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

GitHub Copilot is generally available

github.blog

Post reply on HN
Print view View original

411–420 of 796 posts

Re: GitHub Copilot is generally available

#411

Post by bee_rider »

causi wrote: post #278
bastardoperator wrote: post #234

Earlier quoted context omitted.

Or the fact that you grant GitHub an implicit license as outlined in the ToS.

GitHub isn't liable. That's been established in court with regards to training AIs. Who is liable is you who may or may not have the legal right to use the code CoPilot spits out for you.

It seems like this space will open up all sorts of interesting novel legal questions.

It is possible to provide CoPilot with a sequence of inputs that produces some of the input, which was copyrighted. Let's say you want to help people violate copyright, so you as a third party distribute a script that provides that sequence of inputs. Who's violating the copyright there?

Alternatively -- it is apparently legal to produce a clean-room implementation that duplicates a copyright implementation. Supposing you were to use a tool like CoPilot, which has just been trained on that copyright implementation. Is your room still clean? You might even be able to get it to spit out identical functions!

Or, if you have a ML algorithm which has been trained on leaked closed source code, and it is sufficiently over-fitted as to just provide the source code given the filename or the original binary, who is violating copyright when this tool is used? If it is just the end user, then this seems like a really convenient way to launder leaked closed source code.

Re: GitHub Copilot is generally available

#412

Post by CapsAdmin »

I really like copilot, but my outside of the content being generated it still feels a bit slow and somewhat hacked into vscode. It sometimes interferes with regular "intellisense suggestions" as well.

I've been in the beta since almost the beginning I have not really seen much improvement on the frontend side. Since its release, the changelog only mentions 10 small (or so it seems) improvements

https://marketplace.visualstudio.com/items/GitHub.copilot-ni...

On the backend side, I feel like I've started to "figure out" copilot a little bit. One thing I'd like to see is inline completion which I think gpt3 can do now but copilot which I believe it's based on cannot.

I think I will pay to continue, but I'd like to see some frontend improvements and maybe some backend alternatives. Ideally I'd love this to be open source but compute power doesn't seem feasible (?) unless we start magically crowd sourcing our computers to run a model somehow.

Re: GitHub Copilot is generally available

#413

Post by datatrashfire »

paskozdilar wrote: post #345
omginternets wrote: post #320

Earlier quoted context omitted.

I wholeheartedly agree with your analysis, but feel like it’s ignoring the elephant in the room: writing code is not the bottleneck in need of optimization. Conceiving the solution is. Any time “saved” through Copilot and it’s ilk is immediately nullified by having to check it’s correctness. From there, the problem is worsened by the Frankensteinesque stitching together of disparate parts that you describe. I can’t i…

Can Copilot write tests? That way it could test its own code and tweak it until it works. Of course, one would then ask how to verify tests. I suppose Copilot could write meta-tests - tests that verify other tests. That way it could test its own tests and tweak them until they work. Of course, one would then ask how to verify meta-tests. I suppose Copilot could write meta-meta-tests - tests that verify meta-tests. Th…

I can almost envision a future where human devs write tests, code generating frameworks build code from a spec.

Re: GitHub Copilot is generally available

#414

Post by paskozdilar »

datatrashfire wrote: post #413
paskozdilar wrote: post #345

Earlier quoted context omitted.

Can Copilot write tests? That way it could test its own code and tweak it until it works. Of course, one would then ask how to verify tests. I suppose Copilot could write meta-tests - tests that verify other tests. That way it could test its own tests and tweak them until they work. Of course, one would then ask how to verify meta-tests. I suppose Copilot could write meta-meta-tests - tests that verify meta-tests. Th…

I can almost envision a future where human devs write tests, code generating frameworks build code from a spec.

I think that's the underlying idea of Logic Programming.

Re: GitHub Copilot is generally available

#415

Post by theshrike79 »

jmkni wrote: post #306
Hamcha wrote: post #195

I've been using Copilot non-stop on every hobby project I have ever since they've let me in (2021/07/13) and I am honestly flabbergasted they think it's worth 10$/mo. My experience using it till this day is the following: - It's an amazing all-rounder autocomplete for most boilerplate code. Generally anything that someone who's spent 5 minutes reading the code can do, Copilot can do just as well. - It's terrible if y…

> I've been using Copilot non-stop > I am honestly flabbergasted they think it's worth 10$/mo These two statments seem contradictory to me. Why are you using it 'non-stop' if it isn't even worth $10/month?

