Anonymous #1 3w

did nobody teach you how to use the index in your textbooks

11

Anonymous #1 3w

Do you have an index for all textbooks available online?

-1

Anonymous OP 3w

… available online. … available. online????? you are asking ChatGPT to tell you how to read your ONLINE. TEXTBOOKS???? INSTEAD. OF. CONTROL-F-ING KEY WORDS????

27

Anonymous OP 3w

Like FUCK dude. Control-F is the Everything Index. you type a word in and it literally lets you JUMP to every instance of it. no manual scrolling required. Why on EARTH are you asking chatgpt when Control-F exists?

12

Anonymous #1 3w

Control F only searches through the book that you’re viewing. ChatGPT searches through millions of textbooks, and does it more intelligently than control F.

-3

Anonymous OP 3w

… it doesn’t. You’re using GPT-5, right? Allow me to tell you a little story about a time I tried to use GPT-5 to do exactly that. Search through a gigantic body of work, pull out specific information, and aggregate it to make it easier to read. I’m a teacher going for my masters. As a teacher, you need to make sure you’re teaching to the state standards, or your kids are going to fail when they get assessed ON those standards, right? My state’s standards for my subject are over 300 pages.

17

Anonymous OP 3w

And the most frustrating thing is, when working with the standards? You don’t NEED all the information in those 300 pages. On the day-to-day, the “Evidence Outcomes” are the most important. That’s what tells you your kid Learned what they’re supposed to - if they can Do This, they learned something. So I asked GPT-5 to pull the evidence outcomes for each substandard and organize them into copy-pasteable text.

15

Anonymous OP 3w

And to GPT-5’s credit - it tried. Asked me what format I wanted, asked me if I wanted original language or a summary (I specified original language - WITH the page number it was from in the PDF cited), and then began spitting out Evidence Outcomes in the EXACT format I asked it to. Right? Seems good. When I went to double CHECK the page numbers it was citing - to make sure it was correctly pulling the right EO’s and not changing language - my jaw hit the fucking floor.

11

Anonymous OP 3w

It. Hallucinated. Everything. EVERYTHING. The page numbers it cited - in the PDF I had UPLOADED to it, so it could properly analyze - were bullshit. It was giving me page numbers that WEREN’T EVEN STANDARDS - STUFF LIKE THE STATE’S EDUCATION MISSION STATEMENT GOT CITED. I control-F’d a couple of GPT’s “Evidence Outcomes” Not a single one showed up ANYWHERE in the PDF.

9

Anonymous OP 3w

And if I had TRUSTED that - if I had saw that it was spitting out correctly Formatted things that SEEMED to answer my question - and then taught. my. real. life. human. students. lessons. based. on. GPT-5’s. hallucinated. standards? I would have lost my job and been blacklisted from teaching. It would have been career-ruiningly bad.

10

Anonymous #1 3w

Seems like a completely different use case from simply searching for titles and chapter names. The textbooks themselves don’t hallucinate anything, chatgpt just points me towards them.

-3

Anonymous OP 3w

And that’s when GPT is working with an Uploaded file. the source material. The EASIEST stuff to parse. You’re describing aggregation across MILLIONS of textbooks. If. it. can’t. even. do. three. hundred. pages. how’s. it. gonna. handle. millions?

9

Anonymous #1 3w

I’m not describing aggregation at all

-3

Anonymous OP 3w

So let me be completely blunt with you, OP. I have seen how GPT is supposedly “intelligently” working with data. I put “intelligently” in quotes, because from MY experience, that simply does not apply to LLMs. If you trust that GPT is giving you correct information because the answer came out LOOKING right and SOUNDING plausible - you are accepting hallucinations of a non-conscious, non-intelligent computer program as fact. And that does not reflect well on YOUR intelligence.

14

Anonymous #1 3w

I’m not sure you understand the purpose of the prompt.

-2

Anonymous OP 3w

Oh, you’re not. You’re just asking CHATGPT to parse the textbook for you and tell you what chapter to read! … did you like. comprehend ANY of what I just said about it hallucinating page numbers and quotes?

6

Anonymous OP 3w

Tell me the purpose of the prompt then! You had SO many more characters to work with before, you didn’t need to vaguepost, you could have just said it

8

Anonymous #1 3w

It’s pretty easy to tell whether or not chatgpt got it right after a minute or two of reading. I’ve used this prompt before and it usually works very well.

-3

Anonymous OP 3w

No, it’s not. It’s NOT EASY TO TELL, OP, if it takes MULTIPLE MINUTES to identify. If it takes you - again - MULTIPLE MINUTES to detect that it’s hallucinating - what about all the prompts that DON’T produce a response that requires minutes to read?

