2023-09-07
48 分钟For all the speculation about the future, A.I. tools can be useful right now. Adam Davidson discovers what they can help us do, how we can get the most from them — and why the things that make them helpful also make them dangerous. (Part 3 of "How to Think About A.I.")
Hey there, it's Stephen Dubner.
You are about to hear the third and final episode in our three part series called how to think about AI.
The guest host for this series is Adam Davidson, one of the founders of NPR's planet Money.
Here's Adam can you just tell us who you are and what your job is?
This is a question I dread at parties, but my name is Anna Bernstein.
I am the prompt engineer at a company called Copy AI.
Have you heard of this new job, prompt engineer?
It's a job that could only exist right now, a job that satisfies a need that almost none of us even knew we might have until the last few months.
So what is it exactly?
The prompt engineer essentially is an expert in being the linguistic intermediary between user input and AI output.
Please note how precise Anna Bernstein's language is.
That is her superpower, being really, really precise about language.
If you've played around with AI tools like chat, GPT, or Google's bard, you've seen it.
They respond to the precise words you type in.
They can't figure out a vibe or a hidden intention.
Also, unless you happen to work at an AI company, you don't ever interact with the raw model itself.
Whatever you type in is put through a filter, a filter that people like Bernstein design.
The filter between you and the raw AI model is there for a bunch of reasons.
The first is to avoid the ugly stuff.
Since the AI models were all trained on an unfiltered mass of text from the Internet, they can easily call up truly horrific, offensive words and ideas if left unfiltered.