I am in bed, but the hallucination problem doesn't need to be solved to have value. LLMs are shit in, shit out. But, an LLM that only has access to high-quality content will provide better outputs than a human with the same content, with limited hallucination. This already happens, where the AI is contained only to a set of internal corporate documentation created by highly paid, role-specific humans. The return on value is immense for a company.
Will get into this more no doubt :)
Goodnight!
Goodnight?!? It's only 6pm!! I'm kidding. Hope you're now having a lazy Sunday.
Maybe I'm just not getting it... but I don't really see the immense value in a system that regurgitates existing internal corporate documentation slightly differently that may also include hallucinations - especially when you remove highly confidential or sensitive information that you don't want everyone in the company to access.
I would think companies would more highly value accurate data or new information. We have a company-specific GenAI available to us now but I since I deal with sensitive data I haven't found a usecase for myself yet.
A simple usecase:
Let's say you have a standard contract for a customer and then each actual contract could change terms and conditions.
"What are the differences between our stand TandCs, and this specific contract?"
Saves hours for a lawyer. That lawyer charges a lot per hour. Either less lawyers needed, more billable hours, or more efficient hours, billing hours that only took minutes.
Oh? Would that comparison be done by GenAI or by a AI/ML tool? I definitely never thought of the analytical potential of GenAI, I was stuck on usecases where its creating something.
This does seem useful. I guess it would transfer the responsibility of getting it correct from your lawyers to you though. If something is wrong in the contract that ends up costing you money, that should be on the lawyer, but now it would be your responsibility. I'd be curious if companies would risk the time/money saved with the new accountability, but I guess we'll see.