Many users pay for LLM subscriptions. But the margins are small, because what companies can charge for these services is barely above the cost of running them. There is also a lot of competition between different providers. The amount of investment is just completely disproportionate; it is a thousand times too high.
Why do you think that is?
There is just a ton of hype and outlandish expectations. Newspapers are running headlines like, «all jobs will be replaced soon» – «The 2028 U.S. elections will no longer be run by humans.» There is talk of artificial general intelligence. But these LLMs are more similar to large databases.
Artificial general intelligence (AGI) refers to a program that could solve all conceivable tasks. Do you doubt that LLMs are a step in this direction?
I don't believe that LLMs bring us any closer to human-like or general intelligence. These exaggerated expectations are also due to prominent studies which claimed that AI-models performed better than humans in law and math exams. We now know that language models simply memorized the right answers." https://www.nzz.ch/english/google-researcher-says-ai-hype-is-skewing-investment-ld.1825122
@remixtures I pay for ChatGPT and get my money's worth, but I find local LLMs are almost as powerful and I can give them much largers tasks like writing descriptions of hundreds of photos. ChatGPT does an amazing job of turning handwritten pages into markdown formatted text.
@njrabit: yes, I agree that the technology can be very useful but I'm afraid the companies haven't yet been able to find a viable business model - specially because there are lots of LLMs always emerging. Have you tried this blind test -> https://chat.lmsys.org/?
Add comment