- A new study has found that artificial intelligence systems can now pass the Turing Test, with large language models being perceived as more human than actual people.
- The research rigorously tested modern large language models, including the latest versions of ChatGPT (GPT-4.5) and Meta’s LLaMa, against human participants.
- GPT-4.5 was identified as human 73 per cent of the time, significantly outperforming real people, while LLaMa scored similarly to humans at 56 per cent.
- Researchers suggest that the study's findings necessitate a re-evaluation of the Turing Test, shifting its focus from raw intelligence to measuring 'humanlikeness'.
- The study also highlighted the critical role of specific prompts and personas in enabling AI models to convincingly display human-like social and behavioural traits.
IN FULL
