Can LLMs Solve Unseen Problems? A Challenge for AI Skeptics

I told a colleague that AI tools are advancing rapidly.

He said he thinks LLMs are just internet search result simulators. He also claimed that the TypeScript problem he was working on was something an LLM couldn’t solve and showed me the code.

I told him that with enough context and reasoning resources, an LLM could probably solve it.

He admitted that he hadn’t used any advanced LLM models but believed that LLMs wouldn’t be able to solve problems that had never appeared on the internet before.

I ended the conversation with a simple laughing emoji. It made me think that I should come up with an example to show him.