6 ms·
I don't know how intentional it is / was, but LLMs in general just love to hear themselves talk!
by osigurdson 2mo ago
I don't know how intentional it is / was, but LLMs in general just love to hear themselves talk!
- arjie 2mo agoThey do, and I want to encourage them to do so because they think through talking. What I don’t want to do is spend time reading all that. We will probably just get reader-side affordances for this like auto-folded justification and introduction sections and so on. Doubtless some chat interface will add this the way they’ve added reasoning folding.
- bcrosby95 2mo agoThinking models think through talking, don't reveal that talking, then answer by again thinking through talking. It's kinda funny in a way.
- mdgld 2mo agoI always expose reasoning traces. How else can you seriously debug?
- girvo 2mo agoThe closed models aren't giving you the real thinking traces, though.
- joquarky 2mo agoThose traces are just summaries of the reasoning.
- alfiedotwtf 2mo agoIt’s wild the people don’t want to see the thinking
- jimbokun 2mo agoIs it just a coincidence that the companies creating them charge by the token?
- pizzafeelsright 2mo agoThe aligned incentive appears to be realigning in favor of the corporation. Pray they do not realign them further. There are times I require single word answers. I will use whatever model responds as I desire and at this point those models are just a few.
- minimaxir 2mo agoThe cost-per-task benchmarks align incentives toward more efficient output and those are the ones gaining steam.
- mr_toad 2mo ago> LLMs in general just love to hear themselves talk! Because that’s what’s in the training set. Reticent humans don’t have blogs.
- flir 2mo agoInteresting idea. I think they're getting more wordy over time, personally, so I think it's more to do with the training than the raw data.