4 ms·
That's the idea, yes. However none but OpenAI knows exactly what ChatGPT was trained on. In fact, the dataset that it was trained on is so vast that they probab
by EdiX 3y ago
That's the idea, yes. However none but OpenAI knows exactly what ChatGPT was trained on. In fact, the dataset that it was trained on is so vast that they probably don't know either if it contains any given question.
IIRC last week I saw a study where GPT4 can solve some leetcode problems simply by giving it the number, no description. A clear example of overfitting.
- viraptor 3y agoThere was more information provided. And it's possibly not even overfitting. See https://news.ycombinator.com/item?id=38205153 https://news.ycombinator.com/item?id=38205153
- falcor84 3y agoThese tests should be conducted on new questions. And if we as humans no longer have the ability to create original questions, then maybe we should just retire.