Their goal is to improve the results from companies’ AI tools. Willison says in order to do this effectively, prompt engineers must not be misled by AI’s many capabilities.
“It’s very easy when talking to one of these things to think that it’s an AI out of science fiction, to think that it’s like the Star Trek computer, and it can understand and do anything. And that’s very much not the case,” Willison told CBC.
“These systems are extremely good at pretending to be all powerful or knowing things, but they have massive, massive flaws in them. So it’s very easy to become superstitious to think, “Oh, wow, I asked it to read this web page. I gave you a link to an article and it read it.” It didn’t read it.”
The expert highlighted how the AIs are simply mimicking what is expected of them.
“A lot of the time it will invent things that look like it did what you asked it to. But actually, it’s really just sort of imitating what it thought you might … but really it’s sort of imitating what would look like a good answer to the question that you asked it,” Willison said.