Jan LeCun “There’s no point studying LLM… but instead focus on next generation AI”
(*I think it’s a good read because other arguments against it are also included.)

Jan LeCun, chief AI scientist at Meta, said, “Students who dream of becoming future AI developers should not study LLM.” Instead, we should focus on next-generation models.

“LLM is in the hands of large companies. There is nothing we can bring,” he said. “We need to study next-generation AI systems that go beyond LLM.”

The company shared such information on Twitter and sparked an active discussion. AI developers, data scientists, and AI experts joined the discussion. They cited multimodal AI, advanced reasoning, robotics, unsupervised and self-supervised learning, and general artificial intelligence (AGI) as their next-generation systems.

There are also people who oppose this. They pointed out that this is the perfect time for students and others to work in LLM. For example, it is pointed out that there is still a lot to learn regarding prompts, jailbreaks, and accessibility.

In an interview with the Financial Times, he asserted that “LLM can never reach AGI.”
For that reason, LLM explained that the limitations are too obvious, such as very limited understanding of logic, not understanding the physical world, not having a lasting memory, not being able to deduce a reasonable definition of terms, and not being able to plan.
It also stated that LLM is inherently unsafe, as it can only respond to prompts accurately if the correct training data is provided.

Jeffrey Hinton, a professor at the University of Toronto, is also famous for having a different opinion. Professor Hinton believes that LLM is very similar to the human brain and should be all-in on LLM