Using AI to prepare and evaluate environmental assessments risks ‘robodebt-style’ failures, scientists say

Here is a summary of the key points outlined in the article:

– The Mining Council of Australia has proposed spending $13 million on a trial to use AI in environmental approvals, citing potential efficiency gains.

– The Biodiversity Council, comprised of independent experts across 11 universities, has expressed concern that AI automation might replicate “Robodebt-style” errors, which could fail to protect threatened species due to flawed decisions and lack of transparency.

– Lis Ashby from the Biodiversity Council argues that vague legal language in Australia’s key environmental protection law creates extra complexity for human assessors, which would be even more problematic for AI tools.

– There’s skepticism about using AI for environmental assessments as the data available on many threatened species is insufficient and outdated, potentially leading AI to make flawed decisions.

– Brendan Sydes from the Australian Conservation Foundation suggests that instead of relying on AI, the government should fill existing gaps in data regarding threatened species and habitats.

– Prof. David Lindenmayer notes that current gaps in knowledge about Australia’s threatened species could lead to poor AI decisions based on incomplete or outdated information.

– The federal government should focus on employing more people to carry out accurate assessments rather than trusting AI, according to Prof. Hugh Possingham.

– Tania Constable, chief executive of the minerals council, defends the AI proposal, describing it as innovative and highlighting that AI would support human decision-making rather than replace it.

– A federal government spokesperson emphasized that while AI could help in the process, final approval decisions must always be made by human assessment officers.