t a reliable measure of what real people would actually say.
Large language models can uphold falsehoods they or human users state, despite being presented with evidence to the contrary.
AI-powered cyber-security: the rise of new frontier platforms like Mythos highlights nuclear arsenals’ potential weak spots, and complicates the gambles inherent in nuclear deterrence.
Today’s AI systems are powerful, and it’s natural to see them as having humanlike intelligence. Shaking that illusion is important – and difficult to do.
There are good reasons why we see AI chatbots as more than what they truly are.
The future of artificial intelligence might not be as much a story about engineering as a story about evolution.
Top contributors
-
Director of Sydney Executive Plus, Business School, University of Sydney
-
Professor of Information Technology and Organisation, University of Sydney
-
Professor and Associate Director of the Centre for Change and Complexity in Learning (C3L), Education Futures, Adelaide University
-
Associate Professor of Linguistics and Director of Language Lab, The University of Western Australia
-
Professor of AI, Research Group Leader, UNSW
-
Professor of Statistics and Professor of Electrical Engineering and Computer Science, University of Michigan
-
Research Assistant Professor of Industrial & Systems Engineering, University of Southern California
-
Lecturer in Business Analytics, University of Sydney
-
Senior Research Fellow in AI Accountability, Queensland University of Technology
-
Associate Professor of Computer Science, Affiliate Professor of Information Science, University of Colorado Boulder
-
Professor of Philosophy and Director of Applied Ethics Center, UMass Boston
-
Doctoral Student in Cognitive Science, University of California, San Diego
-
Emeritus Professor of Psychology, Arizona State University
-
Associate Professor of Computer Science, Florida International University
-
Assistant Professor, Oregon State University