I've had Copilot enabled since early beta. I think it has saved me ... 15 minutes of typing in total? A few times it has caught on to a repeating pattern and filled a tedious bit of [({}{})] -style javascript correctly.

Most of the things it does for me I could replace with a library of snippets if I could be bothered to set one up.

Not really worth a monthly cost equivalent to, say, Disney+ - which I use tens of hours every month just by myself.

If my employer paid for it, I wouldn't scoff at it, but I'm not paying a cent of my own money for it.

Re: GitHub Copilot is generally available

#416

Post by Snild »

albertzeyer wrote: post #188

I don't really get this argument why it should be a problem that it is being trained on other public code. Every human was just trained in the same way. Why isn't this a problem for every human? I really don't see the difference. One is an artificial neural network while the other is a biological neural network?

I guess it's a bit of a philosophical disagreement.

In my view: I don't believe a machine (at least not any we're capable of creating) can truly learn.

Copilot is a machine working on its inputs. Humans think and create. Maybe it can be argued that humans are just more complicated machines, but I don't think most people would agree with such an equivalency.

Copilot is constructed almost entirely from others' code. There's a tiny fraction of original "ai glue" in there, but the end product is arguably a derivative work of all that code it was trained on. As is its output.

It can also be argued that the AI part is really just an obfuscating copy machine. One that was created specifically for that task.

And of course, the real killing blow: if/when it reproduces training code verbatim, and you don't notice... will "copilot did it" be a valid defense in court? There are different opinions on that I guess, but no one knows for sure -- and I wouldn't take that risk.

Re: GitHub Copilot is generally available

#417

Post by danuker »

mcluck wrote: post #382

Like many people I thought Copilot was neat but ended up uninstalling it because it caused more problems than it solved. Reading the comments here, it seems that most of the people who get value out of it would be better served creating a set of snippets. If all you need is to fill in boilerplate all the time or repeat general test structures but with different arguments, just make a snippet. Every major code editor…

Or better yet, save the snippet as a function/procedure in the code, and avoid needless duplication (DRY/Occam's razor).

Re: GitHub Copilot is generally available

#418

Post by xpe »

shadowgovt wrote: post #173
lamontcg wrote: post #146

Earlier quoted context omitted.

Another concern is that nearly every stackoverflow answer or wikipedia article that isn't a trivial algorithm tends to be buggy at its edge conditions. Most of them look like they were submitted by college students and not experts.

Remember when we believed that experts were over because the wisdom of the crowds would reign supreme? Been a hell of a decade, hasn't it.

The "wisdom of the crowds" doesn't mean what many people think it means.

The wisdom of crowds works best when:

1. participants are independent (otherwise you may get failure modes, such as "groupthink" or "information cascades")

2. participants are informed, but in different ways, with different opinions;

3. there is a clear, accepted aggregation mechanism, where individual errors "cancel out" to some degree

I view the topics in James Surowiecki's book (or the Wikipedia summary of it, at least) as required thinkinpg for everyone, preferably synthesized with a study of statistics and political economy.

In particular, the Wikipedia article's section on "Five elements required to form a wise crowd" is a slightly different slicing of the required elements that I offer above.

* If you read that section, trust is listed. I, however, don't see trust as a necessary condition for a "wise crowd". Trust is often useful (or even necessary) when a collective decision is used for governance, decision-making, and policy.

Re: GitHub Copilot is generally available

#419

Post by theshrike79 »

matsemann wrote: post #397
jwpapi wrote: post #386

I’m honestly shook at all the comments here. I don’t make any money coding and I’m probably in the lower 25% of HN readers in terms of skills, but I’m more than happy to pay $10/m. I would pay Github $10/m for what they already give me. What is your time worth? You should easily get $60/hr, so you need to save 12 minutes per month to make it worth. I would pay that for all my employees. CoPilot is not a replacement f…

Could it perhaps be you not coding for a living (and being lower in skill as you say) that make you think it's worth it? For me the bottleneck is seldom typing. And while Copilot can sometimes dish out some more advanced stuff, I still have to verify it and understand it. Since I can basically solve every problem I encounter day-to-day, Copilot's contribution is not that useful.

> For me the bottleneck is seldom typing

This is one of my pet peeves in this field. People create whole programming languages that are "expressive", just to save typing a few dozen characters and have huge tirades against "verbose" languages that require typing a bunch of boilerplate.

If typing the code is the bit that takes the longest for you in a project, stop and take a good look in the mirror. There's something else wrong in the process.

Post reply on HN