AI systems are great at tests. But how do they perform in real life?

By Peter Douglas, Lecturer, Monash Bioethics Centre, Monash University

Earlier this month, when OpenAI released its latest flagship artificial intelligence (AI) system, GPT-5, the company said it was “much smarter across the board” than earlier models. Backing up the claim were high scores on a range of benchmark tests assessing domains such as software coding, mathematics and healthcare.

Benchmark tests like these have become the standard way we assess AI systems – but they don’t tell us much about the actual performance and effects of these systems in the real world.

Read complete article

~	Kids with ‘developmental delay’ will be diverted from the NDIS. But how do you know if your child is delayed?

~	No room for the timid: setting Australia’s 2035 emissions target is a daring tightrope act

~	5 million small business employees now have a right to disconnect from work unless it’s ‘unreasonable’. What does that mean?

~	W.G. Sebald’s early critical essays mine his great literary themes – exile, trauma, memory and war

~	Zimbabwe's public debt servicing crisis leaves countless mother grieving

~	Australia has banned 3 ‘forever chemicals’ – but Europe wants to ban all 14,000 as a precaution

~	The Liberals used to be the party for women – then John Howard came along

~	Australia has banned 3 ‘forever chemicals’ – but Europe wants to ban all 14,000. This precautionary approach makes clear sense

~	Buckling rails and lines underwater: how Australia’s ageing train networks are crumbling as the climate changes

~	Destiny is a fierce new stage show exploring love, loss and rebellion under the shadow of apartheid South Africa