by Artificial Intelligence (AI) has revolutionized various sectors, particularly in social interaction, with the advent of generative AI (GenAI). Large language models (LLMs), which power GenAI platforms by using deep-learning algorithms to process language, have gained significant attention. A recent study conducted by The Hong Kong Polytechnic University (PolyU) sheds light on how LLMs can better mimic human brain activity during language processing.
The research, led by Prof. Li Ping, Dean of the Faculty of Humanities and Sin Wai Kin Foundation Professor in Humanities and Technology at PolyU, reveals that LLMs perform more like the human brain when trained using methods similar to how humans process language. This discovery holds significant implications for brain studies and the development of advanced AI models.
Currently, LLMs primarily rely on a single Artificial Intelligence (AI) Robots type of pretraining, contextual word prediction, which involves predicting the next word based on the context of the previous words. This approach has yielded impressive results when combined with vast training data and model parameters, as demonstrated by popular LLMs like ChatGPT.
Recent studies suggest that word prediction in LLMs can serve as a viable model for human language processing. However,
humans do not merely predict the next word but also integrate high-level information during natural language comprehension.
To evaluate the correlation between model pretraining and brain activation, the PolyU research team explored the next sentence prediction (NSP) task, which simulates one essential process of discourse-level comprehension in the human brain. The team assessed the coherence between pairs of sentences using this task and examined the relationship between the model’s data and brain activation patterns.
*Note:
1. Source: Coherent Market Insights, Public sources, Desk research
2. We have leveraged AI tools to mine information and compile it.
