General Politics XII World • Page 379

I asked CGPT:

“When giving answers about history, how do you choose which sources to use for your answers? Or do you take the most common answers across all sources and provide that to the user?”

Answer:

“When providing historical information, I aim for accuracy, balance, and context. Here’s how I approach it:

1. Primary Sources: When possible, I use primary sources or data from reputable historical archives, as they provide direct evidence about the topic.

2. Secondary Sources: For analysis and interpretation, I rely on well-regarded historians, academic publications, and authoritative books.

3. Cross-Referencing: I cross-check information across multiple sources to identify consensus and spot discrepancies. This ensures the answer is not skewed by one perspective.

4. Balancing Views: For controversial topics, I include multiple perspectives and note differing interpretations, so users see the nuance.

5. Current Scholarship: I prioritize sources that reflect the latest research while also acknowledging the contributions of foundational works.


When using tools like browsing, I prioritize established institutions, peer-reviewed content, and recognized experts. If the sources disagree, I highlight the debate and explain the differing views rather than synthesizing them into a single, potentially misleading “most common” answer.”

For a fast way to get answers on something, without spending a long time searching across multiple sites, it’s not a bad tool by any means to get info from. Of course it’s not 100% all the time (I’ve def. noticed this), but in general it’s pretty solid.

“What % of your answers would you say are 100% accurate, taking into account users correcting you?”

Answer:

“While my goal is to provide accurate and reliable answers, it’s challenging to quantify absolute accuracy in terms of percentages due to a few factors:

1. Complexity of Topics:

• For straightforward factual queries (e.g., historical dates, events), my accuracy is very high, likely above 95%, since these answers rely on well-established facts.

• For nuanced or debated topics, interpretations may vary, so “accuracy” depends on presenting balanced views rather than a single “right” answer.

2. User Corrections:

• Users occasionally provide corrections, sometimes due to my errors, but other times due to misunderstandings or differing perspectives. When corrected, I improve future responses by clarifying or re-evaluating the information.

3. Dynamic Knowledge Base:

• My knowledge is updated periodically, so I may lag behind current scholarship or emerging details for recent topics.


Overall, I’d estimate that 90–95% of my answers are factually accurate, with the remaining 5–10% representing areas where there might be:

• Minor errors or omissions.

• Gaps due to unclear user requests or incomplete historical consensus.

• Reliance on outdated or contested information.


Corrections and feedback are invaluable for refining my accuracy, so I always welcome users pointing out where I can improve.”

 

Okay but it is lying

 

Like, during the episode covering the Soviet nuclear program, I asked it "The Soviets launched their first nuke in 1949. How long until then had they been working on developing a nuke? Was one person responsible for the initial research?", and it gave me a great set of summarized information; then asked "Were the Manhattan Project spies ever caught? How close did they operate?", and it gave me a great rundown of all the spies involved and what happened to them after.

Of course its not info you'd want to base any academic research/publications on, but as a general tool to understand what happened during any particular event, I feel its been pretty good so far.


Curious to see what our teacher in residence @David87 thinks about students using it to gain understanding of whatever topics are being taught.

 

Maybe it's gotten better, I just asked for sources on a bunch of random (broad) topics and they were all at least real books and articles

 

But that documentary series is really good; you all should watch it. They even interview Fidel Castro several times, which was pretty cool

 

Ok

 

Are they all trying to avoid prison when the purge happens or something?

 

I was concerned about the environmental impact of ChatGPT until learning it can make a CNN documentary more enjoyable

 

i can use it for history, but only cause i ask really good questions.

 

I’m trying to be nice to our AI systems so that, when the AI overlords eventually take over, they at least spare me

 

bro if anyone posts chatGPT results in this thread again they’re getting a stone cold stunner

 

I'm in a few writer's groups on Facebook and whenever someone posts their book cover that was obviously created with AI for critique it gets torn to shreds in the comments. never gets old

 

I just realllllly don't trust it as a source, I'd rather just google and click around on sources and find different perspectives and etc. I just don't feel confident in what/where it's pulling the info from. Some of it seems legit but then there's so many examples where it just like made shit up haha.

For students specifically, I think it's really as perilous as any internet search where they could be led astray. I'd probably trust it more than social media but less than googling/wiki. My bigger worry with students is how many of them use it to flat out do assignments, which not only does it often not do the assignment correctly, but it robs them of any stamina or practicing of their critical thinking skills. I am kind of scared of what these kids will be like as adults tbh. Like how many doctors and nurses we gonna have in the future (or now, even) that were just on their phone throughout class and did everything with chatgpt? And they're gonna be taking care of us as we get older and have more health problems? Yeesh.

 

I really don't want to get all crotchety and 'BACK IN MY DAY' about it, but yeah even just the act of searching for sources and using some critical thinking skills to figure out if it's trustworthy or not is removed from the equation when you're instantly given the answers you are looking for. It's the instant gratification of the internet on steroids, hidden behind a black box of an algorithm that we are encouraged to trust implicitly. Doesn't seem good for the future or media literacy imo (or actual literacy, whoever shared those stats on literacy rates in this country a while ago, thanks for the nightmares). At least with searching the internet willy-nilly, if you wind up on a site that ends in '.balls' or something you can be like 'hmm maybe I shouldn't trust this'.

 

For me, if youre going to use chat gpt you can't be asking it synthesize info you can't immediately check. that's why it's better to ask it for sources which you can immediately confirm are real, instead of quotes from or analysis of the content of the sources.

 

On an almost daily basis I use AI results/ChatGPT to show my kids how stupid it is. Truly inept at writing what should be a simple high school English paper, down to getting the number of paragraphs wrong when specified in the prompt. Incorrect author names, made up quotes. It's really bad.

 

Ooo would you be able/willing to share an example of this that you've done?

 

y'all should bring me in to your classes to show the kids how to use chatgpt as a good writing and research tool.

 

anyway, what its best used for is math, not writing.

 

Now I'm curious. I'm gonna brun down some trees real quick to ask chatgpt to write my 3 page paper on the murder of Jane McCrae that I've already written and see what it says