No one knows what the real value of LLMs is though. Its a nascent technology. We could be discussing the internet circa 1994 or we could be discussing bitcoin circa 2011.
> the internet circa 1994 or we could be discussing bitcoin circa 2011.
People were clear about the value of both. SMTP by itself was already a clear value for business. The WWW itself was invented in 1989. Same for Bitcoin as the title of the original paper, "Bitcoin: A Peer-to-Peer Electronic Cash System
", explains it all. Some capabilities were not there yet, but it was mostly infrastructure and engineering issues (Acceptance was another issue for bitcoin).
LLMs generate tokens. The generation is governed by the models, but the model itself has no concept of usefulness or truth. Books are not enough to train them. The WWW has a lot of garbage informations. And natural language is insufficient to describe exact processes without formalizing it.
The fact is that the human mind can already do what the LLMs fans hope them to achieve. And we can transfer knowledge through media. Getting something done is mostly assembling the right people in a room, recursively. And we built tools to enhance productivity.
I believe that Microsoft and others are hoping that the general public will be ok with wrong results. But the fact is that all the capabilities exposed so far have an alternative and more deterministic process to achieve the same result. But it often requires learning or hiring a specialist and it's sad to see how many people are balking at that.