Chris Howard, the chief of research at Gartner, talked about the shift from fascination to implementation of Generative AI.
"I wanted to put it in focus for us because I think it's really important at this time in our history as we reckon with humans and technology. The first thing is that language itself is a technology, and writing as an aspect of language is something that, as far as we know, is unique to humans. And it is actually part of the way that we communicate with one another. And what's changing right now is it's becoming how we communicate with machines," he said.
Prompt engineering
Howard observed that most companies experimenting with large language models use the "prompt engineering" technique.
"You could ask a simple question to a chat interface, which will give you a reasonable answer. However, prompt engineering allows you to put more context into it as if speaking to a 12-year-old. And then a lot of what happens next is you search for information. So you may be searching for information inside your enterprise, and it pulls that context or that text back into the prompt and then bounces it against the GPT or the large language model," he explained.
From fascination to implementation
Training large language models for Howard builds a technical representation to make something work and talk like a human.
"That's as we move from a fascination about what this is, what this technology is, that's gripped us so in such an interesting way, to the true mechanics of building it so that it can produce the outcomes that maybe lead people to a better action, a better decision, keep them safe, or to interact with a customer or a client or a patient or a student," he said.
Originally posted on Gartner.
