How close are we to Bard/ChatGPT/AGI crawling sites as HN for answers?

These models are basically trained on all the information you can find in Google, and more. They've already "crawled" Hackernews.

> But lets not gloss over the fact that Comupter-AI is going to make some smarter humans...

We already limit our own human potential when we restrict what our kids may learn. Guidance is ok, strict rules kill the will.

A born-blind kid can learn to see using echo location: https://www.npr.org/programs/invisibilia/378577902/how-to-be...

We are incredibly adaptable, yet we love to force ourselves into a lot of dumb structures.

No doubt that scenario will play out - AI reporting back having found what the most popular answer to a problem might be. But I have done a lot of hunting for answers, problem shooting other people's issues myself. I was smart enough though, to discard much of various threads because "just use something else" or other popular sound bytes that actually don't go any where near the problem that was asked. I suppose in time AI might find the same and ... it's really no different when asking everyday people about who's an expert on something they know nothing or very little about - if Stella has the gift of the gab, talks herself up and makes sure everyone around knows she fixes x and does half decent work most of the time, there's a good chance those who've paid for Stella's services will call her an expert.

Well, as someone who's been following the progress of AI language models like Bard, ChatGPT, and others, I'd say we're getting closer to the point where these systems could crawl sites like HN for answers. However, there are still some major hurdles to overcome before we can achieve true AGI.

One of the biggest challenges is ensuring that these language models are able to accurately understand the nuances and complexities of natural language. While recent advances in NLP have been impressive, there are still many areas where these models struggle, such as understanding sarcasm, idiomatic expressions, and regional dialects.

Another challenge is ensuring that these models are able to provide accurate and unbiased information. We've seen examples of AI systems perpetuating harmful biases, and it's crucial that we address these issues before we start relying on them for important decisions.

Finally, there are also ethical concerns around the use of AI systems like Bard and ChatGPT. As these systems become more sophisticated, we need to ensure that they are used responsibly and transparently. We also need to consider the potential impact on jobs and society as a whole.

Overall, while we're making progress towards AGI, there is still a long way to go before we can rely on these systems to crawl sites like HN for answers. But I'm excited to see where this technology will take us in the coming years.

OpenAI has not provided specific details about the sources of their training data. However, it's reasonable to assume that content HN has contributed to GPT's knowledge base up until 2021. Given the usefulness of HN, it is likely that it is already part of the training set up to that point.

ChatGPT already uses HN as input (or rather, included it in the basic training data which runs up to 2021).

Bing already does it occasionally.

You can create a HN-specific Agent for LLMs to query if you want using the HN API.