3 ms·
There is no such thing as prompt "engineering". It's basically just trying different approaches until you come up with something that subjectively seems good en
by nradov 3y ago
There is no such thing as prompt "engineering". It's basically just trying different approaches until you come up with something that subjectively seems good enough. Nothing wrong with that, but let's not confuse the issue by labeling it as engineering.
- fl7305 3y ago> It's basically just trying different approaches until you come up with something that subjectively seems good enough. > let's not confuse the issue by labeling it as engineering. In my view, "trying different approaches" is a good description of engineering throughout history. Sure, it's excellent if you can base your engineering on a detailed physical model that lets you mathematically optimize a solution based on your boundary conditions. But compare that with metallurgy before we had atomic models. It was a process of trial and error. "Let's add small amounts of different alloy metals and see which ones makes the metal harder / more pliable / stainless / etc". That's still engineering to me. If anything, it could also be called science.
- nradov 3y agoWhat you're describing is artisanal or craft work. It's a crucial aspect of human endeavor but it's simply not engineering. Real engineering requires a foundation in accepted scientific theory and a consistent body of knowledge. Call it "prompt crafting" or something like that.
- fl7305 3y agoWas there an "accepted scientific theory" on metallurgy in the 1800s when gigantic metal ships were built? Were they not designed by engineers?
- nradov 3y agoYes, scientific knowledge of metallurgy based on analytical chemistry was fairly well established by the 1880s when the first successful large steel ships were built. It's impossible to produce large volumes of steel with consistent properties on an artisanal basis because the raw materials are inconsistent. Ship designers had also developed sophisticated mathematical techniques for calculating the strength of metal structures, and optimizing for weight and cost. They certainly weren't just riveting pieces of steel together by intuition and hoping it would float. I do consider those naval designers from the 1880s onward to be true engineers in the modern sense of the word. (At the time, engineers were mostly steam engine operators, so the meaning has changed since then.) Prior to the 1880s, large ships were generally composite wood and cast iron construction. While there was an aspect of engineering involved it didn't require the same level of theoretical knowledge and design was more artisanal. But that's a gray area.
- koolala 3y agoIt reminds me of 'social engineering'. Convincing it to do what you want.
- ianbicking 3y ago"It's basically just trying different approaches until you come up with something that subjectively seems good enough" Sounds like a lot of my engineering! Especially architecture, but generally any higher level code/object/function organization is exactly like this, and in practice even though I know a lot of patterns and have lots of experience and opinions, I often refactor architecture when I'm in a new domain. Which is also true of prompt engineering.
- nradov 3y agoSo what you're doing is probably software development rather than engineering per se. And I don't mean that in a negative or critical way. Most software domains don't necessarily require an engineering approach in order to produce good results. I have done the same thing myself. At some level we're just arguing semantics but I think there is intellectual value in being precise with labels.