Learning to Reason with LLMs • Hacker News Board

Re: Learning to Reason with LLMs

#491

Post by sbaidon94 » Thu, Sep 12, 2024, 7:45 PM UTC

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?

My two cents thinking about different scenarios:

- AI comes fast, there is nothing you can do: Honestly, AI can already handle a lot of tasks faster, cheaper, and sometimes better. It’s not something you can avoid or outpace. So if you want to stick with software engineering, do it because you genuinely enjoy it, not because you think it’s safe. Otherwise, it might be worth considering fields where AI struggles or is just not compatible. (people will still want some sort of human element in certain areas).

- There is some sort of ceiling, gives you more time to act: There’s a chance AI hits some kind of wall that’s due to technical problems, ethical concerns, or society pushing back. If that happens, we’re all back on more even ground and you can take advantage of AI tools to improve yourself.

My overall advice; and it will probably be called out as cliche/simplistic just follow what you love, just the fact that you have an opportunity to pursue to study anything at all is something that many people don't have. We don't really have control in a lot of stuff that happens around us and that's okay.

Re: Learning to Reason with LLMs

#492

Post by exe34 » Thu, Sep 12, 2024, 7:45 PM UTC

chairhairair wrote: post #289

RandomLensman wrote: post #218

How could it fail to solve some maths problems if it has a method for reasoning through things?

Simple questions like this are not welcomed by LLM hype sellers. The word "reasoning" is being used heavily in this announcement, but with an intentional corruption of the normal meaning. The models are amazing but they are fundamentally not "reasoning" in a way we'd expect a normal human to. This is not a "distinction without a difference". You still CANNOT rely on the outputs of these models in the same way you can…

it depends who's doing the simple reasoning. Richard Feynman? yes. Donald Trump? no.

Re: Learning to Reason with LLMs

#493

Post by crop_rotation » Thu, Sep 12, 2024, 7:45 PM UTC

ljm wrote: post #478

crop_rotation wrote: post #434

Earlier quoted context omitted.

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 advice is unrelated to the model and related to the last year's worth of development. In any case I am advising a relook which is perfectly warranted for anyone pre-university or in university.

Re: Learning to Reason with LLMs

#494

Post by rvz » Thu, Sep 12, 2024, 7:45 PM UTC

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?

Unlike the replies here I will be very honest with my answer. There will be less engineers getting hired as the low hanging fruit has already been picked and automated away.

It is not too late. These LLMs still need very specialist software engineers that are doing tasks that are cutting edge and undocumented. As others said Software Engineering is not just about coding. At the end of the day, someone needs to architect the next AI model or design a more efficient way to train an AI model.

If I were in your position again, I now have a clear choice of which industries are safe against AI (and benefit software engineers) AND which ones NOT to get into (and are unsafe to software engineers):

Do:

   - SRE (Site Reliability Engineer)

   - Social Networks (Data Engineer)

   - AI (Compiler Engineer, Researcher, Benchmarking)

   - Financial Services (HFT, Analyst, Security)

   - Safety Critical Industries (defense, healthcare, legal, transportation systems)

Don't:

   - Tech Writer / Journalist

   - DevTools

   - Prompt Engineer

   - VFX Artist

The choice is yours.

Re: Learning to Reason with LLMs

#495

Post by magicalhippo » Thu, Sep 12, 2024, 7:45 PM UTC

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're just there to churn out code, then yeah, perhaps find something else.

But if you're there to improve your creativity and critical thinking skills, then I don't think those will be in short supply anytime soon.

The most valuable thing I do at my job is seldom actually writing code. It's listening to customer needs, understanding the domain, understanding our code-base and it's limitations and possibilities, and then finding solutions that optimize certain aspects be it robustness, time to delivery or something else.

Re: Learning to Reason with LLMs

#496

Post by fsloth » Thu, Sep 12, 2024, 7:46 PM UTC

zamadatix wrote: post #405

gibsonf1 wrote: post #366

Yes, but it will hallucinate like all other LLM tech making it fully unreliable for anything mission critical. You literally need to know the answer to validate the output, because if you don't, you won't know if output is true or false or in between.

You need to know how to validate the answer to your level of confidence, not necessarily already have the answer to compare itself. In some cases this is the same task or (close enough to) that it's not a useful difference, in other cases the two aren't even from the same planet.

This. There are tasks where implementing something might take up to one hour yourself, that you can validate with high enough confidence in a few seconds to minutes.

Of course not all tasks are like that.

Re: Learning to Reason with LLMs

#497

Post by hmottestad » Thu, Sep 12, 2024, 7:47 PM UTC

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?

I think this new model is a generational leap above Claude for tasks that require complex reasoning.

Re: Learning to Reason with LLMs

#498

Post by baq » Thu, Sep 12, 2024, 7:47 PM UTC

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’re any good at SWE with a sprinkle of math and CS, your advantage will get multiplied by anywhere from 2 to 100x if you use the leverage of co-intelligence correctly. Things that took weeks before now easily take hours, so if you know what to build and especially what not to build (including but not limited to confabulations of models), you’ll do well.

Re: Learning to Reason with LLMs

#499

Post by ponector » Thu, Sep 12, 2024, 7:49 PM UTC

al_borland wrote: post #473

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 us…

But excel eliminated need in multiple accountants. One accountant with excel replaced ten with paper.

Chatgpt already eliminated many entry-level jobs like writer or illustrator. Instead of hiring multiple teams of developers, there will be one team with few seniors and multiple AI coding tools.

Guess how depressing to the IT salaries it will be?

Re: Learning to Reason with LLMs

#500

Post by Gee101 » Thu, Sep 12, 2024, 7:49 PM UTC

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?

I'm wondering if the opposite might happen, that there will be more need for software engineers.

1. AI will suck up a bunch of engineers to run, maintain and build on its own.

2. Ai will open new fields that is not yet dominated by software. Ie. Driving ect.

3. Ai tools will lower the bar for creating software meaning industries that weren't financially viable will now become viable for software automation.