There are more involved manners like this: https://github.com/williamcotton/transynthetical-engine/blob...

The New York Mets won the NL East Division titles in the following years: 1969, 1973, 1986, 1988, 2006, and 2015.

This is correct, btw.

"What years did the Mets when the NL East division?"

Without any reference to anything and it will give you the correct answer as well.

You have been duped into believing that ChatGPT reads websites. It doesn't.

without that data, it might, or might not, give correct info, and often it won't.

It also doesn't change the point that copying data has no effect. You could just ask ChatGPT what years the Mets won and it will tell you the correct answer.

To test this, I pasted the Wikipedia information but I changed the data, I just gave ChatGPT incorrect information about the Mets, and then I asked it the same question.

ChatGPT ignored the data I provided it and instead gave me the correct information, so what I pasted had no effect on its output.

Or it might make something up.

You don't ever know, and you can't know, because it doesn't know. It's not looking up the data in a source somewhere, it's making it up. If it happens to pick the right weights to make it up out of, you'll get correct data. Otherwise you won't. The fact that you try it and it works means nothing for somebody else trying it tomorrow.

Obviously putting the data in and asking it the data is kind of silly. But you can put data into it and then ask it to provide more nuanced interpretations of the data you've given it, and it can do reasonably well doing that. People are using it to debug code, I've personally used the ghidra plugins to good effect -- the way that works is to feed in the whole function and then have chatgpt tell you what it can deduce about it. It generally provides reasonably useful interpretations.

You can't give ChatGPT a Wikipedia article and ask it to give you facts based off of it. Ignoring its context window, even if you paste a portion of an article into ChatGPT, it will simply give you what it thinks it knows regardless of any article you paste into it.

For example I just pasted a portion of the Wikipedia article about Twitter and asked ChatGPT who the CEO of Twitter is, and it said Parag Agrawal despite the fact that the Wikipedia article states Elon Musk is the CEO of Twitter. It completely ignored the contents of what I pasted and said what it knew based on its training.

The person I was replying to claimed that if you give ChatGPT the complete context of a subject then ChatGPT will give you reliable information, otherwise it will "synthesize" information. A very simple demonstration shows that this is false. It's incredibly hard to get ChatGPT to correct itself if it was trained on false or outdated information. You can't simply correct or update ChatGPT by pasting information into it.

As far as your other comments about making stuff up or being unreliable, I'm not sure how that has any relevance to this discussion.

but also giving it the complete context does not guarantee you correct information, because that's not how it works.

Your earlier comment was "You could just ask ChatGPT what years the Mets won and it will tell you the correct answer." -- that is not accurate. That's my point. It doesn't know those facts. You might ask it that one minute and get a correct answer, and I might ask it that one minute later and get an incorrect answer. chatgpt has no concept of accurate or inaccurate information.

I'm not really sure if you're purposely doing this or not, so I'm just not going to engage further with you.

ChatGPT does not change the answers it gives you on the basis of providing it with an article. The example I replied to about the Mets can be tested by you, or by anyone, right now. Additionally I have provided the example of pasting a portion of the Wikipedia article on Twitter that clearly states that Elon Musk is its CEO, and then asking ChatGPT who the CEO of Twitter is. ChatGPT is not persuaded or corrected on the basis of any article you present to it.

>I'm not really sure if you're purposely doing this or not, so I'm just not going to engage further with you.

As an FYI, this makes you sound very insecure as a way of trying to manipulate someone into having the last word.

In the future, simply don't engage with someone you don't feel comfortable talking to, you are under no obligation to respond to anyone on the Internet. It's like someone who whines about leaving and then is told not to let the door hit you on the way out, just leave without making a fuss about it. It's far more dignified than the manner you have chosen to go about it.

Here is my research: https://github.com/williamcotton/empirical-philosophy/blob/m...

It is clear that analytic augmentations will result in more factual information.

Your claims are unfounded and untested.

"Based on the information provided, the current Prime Minister of Israel is Benjamin Netanyahu. He took office on 29 December 2022 and is leading the 37th government with a coalition that includes Likud, Shas, UTJ, Religious Zionism, Otzma Yehudit, and Noam."

QED.

I've been testing it by using it to generate and debug fairly complex C++ code all day, which it does based on the error messages I copy into the prompt window from the compiler. It's quite good at it, better than most college students at least.

You are very wrong about how these things work.

https://simonwillison.net/2023/Mar/10/chatgpt-internet-acces...