| ▲ | kibwen 12 hours ago | |
> One of my fundamental objections to LLM-driven software development was that model output is non-deterministic. Model outputs are deterministic. Your API provider may be injecting untunable amounts of randomness, but if you're not at the whim of an opaque cloud service, your model can be as deterministic as you please. The unease that you describe isn't about models not being deterministic (because they are), it's that generating outputs from LLMs is an inherently inductive process (operating backwards by extrapolating from observations and examples) rather than a deductive process (operating forwards from careful logical reasoning founded on axiomatic principles). Deductive processes are more reliable, predictable, and analyzable than inductive processes, but they're also much harder to produce for many classes of problems. | ||
| ▲ | saghm 2 hours ago | parent | next [-] | |
> Model outputs are deterministic. Your API provider may be injecting untunable amounts of randomness, but if you're not at the whim of an opaque cloud service, your model can be as deterministic as you please. "May be" is doing a lot of work. Can you point to any API provider that gives me deterministic output? I might be able to get away with running slower dumber models on my own hardware for my own projects if I want to tune this, but I'm dubious that there are many employers who are happy to let their employees do that, and when a large portion of developers nowadays are expected to use LLMs at their job, that means that non-determinism is pretty much a fact of life regardless of any theoretical ability to fix it. I don't understand why so many people try to argue against this being a problem because of the feature that pretty much nobody uses because no one can use it. It's like interjecting into a discussion about people getting hurt when falling from various heights by telling them that could avoid it by just going to the moon where gravity is lower; scientifically accurate, but not feasible for me because my job is here on earth and I don't have enough money to give that up and spend all my time on space. | ||
| ▲ | ibraryunus 12 hours ago | parent | prev [-] | |
[flagged] | ||