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

GitHub Copilot is now available for free

github.com

Post reply on HN
Print view View original

381–390 of 412 posts

Re: GitHub Copilot is now available for free

#381

Post by pabs3 »

_heimdall wrote: post #380
fragmede wrote: post #374

Earlier quoted context omitted.

The license aside, the question is what can be done with a carefully arranged blob of binary? Without additional software (Windows) I can't really do anything with Photoshop.exe. Similarly, Llama.gguf is either useful, with Ollama.app, or not, standing alone. So (looking past the difference in license), would you consider Photoshop.exe similar in that it's a binary blob that's useless by itself, or is it a useful col…

The license used isn't important in my opinion, when talking about open source the question is whether the source code is available to be modified and reviewed/interpreted. Photoshop, or any compiled binary, isn't meant to be open source and the code isn't meant to be reviewable. Llama is called open source, though the most important part isn't publicly available for review. If llama didn't claim to be open source I…

The source data for the training needs to be public and freely licensed too, otherwise its IMO not an open source model.

Re: GitHub Copilot is now available for free

#382

Post by Wowfunhappy »

fragmede wrote: post #374
_heimdall wrote: post #365

Earlier quoted context omitted.

Photoshop doesn't claim to be open source like llama does though, I'm not sure of the connection you're making. Don't get me wrong, llama is at least more open than OpenAI and that may be meaningful.

The license aside, the question is what can be done with a carefully arranged blob of binary? Without additional software (Windows) I can't really do anything with Photoshop.exe. Similarly, Llama.gguf is either useful, with Ollama.app, or not, standing alone. So (looking past the difference in license), would you consider Photoshop.exe similar in that it's a binary blob that's useless by itself, or is it a useful col…

I think this discussion is silly in the context of a modern LLM. Nobody really understands how an LLM works, and you absolutely do not actually want to retrain Llama from scratch.

When I said "it's not really open source", I was referring to the fact that there are restrictions on who can use Llama.

Re: GitHub Copilot is now available for free

#383

Post by thot_experiment »

int_19h wrote: post #354
indigodaddy wrote: post #339

Earlier quoted context omitted.

Thanks, I need to figure out something to run on M4 Mac Mini standard 16G

You should be looking at 7-8b sized models, then. https://huggingface.co/Qwen/Qwen2.5-Coder-7B-Instruct is considered pretty strong for its size. That said, you shouldn't expect more than glorified autocompletion at that point.

I'm not familiar with the practical memory requirements on Macs but I suspect that with 16gb of integrated ram you won't have issues running 14B models even at q6_k, certainly would be fine at q4 and it's definitely going to be capable of writing code based on instruction as well as minor refactoring, generating docstrings etc.

Re: GitHub Copilot is now available for free

#384

Post by outworlder »

uludag wrote: post #28

I feel like this free plan is just enough to get someone hooked and require them to upgrade to paid. Like, imagine GPS navigation wasn't widespread and there was a paid service that gave you 20 free trips. Eventually your normal navigation skills would atrophy and you'd be obliged to purchase.

> imagine GPS navigation wasn't widespread and there was a paid service

GPS is something that wouldn't be invented these days. Instead of the US military footing the bill, it would be some private company somewhere.

Re: GitHub Copilot is now available for free

#385

Post by outworlder »

TranquilMarmot wrote: post #305
beepbooptheory wrote: post #64

Working with people who use this stuff a lot has made my current job just so so much harder in every way its astonishing. I used to solve problems with code, now I feel like hermeneut or dream analyzer: absent of human intention, codebases quickly become these weird piles of different idioms, even without considering the hallucinations (those have definitely cost me a few sleepless nights now either way). But I am ju…

Where I work, we have a directive to use AI as much as we can wherever possible. I was handed a codebase that's been working on by many, many, many different people over the past ~2 years and was written almost entirely with AI (starting with GPT 3) The only way you can deal with the codebase is to fully embrace the AI. Whenever I want to make anything beyond a simple API change, I have to boot up Cursor and give it…

This is how I think the current bubble will pop. Yes, these are useful tools that we just now are trying to learn how to use. But wallstreet and bean counters are going apeshit on the prospect of replacing the (expensive!) humans they currently pay for.

