Friday, June 24, 2022

GitHub Copilot Is Generally Available

Thomas Dohmke (Hacker News):

Today, I am thrilled to announce that we are making GitHub Copilot generally available to individual developers. Your AI pair programmer is here.

With GitHub Copilot, for the first time in the history of software, AI can be broadly harnessed by developers to write and complete code. Just like the rise of compilers and open source, we believe AI-assisted coding will fundamentally change the nature of software development, giving developers a new tool to write code easier and faster so they can be happier in their lives.

[…]

GitHub Copilot wouldn’t be possible without GitHub’s vibrant community of students and creators. To support and give back to those communities, we’re making GitHub Copilot available for free to verified students and maintainers of popular open source projects.

FOSS Post (via Hacker News):

Some open source software developers argued that the resulting neural networks is a derivative work of the GPL work, and hence, should be demanded to be released under the GPL license as well.

GitHub’s current CEO said that from their point of view, they see this as a part of “fair use”; which implies that using few lines of modified codes from a public source code is not enough to establish any type of lawsuits against them[…]

However, others argue that the neural network outputs (on the average of a 0.1% probability) copy-pasted snippets from various repositories on GitHub, and hence, it can not fall under fair use[…]

ReinH (via Hacker News):

github copilot is incredible. it just sells code other people wrote, but because it’s an “AI” it is apparently allowed to launder that code without it being a “derivative work”.

Previously:

12 Comments RSS · Twitter

Old Unix Geek

This is a terribly bad idea. It's solving the wrong problem.

Software these days is incredibly wasteful. https://www.positech.co.uk/cliffsblog/2022/06/05/code-bloat-has-become-astronomical/

Why? Because of programmers who cannot program, but only use Google. https://pvs-studio.com/en/blog/posts/0952/

What does Copilot do? It makes googling more efficient, selecting for the incompetent over the competent. And it does so by stealing the work of the competent.

Ask yourself, would Shakespeare use tools like this to write his plays? Just like English, better code is focused on solving a problem simply and elegantly without corner cases so that others can understand it.

Copilot will guarantee that the incompetent appear more competent than they are, and produce more subtle bugs way beyond their ken, since it is harder to understand a bug than to write one.

What copilot should have been solving is debugging, or finding ways to simplify your code, problems where the competent actually need help.

What we need is less but better software. This is just going to make more worse software.

My response to this? I've stopped contributing to open-source / GPL code. Screw it. The idealistic ideas of my youth about helping create a better world of software hit the wall of reality: few people genuinely want to contribute to any field. Most are lazy and greedy.

Everything rots. In the olden days we set up conferences to teach each other cool tricks. Then they became "career boosting points". Now they're full of people who know nothing but want to look cool. Even Scientific papers are failing in the same way: papers that when you work to replicate them clearly could never have worked. Errare humanum est should be replaced by Corruptio humanum est.

Old Unix Geek

Here's an example of the code Copilot made:

https://matthewbutterick.com/chron/this-copilot-is-stupid-and-wants-to-kill-me.html

Eratosthenes' sieve, but totally and utterly butchered, and wrong.

Old Unix Geek

Well, I really went down this rabbit hole...

https://aws.amazon.com/blogs/aws/now-in-preview-amazon-codewhisperer-ml-powered-coding-companion/

Here's amazon's version. It's not trying to do Eratosthenes' sieve but contains identical logic to what GitHub Copilot came up with... For instance, isPrime should stop when i*i > n. (Notice, they even use the same variables).

So I was wrong to write that the code was wrong, in that it would work. But it's awful code: both use an O(N^2) algorithm when an O(N) algorithm has been known of for 2300 years, and Copilot added an irrelevant array/list.

Alright, enough from me.

I've been using Copilot for a month or so now, and it's pretty interesting what it can and can't do. It's bad at just writing whole functions for you, but it's nice for repetitive stuff that looks similar every time you do it, like null checks. It also learns from your own code, and follows your coding conventions.

Another nice thing I've found is that it does translations. Everything I write is multilingual, and Copilot somehow figures out what files contain the translations, what language they are, and then suggests surprisingly usable translations when I type in a new label.

You know, OUG, for somebody who, in other contexts, thinks that it is great for tools (let's say search engines, just to pick a random example) to output wrong information, and that it's the users' fault for not detecting that, I find your take on this tool pretty surprising :-)

Old Unix Geek

@Plume, "wrong" information eh? The propaganda really works. I hope you have been studying the latest edition of the NewSpeak Dictionary in your free moments, as every self respecting member of IngSoc should. :-)

You well know that I believe that not reflecting objectively what is on the internet is wrong information when dealing with search engines, since I believe that's the implicit promise of a search engine. If Google Search said it only searches information approved by Google, I'd have less of a problem with it. But then fewer people would use it.

