Get the latest Science News and Discoveries

AIs get worse at answering simple questions as they get bigger

Using more training data and computational power is meant to make AIs more reliable, but tests suggest large language models actually get less reliable as they grow

None

Get the Android app

Or read this on New Scientist

Related news:

Figuring out how AI models "think" may be crucial to the survival of humanity – but until recently, AIs like GPT and Claude have been total mysteries to their creators. Now, researchers say they can find – and even alter – ideas in an AI's brain.

AIs are irrational, but not in the same way that humans are

ChatGPT trains robot dog to walk on Swiss ball | This quadruped wobbling along, balanced on top of an exercise ball is a fun experiment to watch – but at its core, it demonstrates that AIs like GPT-4 can train robots to perform complex, real-world tasks much more effectively than we humans can.

« Chemists discovered the first new chemical bond in more than a decade

World's oldest cheese found on 3500-year-old Chinese mummies »