Gary Marcus States Opinion on Limitations of Large Language Models

A researcher stated his view on large language models (LLMs). He stated that LLMs are of limited use. This opinion comes from his background as a practitioner in the field.

The researcher reported that hype surrounding LLMs, promoted by AI companies and individuals, is ill-founded. He included self-promoters on social media platforms like X and in academia in this assessment. According to the researcher, such promotion can be dangerous.

Large language models are AI systems trained on vast text datasets to generate human-like responses. They power applications in chatbots, content generation, and translation. The critique focuses on their underlying capabilities despite these uses.

The development of LLMs has accelerated with models from AI companies.

Companies such as Google and Meta have released similar systems, leading to widespread adoption. The researcher has previously written about AI limitations in books and articles. The researcher has worked on AI for decades.

His statements often highlight gaps between AI performance and human intelligence. This recent comment aligns with his ongoing discussions on the topic. The stakes involve investment in AI technology.

Affected parties include tech firms, researchers, and users relying on these tools for productivity. Regulatory bodies are beginning to examine AI risks, including misinformation from generative models.

The opinion contributes to debates on AI ethics and reliability.

Future developments may include improved testing for LLMs to address stated limitations. Researchers could explore hybrid approaches combining LLMs with other AI methods for better outcomes.