If CoPilot were promising "I will help you craft bad code", I would also have less of a problem with it. Of course fewer people would use it if it did. But it's promising to help you write good code. So although I see the irony you mention, in a sense my problem with it is the same -- it does not deliver what it promises.

I do see its value in filling in boilerplate code, and perhaps that's all it should do. But that would require it to know what is boilerplate and what isn't. And then, I'd be wondering why there's a need for boilerplate code...

>Why? Because of programmers who cannot program, but only use Google. https://pvs-studio.com/en/blog/posts/0952/

If I were PVS-Studio, I'd take down that "we do not want you to work here" ad.

The author thinks far too highly of themselves, thinks they're the smartest person in the room (thereby disqualifying them from management), and never even thought to have an interviewee explain the code they wrote, and their thought process in writing it.

On top of that, "productivity plateaued"? Well, yes, of course it did. Did the author think these are five-year-olds who keep learning at a breathtaking pace?

What a toxic work environment.

> What does Copilot do? It makes googling more efficient

That's not necessarily a bad thing.

Having said that, I'm rather skeptical of Copilot. IntelliCode, which is kind of its lesser sibling that only offers ML suggestions for line completion, seems alright; filling out an entire method seems problematic for a number of ways.

You know own that code. When you commit it, the author will be you. When PR comes up, you're the one to defend it. So you better understand it. Wondering if it works correctly? Enjoy picking it apart bit by bit to the point where it may have been faster to write it from scratch; that way, at least you're aware of the thought process that went into it.

Then there's the licensing angle — us mere mortals cannot make an amateur video with pop music in the background without getting taken down, and we cannot take a screenshot in the TV app because someone at MovieCo thinks we could take 24 screenshots a second and add an audio track in order to pirate the thing. But when a corporation — a member of the litigious Business Software Alliance, mind you — flat-out takes code from GitHub repos wholesale, without so much as consultation, much less attribution, and sells the result back to you for $10/mo, that's somehow cheered by indie developers.

I'm not saying that's expensive for the service. I'm saying the least they could've done is create a pool where 70% of the revenues get paid to the maintainers of the repos they took the code from.

Old Unix Geek

> If I were PVS-Studio, I'd take down that "we do not want you to work here" ad.

I don't think you're Russian, Sören. Ivan Belokamentsev and PVS studio are. Cultural expectations are different. Asking someone to explain their work might be seen as insulting there.

I'm curious whether Russian programmers would find this a toxic "advertisement". They might instead see this as an advert for excellent programmers who want a challenging problem to work on without colleagues who need hand holding.

"@Plume, "wrong" information eh? The propaganda really works"

If you can't tell the difference between a free, independent press and state propaganda, there might be nothing in the world that could possibly change your mind about this.

"If Google Search said it only searches information approved by Google"

But they *do* say that they prioritize reliable information. Google literally says that their mission is to "deliver the most relevant and reliable information available." It's not a secret. That's what they say.

More to the point, though, I'm not sure why you're conflating ranking with removing. Google obviously ranks results, and they've always attempted to rank correct information above incorrect information. That's neither a secret, nor in any way confusing to the vast majority of Google's users. In fact, it's the reason they use it in the first place.

I've never argued that Google should remove state-controlled media like RT from its index (and Google has, in fact, not removed them), so I'm not sure who that specific point about "only searches information approved by Google" is addressed to.

And just to make sure I've said this: it is correct to rank results from free, independent media above state-controlled media, because a priori, they have a higher likelihood of being correct. This should not be controversial.

Whether we're talking about Google or Copilot, tools should do the best they can to provide correct results. That, too, seems like it should be a pretty uncontroversial statement.

"I'd be wondering why there's a need for boilerplate code..."

Sure, but until we've reached a point where boilerplate code is no longer needed, tools like this remain useful. Mind you, I do think that what they're doing is incredibly unethical, and possibly illegal. But something can be unethical, illegal, *and* useful at the same time.

Old Unix Geek

@Plume: I think you put too much stock into the difference between "state controlled media" and "free, independent media" as you call it. It might be comforting but it's not relevant. What's relevant is whether the information you get from the news is predictive.

For what it's worth, the BBC, Deutsche Welle, Radio France, etc are all "state media" just as much as "RT", "Al Jazeera" and "CGTN", yet some are supposedly credible and others aren't.

In the US most media consumed by the population would fall into the latter category. They all happen to be owned by 6 or so corporations, many of whom have defense industry ties, but more importantly pretty much all sing from the same choir book.

What's truly freaky is that as time has gone on, other Western language media has also converged on the same choir book.

Why do I call it a choir book? Because if it were the truth, it should help one predict the future accurately. It doesn't. Instead the exact same faulty biases appear in every media. One has to search deeply to find material that better predicts the future. That's a problem, and did not use to be the case.