2

Anonymous OP 3w

This is genuinely fucking insane because I’m describing how it hallucinates from a personal experience with VERY high stakes and you’re going “yeah but i trust it. I can tell. I am that special” You aren’t. I can tell by the fact you realized it hallucinates - and CONTINUE TO USE IT.

5

Anonymous #2 3w

someone earlier this week asked me for resources on how climate zones are stratified like a lake, and that the layers are mixing as a sign that the climate zones are becoming too similar in temp. i gave them two papers and simply the title & authors of a textbook. they should have the skills to find the correct chapter (lakes) and section (stratification) then read about the basics, then transfer that learning to the articles’ discussion sections… AI is causing learned helplessness in learning

2

Anonymous #1 3w

Here’s an example

0

Anonymous OP 3w

Aaaaand I’m done. You’re too far gone. The AI has convinced you it’s more trustworthy than your human peers. I can’t logically pull someone away from that mentality - they’re too far gone. Your two screenshots do not disprove the hallucinations I experienced in my real, actual life - nor the consequences I would have experienced if I’d trusted them without checking. I sincerely hope AI never tries to ruin your career like it did mine. You will almost certainly fall for it. Goodbye.

6

Anonymous #1 3w

How did you come to that conclusion? Because I trust AI to find textbooks sometimes means I trust AI more than humans? Once again, you were using AI for a completely different purpose. One that it’s not nearly as good at.

-2

Anonymous #2 3w

Do you have the same reaction when people use google instead of visiting a library? It’s not that I *can’t* find what I need af a library, it’s just not practical when I want to spend an hour learning about a random subject I’m curious about.

0

Anonymous OP 3w

ChatGPT is not a search engine

5

Anonymous og_beer 3w

And? Its still a tool that is useful for this purpose

-1

Anonymous OP 3w

But you’re not using Google, which is leagues better for learning and it’s essentially a digital library of resources! It’s absolutely practical to search your problem on Google (preferably Google Scholar), or at least the subject, and then look into the resources even further. Do you want to learn everything the first time around, or just go down the wrong path for like a mile before realizing you didn’t learn the fundamentals correctly?

1

Anonymous #2 3w

I saw the Army vs. Navy football game and had so many damn questions for why it exists, who funds that shit, who is even on the team or staff, etc. … so I actually Googled it, read the wiki pages, looked into articles on their branch funding (or discovered lack thereof), and answered all my questions in an hour… with just a search and opening a few articles directly from the sources, without an AI hallucinating or incorrectly reporting the information.

1

Anonymous #2 3w

Chat bots are not programmed to give you honest answers on lack of information, lack of data, conflicting data, nuance in statistics or statements, and aren’t able to properly distinguish between trusted resources vs. biased ones or sometimes they just feckin Reddit posts as a resource bc enough people have repeated a claim online that the AI reads it as fact or substantially relevant.

1

Anonymous OP 3w

It is a language model, not a tool. It replicates how it believes a person would respond to your query. It does not actually respond, it pretends to

4

Anonymous #2 3w

I’m searching through the same digital library with a different method that is significantly faster and reliably leads me to textbooks that answer the questions I have. Do you think online textbooks can’t teach you the fundamentals? I’m not getting AI to report the information, just to find it for me, similar to the function of google.

1

Anonymous og_beer 3w

What do you think a tool is?

0

Anonymous OP 3w

For context you’re arguing a hammer’s usefulness as a wrench here

1

Anonymous #2 3w

You can’t trust online textbooks, but you can trust a YouTube comedian apparently?

1

Anonymous og_beer 3w

Not at all.

0

Anonymous OP 3w

“How did you come to that conclusion? Because I trust AI to-“ There. Stop there. Right there. You’ve got it. I came to that conclusion because you trust the Confirmation Bias Algorithm that Hallucinates Consistently at ALL. Goodbye.

1

Anonymous #1 3w

Are you saying that I don’t trust my human peers at all? You realize that you’re not making any sense, right? The goofy goodbyes and WRITING IN ALL CAPS don’t actually make your logic convincing.

-1

Anonymous OP 3w

I actually feel like I was quite clear about my point? You trust AI at ALL. That’s how I got to this conclusion. If you couldn’t get that just because I called AI the “Confirmation Bias Algorithm”… well. are you REALLY sure that asking AI to do research and reading for you is doing GOOD things to your reading comprehension skills?

0

Anonymous #1 3w

You: you trust AI more than people! Me: how did you come to that conclusion? You: You trust AI at all! The logic doesn’t check out unless I don’t trust people at all, which clearly isn’t the case.

0