From the course: Advanced Guide to ChatGPT, Embeddings, and Other Large Language Models (LLMs)
Join today to access over 25,200 courses taught by industry experts.
Evaluating alignment plus ethics - ChatGPT Tutorial
From the course: Advanced Guide to ChatGPT, Embeddings, and Other Large Language Models (LLMs)
Evaluating alignment plus ethics
whether or not a model works or not. Of course, it is about that, but that's really the first step. Because it's also meant to be a step to understand how well it is impacting the usefulness of the model in a real world scenario. So it's not just true or false, And if I give you a second one, is the second one better or worse? Just like we saw in our last section where OpenAI was choosing the better response of the two for reinforcement learning from human feedback, they had a human do that, at least in the beginning. The idea of human evaluation is not new of course. We've had AWS's Mechanical Turk for a while. I, by the way, as an organic Turk, I find that it slightly offensive. Scale AI is a big up and comer, not even up and comer at this point. A newcomer in this space at least, but still a behemoth in the field of human evaluation and labeling. And it's expensive. On average, of course, you can find this in better prices depending on who you work with but usually, you're thinking…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.