i imaging any site made afer 2021 will require web access?
but apart form that, are all the template run with out web assess?
1 Like
Yes.
Read this article to understand it better
Can ChatGPT crawl live data from URLs? - No, but this is often not necessary
1 Like
The work-a-round is to scrape the website data in python, then train your prompt to extrapolate the necessary details from the scraped data. You can run an entire routine this way. You can scrape data, cleanse it, store it, then train your prompt on the data. You can also set multiple routines this way rather than fine tuning a single prompt, you can run prompts concurrently. IMO, this method is best if you’re managing you tokens.
3 Likes
As everyone is saying, the only ‘internet access’ that chatGPT has by default is that the internet (and you through it) can input to the prompt box. It has no other form of connectivity to the web other than output and the API. It cannot browse a page or access a URL, though it may sometimes seem to if the URL is one it can reasonably predict the likely content of.
If you want it to actually use the content of any URL you need to copy that content into your prompt yourself. It cannot do so. Only your prompt is dynamic, changing in accordance with variables. Everything past that is pre-trained.
1 Like
I wish I could understand your ramblings but I cannot. I can only offer my opinion as it relates to using the bot to crawl live data. Since it does not currently have this capability, I am suggesting the site be crawled and data scraped , the train prompt to manage the scraped data in the manner you wish. This is the work-a-round.
Do you have python code that you could share?
thanks i did this but it said the data was too much to enter, so i entered in 3 sections but it didnt work out very well as chatty seems to get confused lol
We all know that AI is really, really big right now. Lots of attention. And lots of attention means lots and lots of money pumping in to research and development. Then, we also know that AI is (at least in the public domain) still very new, and that these are just the first tentative baby-steps of a long, long journey.
Add those things together and you know that there is going to be change after change after change, every few months from now until it all hits a later stage of the growth process, or until all those R&D dollars go into some newer even shinier thing.
So, in the coming days, now and for the foreseeable future, things are going to change. In specific, in the next little while AIPRM are adding new functionality to crawl, scrape, and inject fresh data into prompts. AND in the next little while, OpenAI themselves have announced they’ll be releasing their own plugin that does the same or very similar things, direct from the makers of ChatGPT, that AIPRM will have to support anyway, in addition to their own version.
That means that 2 powerful tools that change the whole situation are both due out, perhaps in just days.
Given that, the sensible advice right now, to anyone, is hold off on any decision you can, because in just days that whole situation you’d make your decision about is going to change.
I don’t expect that OpenAI’s plugin will be exactly the same as AIPRM’s ‘Live Crawling’ functionality. However, I absolutely expect that one will be better for some specific uses and cases than the other, and that the other option will be better for other specific cases. What none of us can know, right now, is which is better for which things or uses, or why.
It’s a waiting game - but ought to be a very, very short one.
1 Like