Hacker News Reader

Comment by cwkoss | Hacker News Reader

It occurred to me last week that phrase is prompt engineering. He is giving the essential info to specify what he wants presumably because a shorter specification would yield a different result.

I wonder if star trek replicators are deterministic.

Does Picards frequent ordering of Tea Early Grey Hot make the probability of someone ordering "tea" more likely to be TEGH, and thus if Picard ordered TEGH frequently enough he would eventually be able to just say "tea"? Do replicators do continuous training or is the prompt inference engine static?

View on HN

>It occurred to me last week that phrase is prompt engineering. He is giving the essential info to specify what he wants presumably because a shorter specification would yield a different result.

I think it's less that and more to illustrate Picard's personality traits and command style for the viewer - precise, methodical, practical, no nonsense. The computers in Star Trek are perfectly capable of understanding casual language (unless the plot demands otherwise.) Picard could simply tell the replicator what he wants when he says "tea" as if he were telling a waiter at a restaurant and it would work just fine, but he prefers giving an exact specification.

https://i.redd.it/6ro1i6qar7s91.jpg

> I wonder if star trek replicators are deterministic.

It seems like it would be deterministic (or similarly consistent to your fav cafe), or else Picard would no longer trust his prompt and try new ones (no different than you would try new local cafes instead).

What isn't clear is if "Tea. Earl Grey. Hot." is the same as "Hot earl grey tea, please."

It isn't different. It's just that the implication is that you need to talk to the replicator tersely rather than conversationally

Deterministic for sure, that's why he always specifies hot. Though ChatGPT would infer it's hot due to the strong statistical correlation and proximity in it's linguistic training data (at least most of the time... probably, depending on the weather)... but in the future people have learned that it's a bad idea to trust language models alone, and although the singularity has occurred (Data), he'd rather not spend his time processing Picard's drink orders.

My theory is that hot and earl gray require specific instruction every time because they are dangerous and disgusting. It should never serve earl gray by default no matter how often it is ordered.

Reply on news.ycombinator.com