AI ethics expert Timnit Gebru cautions on AI 'automation bias,' advocates critical thinking

Munich, Nov 5 (PTI) Artificial Intelligence ethics researcher Timnit Gebru has stressed on responsible approach to AI, urging caution against 'automation bias', and advocating for small task-specific models that guzzle minimal data, compute and resources.

A computer scientist and prominent industry voice on artificial intelligence ethics, and algorithmic bias, Gebru underlined the importance of human cognitive skills, and exhorted audiences not to blindly trust machines over "common sense".

Recognised by TIME magazine for her influential work in the field of AI ethics, Gebru is known for her stance against gender and colour bias in commercial artificial-intelligence systems.

Speaking at Celosphere 2025, the annual event of Munich-headquartered process intelligence company Celonis, Gebru warned against excessive reliance on large AI models even for simplest of tasks.

"Beware of automation bias... think of critical thinking skills and don't overtrust machines," she said at a session titled 'Realistically Optimistic: Responsible AI with Dr Timnit Gebru'.

Gebru also cautioned about the hidden human and environmental costs of current large-scale AI models that are that guzzle large amounts of data and computing, and have huge energy requirements.

Instead, Gebru batted for "small, task-specific models" over one giant AI "machine god", where needed.

Citing an example, she said a speech recognition system should just transcribe speech, for which a specific tool should be sufficient.

She also touched upon the issue of "AI hallucinations," generative AI's tendency to make things up, and noted that it was not an issue in previous auto speech recognition tools. She also flagged "automation bias," where users end up overtrusting machines over their own critical thinking or judgement.

Her advice to leaders and organisations is not to outsource critical thinking skills and focus on the simplest possible solution using the smallest amount of data and resources.

She also favoured critical thinking in education and emphasised sustainability, privacy, and security in processes and systems.

"You know more than you think, see whether a particular tool is necessary... also apply some common sense," she said.

The observations assume significance amid the ongoing debate over whether or not AI may diminish human cognitive abilities over time.

Earlier this year, a study from researchers at MIT's Media Lab on the impact of GenAI on critical thinking abilities, revealed that although AI tools can improve efficiency, in the study, those who relied excessively on GenAI, overtime remembered less. PTI MBI DRR DRR