A new study shows that AI lacks self-learning skills and does not pose an existential threat

A new study published at the 62nd annual conference of the Association for Computational Linguistics (ACL 2024) has found that large language models (LLMs) such as ChatGPT cannot learn on their own and do not pose an existential threat.

Here's What We Know

The study emphasises that while LLMs can generate complex language, they are unable to acquire new skills without clear instructions, making them predictable and controllable. Dr Harish Tayyar Madabushi, from the University of Bath, said the prevailing view that AI poses a threat to humanity is a distraction from the real issues.

Professor Irina Gurevich added that while AI does not pose an existential threat, its potential misuse, for example to create fake news, needs attention. The study dispels concerns that AI can develop complex thinking abilities and emphasises the need to focus on other risks associated with the use of AI.

Source: Neuroscience