Live data from Hacker News

I'd rather read the prompt

claytonwramsey.com

761–770 of 880 posts

Re: I'd rather read the prompt

#761
post #605
post #318

Earlier quoted context omitted.

We were talking about writing, not about vibe coding. No one said anything about vibe coding. Using tools appropriately to accomplish tasks more quickly is just common sense. Deliberately choosing to pay 10x the cost for the same or equivalent output isn't a rational business decision, regardless of whether the task happens to be writing, long division, or anything else. Just to be clear, I'm not arguing against doin…

Don't get me wrong: I don't deny that LLMs can help tricking other humans into believing that the text is more professional than it actually is. LLMs are engineered exactly for that. I'd be curious to know whether your legal documents are as good as without LLMs. I wouldn't be surprised at all if they were worse, but cheaper. Talking about security audits, that's actually a problem I've seen: LLMs makes it harder to…

There's no "tricking" involved, and no basis for your assumption that LLMs lower the quality of work. I would suggest that what you and the author are observing is actually the opposite effect: LLMs broadly help improve the quality of work, all else being equal. The caveat is that when all else is not equal, this manifests in bad work being improved to a level that's still bad. The issue here is students using advanced tooling as an excuse to be lazy and undercut their own learning process, not the tool itself. LLMs are just this generation's version of Wikipedia and spell check.

As much as the author rightfully complains about the example in the post, a version that only said "explain the downsides of Euler angles in robotics and suggest some alternatives" would obviously be far worse. In this case, the AI helped elevate clear F-level work to maybe a C. That's not an indictment of AI; it's an indictment of low-quality work. LLMs lower the bar to produce passable-looking bad work, but they also lower the bar to produce excellent work. The confirmation bias here is that we don't know how many cases of B-level work became A papers with AI assistance, because those instances don't stand out in the same way.

In the audit example, LLMs aren't doing the audit. They synthesize my notes into a useful starting point to nullify writer's block, and let me focus more of my time on the hard or unique aspects of a given report. It's like having an intern write the first draft for me, typically with some mistakes or oversights, occasionally with a valuable additional insight thrown in, and often with links to a few helpful references for the customer that I wouldn't necessarily have found and included on my own. That doesn't lower the quality; it improves it.

As far as the legal example, it really depends on the complexity of a given instance and the guidance you've provided to your lawyers. A good lawyer won't sign off on something that fails to meet the requested quality bar (if anything, the financial incentive would be for them to err on the side of conservatism and toss out the draft you'd provided). But of course this all depends on you having a clear enough understanding of what you're trying to accomplish, and enough familiarity with legal documents and proficiency with language to shape everything into a passable first draft. AI speeds this up, but if you don't know what you're doing then the AI won't solve that for you. It's a tool like any other, and can be used properly or improperly.

Re: I'd rather read the prompt

#762

Earlier quoted context omitted.

I'm teaching faculty at a university and, at least where it comes to lecture courses, I don't find this a bubble. There are definitely plenty of students who would do just fine, if not even better, without the external discipline and structure. And even more those who could do it with something like a MOOC or just a posted curriculum. And to be able to work without an external discipline is IMHO one of the main learn…

Those students do exist, but there are an exceedingly small minority. Of course, they won't tell you this, because everyone likes to believe they're self-motivated. Most people just aren't. The simple question to ask is, when you go home, what do you do? If the answer is learn how to sew or work on your project car you've had for 10 months, you can probably learn on your own. If your answer is watch TV, play video ga…

There are plenty of students studying out of interest. It's very common to e.g. study excess courses. Many doing their theses can work very independently and go way beyond what's required.

Highly self-driven students are a minority, but not a rarity. People do things out of being interested and enjoying learning. It shouldn't be a surprise in a website called hacker news.

Re: I'd rather read the prompt

#763
post #275

Earlier quoted context omitted.

> I'm there for the degree. If I wanted to learn and engage with material, I could save $60,000 I would argue that if it costs $60,000, both your education system and the recruitment in those companies that require this degree are broken. It's not the case in all countries though. Not that it is your fault, just stating the obvious.

My tuition was $35k a year 25 years ago. Just checked, and now it is $60k a year. Before room and board. Broken? Saddling individuals with a quarter million in debt when they are just starting life is absolutely broken. That they must indenture to be a modern professional (and buy hope for at least a middle class landing) is broken. The notion that everything must return a (generally, near-term) accounting profit is…

$35k in 2000 is $65k now so that's actually decreased in real terms.

Re: I'd rather read the prompt

#764
post #753

Earlier quoted context omitted.

I think software developers might be somewhat of an outlier. Industry wants good programmers but universities teach computer science which really should be called "computation science". Much of what we learn in university will hardly ever be used while many practical skills are at best learned as a side effect. Dijkstra favorite said that computer science is as much about computers as astronomy is about telescopes. S…

> while many practical skills are at best learned as a side effect. I strongly disagree, that’s the intent not a side effect. It’s IMO a common misconception that early algorithm classes are just designed around learning algorithms. Instead basic algorithms are the simplest thing to turn abstract requirements into complex code. The overwhelming majority of what students learn are the actual tools of programming, debu…

That is true, in the sense that the 100/200 level classes are covering programming basics in addition to whatever algorithmic theory is being presented. But beyond that that, programs really seem to differ pretty strongly on applied projects and software engineering practices (basic stuff like source control) and more theoretical/mathematical concepts. One type of capstone style class commonly seen is compiler design. To a certain extent, a good school will teach you how to learn, and give you enough of a background, class projects, internships, electives with applied options, that you get a well rounded education and can quickly ramp up in a more typical software organization after graduation. But as someone who has hired many new grads over the years, it always surprises me what sort of gaps exist. It rarely is about programming basics, and almost always about "software engineering" as a discipline.