Once the codebases become an unmanageable mess I think the pendulum will swing back, hard.

That should buy sometime for anyone not entering the industry right now.

Re: GitHub Copilot is now available for free

#386

Post by int_19h »

thot_experiment wrote: post #383
int_19h wrote: post #354

Earlier quoted context omitted.

You should be looking at 7-8b sized models, then. https://huggingface.co/Qwen/Qwen2.5-Coder-7B-Instruct is considered pretty strong for its size. That said, you shouldn't expect more than glorified autocompletion at that point.

I'm not familiar with the practical memory requirements on Macs but I suspect that with 16gb of integrated ram you won't have issues running 14B models even at q6_k, certainly would be fine at q4 and it's definitely going to be capable of writing code based on instruction as well as minor refactoring, generating docstrings etc.

The model itself will fit just fine, of course, but you'll also want a large context for coding. And then since it's integrated RAM, it's also used by everything else running on your system - like, say, your IDE, your compiler etc, which are all fairly memory hungry.

Also keep in mind that, even though it's "unified memory", the OS enforces a certain quota for the GPU. If I remember correctly, it's something like 2/3 of the overall RAM.

Re: GitHub Copilot is now available for free

#387

Post by jodrellblank »

nunez wrote: post #358
hkaal wrote: post #152

Earlier quoted context omitted.

It is part of the great plan: - Buy GitHub and devalue all individual projects by soft-forcing most big projects to go there and lose their branding. - Gamify and make "development" addictive. - Use social cliques who hype up each other's useless code and PRs and ban critics. - Make kLOC and plagiarism great again! This all happened before "AI". "AI" is the last logical step that will finally destroy open source, as…

Don't know why this was downvoted. This is an interesting take that I, at least, haven't seen before.

A good conspiracy can be interesting even if it's false, so I don't criticise you for finding it interesting; but I downvoted because I didn't - HN is chock full of "Microsoft bad, Embrace Extend Extinguish" and it's fucking tiresome. I wrote 1700 words on how that comment makes no sense but I'll reduce it:

- Have you ever read a PR and thought "this code is useless" and the result was you deciding that "kLOC is great"? Any way I put those things (Microsoft, kLOC, AI, Github, social cliques,...) together I don't get anything sensible; Microsoft spent 7.5Bn on Github to make kLOC great to help them destroy open source? It's a crackpot word salad, not a reasoned position. At least, if it's a reasoned position they should post the reasoning.

- Github has 200,000,000+ public repositories and makes $1Bn/year revenue. How will putting AI into Github 'finally' destroy open source and why does Microsoft want to screw up a revenue stream roughly the size of Squarespace's, bigger than DigitalOcean's or Netgear's, and getting on for as big as CloudFlare's?

Re: GitHub Copilot is now available for free

#389

Post by weikju »

int_19h wrote: post #351
weikju wrote: post #331

Earlier quoted context omitted.

Because the rich and powerful people who will reap the most benefit from all the automation will not redistribute the wealth to the now-useless ex-labor force.

Oh, they will. The only question is whether this will happen sooner and voluntarily, or later with torches and pitchforks.

Better bring out the pitchforks earlier while we still have access to them!!

Re: GitHub Copilot is now available for free

#390

Post by AlienRobot »

8n4vidtmkvmk wrote: post #292
AlienRobot wrote: post #245

Earlier quoted context omitted.

"Useful contributions" is subjective. Not everyone is born a senior developer. Juniors, and even children who aren't even juniors yet ask questions on these channels. Source: I bothered a lot of people on the Internet about C++ when I was child.

Are you saying these child questions are useful contributions?

To the child that asked it? Absolutely. :-)

Information is only useful if it's accessible. If they are asking questions it's because the information they want is in practice not accessible to them.

Post reply on HN