Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Study mode

openai.com

Post reply on HN
Print view View original

481–490 of 828 posts

Re: Study mode

#481

Post by baby »

zvmaz wrote: post #387
jacobedawson wrote: post #379

An 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…

The fear of asking stupid questions is real, especially if one has had a bad experience with humiliating teachers or professors. I just recently saw a video of a professor subtly shaming and humiliating his students for answering questions to his own online quiz. He teaches at a prestigious institution and has a book that has a very good reputation. I stopped watching his video lectures.

You might also be working with very uncooperative coworkers, or impatient ones

Re: Study mode

#482

Post by throwaway31131 »

r3trohack3r wrote: post #458
tekno45 wrote: post #444

Earlier quoted context omitted.

How do you know? its literally non-deterministic.

Most (all?) AI models I work with are literally deterministic. If you give it the same exact input, you get the same exact output every single time. What most people call “non-deterministic” in AI is that one of those inputs is a _seed_ that is sourced from a PRNG because getting a different answer every time is considered a feature for most use cases. Edit: I’m trying to imagine how you could get a non-deterministic…

> I’m trying to imagine how you could get a non-deterministic AI

Depends on the machine that implements the algorithm. For example, it’s possible to make ALUs such that 1+1=2 most of the time, but not all the time.

…

Just ask Intel. (Sorry, I couldn’t resist)

Re: Study mode

#483

Post by huitzitziltzin »

I would love to see more of their game theory example.

Having experience teaching the subject myself, what I saw on that page is about the first five minutes of the first class of the semester at best. The devil will very much be in the other 99% of what you do.

Re: Study mode

#484

Post by ZYbCRq22HbJ2y7 »

qualeed wrote: post #300
ZYbCRq22HbJ2y7 wrote: post #118

Earlier quoted context omitted.

This is important, as benchmarks indicate we aren't at a level where a LLM can truly be relied upon to teach topics across the board. It is hard to verify information that you are unfamiliar with. It would be like learning from a message board. Can you really trust what is being said?

> we aren't at a level where a LLM can truly be relied upon to teach topics across the board. You can replace "LLM" here with "human" and it remains true. Anyone who has gone to post-secondary has had a teacher that relied on outdated information, or filled in gaps with their own theories, etc. Dealing with that is a large portion of what "learning" is. I'm not convinced about the efficacy of LLMs in teaching/studyin…

Sure, humans aren't without flaws in this area. However, in real time, humans can learn and correct themselves, we can check eachother, ask for input, etc, and not continue to make mistakes. This isn't the case with LLMs as a service.

For example, even if you craft the most detailed cursor rules, hooks, whatever, they will still repeatedly fuck up. They can't even follow a style guide. They can be informed, but not corrected.

Those are coding errors, and the general "hiccups" that these models experience all the time are on another level. The hallucinations, sycophancy, reward hacking, etc can be hilariously inept.

IMO, that should inform you enough to not trust these services (as they exist today) in explaining concepts to you that you have no idea about.

If you are so certain you are okay to trust these things, you should evaluate every assertion it makes for, say, 40 hours of use, and count the error rate. I would say it is above 30%, in my experience of using language models day to day. And that is with applied tasks they are considered "good" at.

If you are okay with learning new topics where even 10% of the instruction is wrong, have fun.

Re: Study mode

#485

Post by cindyllrn »

breve wrote: post #455
mediaman wrote: post #454

Earlier quoted context omitted.

This is endlessly brought up as if the human operating the tool is an idiot. I agree that if the user is incompetent, cannot learn, and cannot learn to use a tool, then they're going to make a lot of mistakes from using GPTs. Yes, there are limitations to using GPTs. They are pre-trained, so of course they're not going to know about some easter egg in DDG. They are not an oracle. There is indeed skill to using them.…

The ability to say "I don't know" is not a high bar. I would say it's a basic requirement of a system that is not magic.

I see your point

Some of the best exchanges that I participated in or witnessed involved people acknowledging their personal limits, including limits of conclusions formed a priori

To further the discussion, hearing the phrase you mentioned would help the listener to independently assess a level of confidence or belief of the exchange

But then again, honesty isn't on-brand for startups

It's something that established companies say about themselves to differentiate from competitors or even past behavior of their own

I mean, if someone prompted an llm weighted for honesty, who would pay for the following conversation?

Prompt: can the plan as explained work?

Response: I don't know about that. What I do know is on average, you're FUCKED.

Re: Study mode

#486

Post by no_wizard »

chpatrick wrote: post #475
no_wizard wrote: post #440

Earlier quoted context omitted.

Pattern matching has a definition in this field, it does mean specific things. We know machine learning has excelled at this in greater and greater capacities over the last decade The other part of this is weighted filtering given a set of rules, which is a simple analogy to how AlphaGo did its thing. Dismissing all this as vague is effectively doing the same thing as you are saying others do. This technology has lim…

> Pattern matching has a definition in this field, it does mean specific things. Such as? > They’re fundamentally wholly understandable systems that work on a consistent level in terms of the how they do what they do (that is separate from the actual produced output) Multi billion parameter models are definitely not wholly understandable and I don't think any AI researcher would claim otherwise. We can train them but…

You’re welcoming to provide counters. I think these are all sufficiently common things that they stand on their own as to what I posit

Re: Study mode

#487

Post by beefnugs »

I have no evidence of this but: I think this is the ultimate scam?

human: damn kids are using this to cheat in school

openai: release an "app"/prompt that seems really close to solving this stated problem

kids: I never wanted to learn anything, I just want to do bare minimum to get my degree, let my parents think they are helping my future, and then i can get back to ripping that bong

It doesn't matter the problem statement: the 80% or less solution seems can be made and rather quickly. Such a huge percentage of the population judges technology solutions as "good enough" way lower than they should. This is even roping in people from the past who used to be a higher level of "rigorous correctness" because they keep thinking, "damn just a bit more work and it will get infinity better, lets create the biggest economic house of cards this world will ever collapse under"

Re: Study mode

#488

Post by ay »

romaniitedomum wrote: post #412
jacobedawson wrote: post #379

An 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…

My favourite story of that involved attempting to use LLM to figure out whether it was true or my hallucination that the tidal waves were higher in Canary Islands than in Caribbean, and why; it spewed several paragraphs of plausibly sounding prose, and finished with “because Canary Islands are to the west of the equator”.

This phrase is now an inner joke used as a reply to someone quoting LLMs info as “facts”.

Re: Study mode

#489

Post by Avicebron »

ramraj07 wrote: post #441
romaniitedomum wrote: post #412

Earlier 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.

OpenAI's o3/40 models completely spun out when I was trying to write a tiny little TUI with ratatui, couldn't handle writing a render function. No idea why, spent like 15 minutes trying to get it to work, ended up pulling up the docs..

I haven't spent any money with claude on this project and realistically it's not worth it, but I've run into little things like that a fair amount.

Re: Study mode

#490

Post by tekno45 »

r3trohack3r wrote: post #458
tekno45 wrote: post #444

Earlier quoted context omitted.

How do you know? its literally non-deterministic.

Most (all?) AI models I work with are literally deterministic. If you give it the same exact input, you get the same exact output every single time. What most people call “non-deterministic” in AI is that one of those inputs is a _seed_ that is sourced from a PRNG because getting a different answer every time is considered a feature for most use cases. Edit: I’m trying to imagine how you could get a non-deterministic…

So by default. Its non-deterministic for all non power users.
Post reply on HN