Re: I'd rather read the prompt

#765
post #255

Earlier quoted context omitted.

You're missing trees for forest. When I was a kid and got an assignement for writing an essey about "why good forces prevailed in Lords of the Rings" as a gate check to see if I actually read the novel I had three choices: (a) read the novel and write the essey myself (b) find an already written essey - not an easy task in pre-internet era but we had books with esseys on most common topics you could "copy-paste" - an…

> Today I can ask ChatGPT to write me a 4 pages essay about a novel I've never heard of and call it a day. There's no value gained in the process. If we take the original article at face value, no you can't do that. ChatGPT will apparently produce something that is obviously ChatGPT produced and fail to fool even the most absent minded of instructors that you have read the material. So even with a ChatGPT LLM to help…

> If we take the original article at face value, no you can't do that. ChatGPT will apparently produce something that is obviously ChatGPT produced and fail to fool even the most absent minded of instructors that you have read the material.

Only if you don't have any custom instructions about style and don't proofread it afterwards. All the usual "tells" of ChatGPT are very obvious to scrub out, and you don't have to use OpenAI's chat wrapper to begin with.

Re: I'd rather read the prompt

#766

Ever since AI came out I’ve been talking about the prompt to output ratio. We naturally assume that the prompt will be smaller than the output just because of the particulars of the systems we use, but as you get more and more particular of what you want, the prompt grows while the output stays the same size. This is logical. If instead of writing an essay, I just describe what I want the essay to say, the descriptio…

This is why it’s unlikely these systems will effectively replace software development. By the time you’ve specified the novel system you want to build well enough in English such that you get exactly the system you want you might as well have written the code.

This is simply not true.

I can describe a novel physics model for a video game. I can do a refresher on concepts like friction, air resistance, gravity, etc. that I don't remember well from school. Then I can describe the constraints and generate code to satisfy it.

If I were to go and learn the physics really in depth and then code it myself, it would take 10x longer.

Re: I'd rather read the prompt

#767
post #221

> I should hope that the purpose of a class writing exercise is not to create an artifact of text but force the student to think; a language model produces the former, not the latter. It's been incredibly blackpilling seeing how many intelligent professionals and academics don't understand this, especially in education and academia. They see work as the mere production of output, without ever thinking about how that…

I’ve come across this analogy that I think works well: Using an LLM to do schoolwork is like taking a forklift to the gym. If all we were interested in was moving the weights around, you’d be right to use a tool to help you. But we’re doing this work for the effect it will have on you . The reason a teacher asks you a question is not because they don’t know the answer.

I love this analogy because it's also not a waste of time to learn how to use a forklift!

But the gym isn't the best place to engage in forklift training. And you engage in forklift training at the gym, expect to learn how to use a forklift to lift gym weights. Don't expect to also get the benefits that the gym is designed to impart.

Re: I'd rather read the prompt

#768
post #6

I used to teach, years before LLMs, and got lots of copy-pasted crap submitted. I always marked it zero, never mentioning plagiarism (which would require some university administration) and just commenting that I asked for X and instead got some pasted together nonsense. As long as LLM output is what it is, there is little threat of it actually being competitive on assignments. If students are attentive enough to par…

When I was kid in school I would write original essays, and I mean truly original creative ideas. But of course any new idea has a chance of failure, so these essays were mostly bad and got bad grades. At a loss for what to do I quickly stopped reading the books I was assigned basing my essays on Wikipedia summaries and other people’s reviews. I saw my first few As and even A+s and I realized if I write something ori…

Or you're just a bad writer. I certainly could not understand your main point, particularly the sentence "For an idea to truly be considered intelligent in literature it has to be appealing to people have no actual memory of the things they’ve read." which is ungrammatical.

Re: I'd rather read the prompt

#769
post #657

Earlier quoted context omitted.

I think you missed the parent's point. Maybe consider this: Why don't employers recognize the credentials of a MOOC to the same degree that they would a university degree? We could similarly ask why employers value the degrees of some universities more than others. I think it's important to realize that ultimately the decisions come from the employers, not the universities. No one is making the employers do anything.…

I brought up that I think the root problem is employers in a different comment, so I’m not missing that point.[1] The distinction is that I don’t think it’s a dichotomous problem. While colleges may not be the root problem, they can still be a contributor to the problem. (Besides, if an employer is to recognize a MOOC, that course completion has to be documented, which means it’s just another version of credential) […

I was responding to the comment you wrote. But I also disagree with the other one.

It is a confounding variable[0]. The problem with trying to go after the confounding variable is you 1) don't solve the problem after fixing it 2) let's the current negative feedback loop continue growing.

[0] https://en.wikipedia.org/wiki/Confounding

Re: I'd rather read the prompt

#770
I think AI can be an amazing tool that can help us learn even better when used correctly and when not used as a substitute for learning and understanding.

It can be used as a personal tutor. How awesome is it to have a tutor always available to answer almost any question from any angle to really help you understand? Yes, AI won't get everything right 100%, but for students who are still learning basics, it's fair to assume that having an AI tutor can yield far better results than having no tutor at all.

It can also be used as a tool for doing mundane work, so you can focus more on the interesting and creative work. Kind of like a calculator or a spreadsheet. Would math majors become better mathematicians if they had to do all calculations by hand?

I think instead of banning AI, education needs to reform. Teaching staff should focus less time on giving lectures and grading papers (those things can be recorded and automated) and more time on ORAL EXAMS where they really probe student's knowledge and there's no possibility of cheating.

Students can and should use AI to help them prepare. E.g. don't ask AI to write an essay for you, write it yourself and ask it to critique it. Don't ask it to give you answers for a test, ask it to ask you questions on the topic and find gaps in your knowledge. Etc.

Post reply on HN