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

Learning to Reason with LLMs

openai.com

Post reply on HN
Print view View original

471–480 of 1001 posts

Re: Learning to Reason with LLMs

#471

Post by candiddevmike »

cs391231 wrote: post #437
hakanderyal wrote: post #418

Earlier quoted context omitted.

Software engineering contains a lot more than just writing code. If we somehow get AGI, it'll change everything, not just SWE. If not, my belief is that there will be a lot more demand for good SWEs to harness the power of LLMs, not less. Use them to get better at it faster.

This thing is doing planning and ascending the task management ladder. It's not just spitting out code anymore.

Management will be easier to replace than SWEs. I'm thinking there will come a time, similar to the show Mrs Davis, where AI will direct human efforts within organizations. AI will understand its limits and create tasks/requirements for human specialists to handle.

Re: Learning to Reason with LLMs

#472

Post by bartman »

romeros wrote: post #428
bartman wrote: post #367

This is incredible. In April I used the standard GPT-4 model via ChatGPT to help me reverse engineer the binary bluetooth protocol used by my kitchen fan to integrate it into Home Assistant. It was helpful in a rubber duck way, but could not determine the pattern used to transmit the remaining runtime of the fan in a certain mode. Initial prompt here [0] I pasted the same prompt into o1-preview and o1-mini and both c…

is it better than Claude?

Neither Sonnet nor Opus could solve it or get close in a minimal test I did just now, using the same prompt as above.

Sonnet: https://pastebin.com/24QG3JkN

Opus: https://pastebin.com/PJM99pdy

Re: Learning to Reason with LLMs

#473

Post by al_borland »

cs391231 wrote: post #408

Student here. Can someone give me one reason why I should continue in software engineering that isn't denial and hopium?

The calculator didn’t eliminate math majors. Excel and accounting software didn’t eliminate accountants and CPAs. These are all just tools.

I spend very little of my overall time at work actually coding. It’s a nice treat when I get a day where that’s all I do.

From my limited work with Copilot so far, the user still needs to know what they’re doing. I have 0 faith a product owner, without a coding background, can use AI to release new products and updates while firing their whole dev team.

When I say most of my time isn’t spent coding, a lot of that time is spend trying to figure out what people want me to build. They don’t know. They might have a general idea, but don’t know details and can’t articulate any of it. If they can’t tell me, I’m not sure how they will tell an LLM. I ended up building what I assume they want, then we go from there. I also add a lot of stuff that they don’t think about or care about, but will be needed later so we can actually support it.

If you were to go in another direction, what would it be where AI wouldn’t be a threat? The first thing that comes to my mind is switching to a trade school and learning some skills that would be difficult for robots.

Re: Learning to Reason with LLMs

#474

Post by novaleaf »

boo, they are hiding the chain of thought from user output (the great improvement here)

> Therefore, after weighing multiple factors including user experience, competitive advantage, and the option to pursue the chain of thought monitoring, we have decided not to show the raw chains of thought to users. We acknowledge this decision has disadvantages. We strive to partially make up for it by teaching the model to reproduce any useful ideas from the chain of thought in the answer. For the o1 model series we show a model-generated summary of the chain of thought.

Re: Learning to Reason with LLMs

#475

Post by Bjorkbat »

idiliv wrote: post #279

In the demo, O1 implements an incorrect version of the "squirrel finder" game? The instructions state that the squirrel icon should spawn after three seconds, yet it spawns immediately in the first game (also noted by the guy doing the demo). Edit: I'm referring to the demo video here: https://openai.com/index/introducing-openai-o1-preview/

Yeah, now that you mention it I also see that. It was clearly meant to spawn after 3 seconds. Seems on successive attempts it also doesn't quite wait 3 seconds.

I'm kind of curious if they did a little bit of editing on that one. Almost seems like the time it takes for the squirrel to spawn is random.

Re: Learning to Reason with LLMs

#477

Post by bartman »

GaggiX wrote: post #465
bartman wrote: post #367

This is incredible. In April I used the standard GPT-4 model via ChatGPT to help me reverse engineer the binary bluetooth protocol used by my kitchen fan to integrate it into Home Assistant. It was helpful in a rubber duck way, but could not determine the pattern used to transmit the remaining runtime of the fan in a certain mode. Initial prompt here [0] I pasted the same prompt into o1-preview and o1-mini and both c…

Did you edit the message? I cannot see anything now in the screenshot, too low resolution

You need to click on the image for the high res version to load. Sorry, it’s awkward.

Re: Learning to Reason with LLMs

#478

Post by ljm »

crop_rotation wrote: post #434
cs391231 wrote: post #408

Student here. Can someone give me one reason why I should continue in software engineering that isn't denial and hopium?

If you have better career ideas, you should not continue. The thing is it is very hard to predict how the world will change (and by how much from very little to a revolutionary change) with all these new changes. Only licensed and regulated professions (doctors/lawyers/pilots etc) might remain high earning for long (and they too are not guaranteed). It really is worth a relook on what you want to do in life while see…

This is pretty extreme advice to offer in response to news that a model that can better understand programming problems is coming out.

In fact, it's more encouragement to continue. A lot of issues we face as programmers are a result of poor, inaccurate, or non-existent documentation, and despite their many faults and hallucinations LLMs are providing something that Google and Stack Overflow have stopped being good at.

The idea that AI will replace your job, so it's not worth establishing a career in the field, is total FUD.

Re: Learning to Reason with LLMs

#479

Post by spaceman_2020 »

canjobear wrote: post #427

First shot, I gave it a medium-difficulty math problem, something I actually wanted the answer to (derive the KL divergence between two Laplace distributions). It thought for a long time, and still got it wrong, producing a plausible but wrong answer. After some prodding, it revised itself and then got it wrong again. I still feel that I can't rely on these systems.

Look where you were 3 years ago, and where you are now.

And then imagine where you will be in 5 more years.

If it can almost get a complex problem right now, I'm dead sure it will get it correct within 5 years

Re: Learning to Reason with LLMs

#480

Post by al_borland »

LouisSayers wrote: post #431
cs391231 wrote: post #408

Student here. Can someone give me one reason why I should continue in software engineering that isn't denial and hopium?

Don't do it, help us keep our high salaries :D Joking aside, even with AI generating code, someone has to know how to talk to it, how to understand the output, and know what to do with it. AI is also not great for novel concepts and may not fully get what's happening when a bug occurs. Remember, it's just a tool at the end of the day.

> may not fully get what's happening when a bug occurs.

And may still not understand even when you explicitly tell it. It wrote some code for me last week and made an error with an index off by 1. It had set the index to 1, then later was assuming a 0 index. I specifically told it this and it was unable to fix it. It was in debug hell, adding print statements everywhere. I eventually fixed it myself after it was clear it was going to get hung up on this forever.

It got me 99% of the way there, but that 1% meant it didn’t work at all.

Post reply on HN