It is an important information that you don’t really need petabytes of common crawl data to make a highly accurate bot. There are a few other open source models that preform well with significantly smaller training data that OpenAI.
That isn't what is being described here. They are just providing additional context to ChatGPT using its plugin API. It's still trained on large amounts of public text data.
>It is an important information that you don’t really need petabytes of common crawl data to make a highly accurate bot. There are a few other open source models that preform well with significantly smaller training data that OpenAI.
Sure, but the tradeoff is in generalization vs specialization. No one is impressed by the fact that ChatGPT is able to recite facts. Google can do that. Where it becomes interesting is in the general applicability of a single tool to thousands of possible domains.
