HN is just flooded with AI BS these days.
Study mode
531–540 of 828 posts
Re: Study mode
#532Earlier quoted context omitted.
And then when you don't find it, you move onto the next book. Problem solved!
Or you know what, just google books about some topic and get a list of... real books recommended by people with names and reputations? Its truly incredible!
And now we get all the way back to the OP, and having so little knowledge on the subject that you don't know what to Google, or which forums are trustworthy for that topic. And so the wheel turns...
Re: Study mode
#533I had made a specialized prompt in Claude projects, for my learning, + added like field notes and lecture transcripts, was going good.
Then I tried to migrate it to chat gpt to try this thing out, but seems to be like it’s just prompt engineering behind. Nothing fancy.
And this study mode is not only not available in chat gpt projects, which students need for adding course work, notes, transcripts.
Honestly, just release gpt-5!!!
Re: Study mode
#534I'd like an LLM integrated spaced-repetition app. It would go along with this study feature quite well.
If LLMs continue to improve, we are going to be learning a lot from them, they will be our internet search and our teachers. If we want to retain some knowledge for ourselves, then we are going to need to learn and memorize things for ourselves.
Integrating spaced-repetition could make it explicit which things we want to offload to the LLM, and which things we want to internalize. For example, maybe I use Python a lot, and occasionally use Pearl, and so I explictly choose to memorize some Python APIs, but I'm happy to just ask the LLM for reminders whenever I use Pearl. So I ask the LLM to setup some spaced repetition whenever it teaches me something new about Python, etc.
The spaced repetition could be done with voice during a drive or something. The LLM would ask the questions for review, and then judge how well we did in answering, and then the LLM would depend on the spaced-repetition algorithm to keep track of when to next review.
Re: Study mode
#535yesterday I read a paper about using gpt 4 as a tutor in italian schools, with encouraging results - students are more engaged, get through homework by receiving immediate and precise feedback, resulting in non-negligible performance improvements:
https://arxiv.org/abs/2409.15981
it is definitely a great use case for LLMs, and challenges the assumption that LLMs can only “increase brain rot” so to say.
Re: Study mode
#536Earlier quoted context omitted.
> Learning something online 5 years ago often involved trawling incorrect, outdated or hostile content and attempting to piece together mental models without the chance to receive immediate feedback on intuition or ask follow up questions. This is leaps and bounds ahead of that experience. But now, you're wondering if the answer the AI gave you is correct or something it hallucinated. Every time I find myself putting…
What exactly did 2025 AI hallucinate for you? The last time I've seen a hallucination from these things was a year ago. For questions that a kid or a student is going to answer im not sure any reasonable person should be worried about this.
Are you using them daily? I find that maybe 3 or 4 programming questions I ask per day, it simply cannot provide a correct answer even after hand holding. They often go to extreme gymnastics to try to gaslight you no matter how much proof you provide.
For example, today I was asking a LLM about how to configure a GH action to install a SDK version that was just recently out of support. It kept hallucinating on my config saying that when you provide multiple SDK versions in the config, it only picks the most recent. This is false. It's also mentioned in the documentation specifically, which I linked the LLM, that it installs all versions you list. Explaining this to copilot, it keeps doubling down, ignoring the docs, and even going as far as asking me to have the action output the installed SDKs, seeing all the ones I requested as installed, then gaslighting me saying that it can print out the wrong SDKs with a `--list-sdks` command.
Re: Study mode
#537Earlier quoted context omitted.
> Learning something online 5 years ago often involved trawling incorrect, outdated or hostile content and attempting to piece together mental models without the chance to receive immediate feedback on intuition or ask follow up questions. This is leaps and bounds ahead of that experience. But now, you're wondering if the answer the AI gave you is correct or something it hallucinated. Every time I find myself putting…
What exactly did 2025 AI hallucinate for you? The last time I've seen a hallucination from these things was a year ago. For questions that a kid or a student is going to answer im not sure any reasonable person should be worried about this.
Just a couple of days ago, I submitted a few pages from the PDF of a PhD thesis written in French to ChatGPT, asking it to translate them into English. The first 2-3 pages were perfect, then the LLM started hallucinating, putting new sentences and removing parts. The interesting fact is that the added sentences were correct and generally on the spot: the result text sounded plausible, and only a careful comparison of each sentence revealed the truth. Near the end of the chapter, virtually nothing of what ChatGPT produced was directly related to the original text.
Re: Study mode
#538An underrated quality of LLMs as study partner is that you can ask "stupid" questions without fear of embarrassment. Adding in a mode that doesn't just dump an answer but works to take you through the material step-by-step is magical. A tireless, capable, well-versed assistant on call 24/7 is an autodidact's dream. I'm puzzled (but not surprised) by the standard HN resistance & skepticism. Learning something online 5…
There is no skepticism. LLMs are fundamentally lossy and as a result they’ll always give some wrong result/response somewhere. If they are connected to a data source, this can reduce the error rate but not eliminate it.
I use LLMs but only for things that I have a good understanding of.
Re: Study mode
#539An underrated quality of LLMs as study partner is that you can ask "stupid" questions without fear of embarrassment. Adding in a mode that doesn't just dump an answer but works to take you through the material step-by-step is magical. A tireless, capable, well-versed assistant on call 24/7 is an autodidact's dream. I'm puzzled (but not surprised) by the standard HN resistance & skepticism. Learning something online 5…
> Learning something online 5 years ago often involved trawling incorrect, outdated or hostile content and attempting to piece together mental models without the chance to receive immediate feedback on intuition or ask follow up questions. This is leaps and bounds ahead of that experience. But now, you're wondering if the answer the AI gave you is correct or something it hallucinated. Every time I find myself putting…
Just have a second (cheap) model check if it can find any hallucinations. That should catch nearly all of them in my experience.
Re: Study mode
#540Earlier quoted context omitted.
> The ability to say "I don't know" is not a high bar. For you and I, it's not. But for these LLMs, maybe it's not that easy? They get their inputs, crunch their numbers, and come out with a confidence score. If they come up with an answer they're 99% confident in, by some stochastic stumbling through their weights, what are they supposed to do? I agree it's a problem that these systems are more likely to give poor,…
but they're not. Ofyen the confidence value is much lower. I should have an option to see how confident it is. (maybe set the opacity of each token to its confidence?)
Logits aren't confidence about facts. You can turn on a display like this in the openai playground and you will see it doesn't do what you want.