Most examples are controversial, so I'll ask Michael's indulgence just to explain what I mean.

I have a better record at predicting the path of the most recent pandemic than the mainsteam media had, simply because I found a range of sources who knew something, including scientific papers. When I first wore N95 masks, it was so strange to people around me that they thought it might be I who was infected. Some of my sources have been censored off social media. Fauci at the time was saying masks don't work.

Similarly, I doubted the sanctions against Russia would work: Russia isn't a gas station with Nukes as Obama put it. I guess people are slowly learning this fact, no thanks to the media which is supposed to inform them. The Ruble is not "rubble", despite Joe Biden's wish. I have little doubt that Ukraine will lose this war, and all we are doing by shipping more weapons there is causing more people to die, something I find reprehensible. I even wonder whether NATO could beat Russia at conventional warfare at this point, despite the US believing it has the best military on the planet, given this article from the Royal Institute of war: https://www.rusi.org/explore-our-research/publications/commentary/return-industrial-warfare But you certainly don't get that perspective from the "free and independent media".

To summarize, there is a British expression that seems apt: they treat us like mushrooms. It means they keep us in the dark and feed us bullshit. I see that happening both in state and "free, independent" media.

Anyway, I know Michael doesn't like content that is not tech-related on his website, so I'll leave it at that.

Old Unix Geek

(By latter category I meant "free & independent".)

>I think you put too much stock into the difference between "state controlled media" and "free, independent media" as you call it. It might be comforting but it's not relevant.

Of course it is.

Unless you believe that elections make zero difference or that the media automatically adopt the entire government position, free, independent media are different in that they get to express opinions that aren't those of whatever the current executive coalition happens to be.

> For what it's worth, the BBC, Deutsche Welle, Radio France, etc are all "state media" just as much as "RT", "Al Jazeera" and "CGTN"

The BBC is not a state medium. Deutsche Welle indeed is, but the apt comparison to the BBC is ARD, ZDF, etc., which are not state media. Al Jazeera is largely state-funded, but its international versions are mostly independent. Lumping Al Jazeera English in with RT is silly.

>yet some are supposedly credible and others aren't.

Yes, imagine that. Some things are different from others.

>They all happen to be owned by 6 or so corporations, many of whom have defense industry ties, but more importantly pretty much all sing from the same choir book.

In foreign policy, maybe.

In other areas? Are you arguing that MSNBC, CNN and Fox News have similar talking points?

(There's also a lot more to media than cable TV, which is losing relevance anyway.)

>Fauci at the time was saying masks don't work.

That's taken out of context. He was indeed saying (for a rather brief portion of the pandemic) that masks weren't effective. This was at a time when 1) the virus was thought to be spreading by contact rather than by respiration, and 2) mask availability was severely limited, so advising people to get masks would have led to a panic regardless.

Old Unix Geek

> This was at a time when 1) the virus was thought to be spreading by contact rather than by respiration, and 2) mask availability was severely limited, so advising people to get masks would have led to a panic regardless.

He was supposed to be telling us the best scientific advice, i.e. the truth. He didn't. There was plenty evidence that it was aerosolized. That's why I wore an N95. I don't care that you think lying was "to prevent panic". He lied, about something that kills people, and therefore has lost credibility, and the same goes for the media who pushed his lie.

As to the BBC today, I disagree. In the 70s perhaps you'd have been right. Personally, I've given up on complaining to Ofcom.

As to the US media, they make a big story about the tiny differences from a to b, leaving out the reset of the alphabet of human experience.

Sure, if you care about abortion, there are different opinions, although simple questions such as why have we been relying on a dubious legal interpretation are simply not asked. Every other Western country has worked on a bipartisan compromise to make it legal. Why hasn't the US in the last 50 years. Even Ireland, which is pretty Catholic, managed to reach a compromise.

There's very little, if any, coverage of ideas such as:

* Is it healthy that the media, food production, social media, etc are so concentrated?

* Is it ok to trade Americans' personal information, or should the US be imposing restrictions like the EU's GPDR?

* Capitalism might be incompatible with fixing climate change, and even more so Globalism. What are we going to do about that?

* Why is climate change rarely mentioned, even though it most likely will wipe out civilization, unless we change our ways really soon? (And that means international cooperation, not war and sanctions).

* Renewables are barely 5% of the energy mix. They are intermittent. Without fossil fuels, we cannot feed the world's current population, let alone adapt (moving cities, making replacement cars, etc all takes energy). What's the plan? Is there one?

* It's taken for granted that democracy and capitalism is the best system, and we have both. Do we? Is it really true that no better ways of organizing people's interactions could be found?

* Why do all things related to war play such an enormous role in the US economy and is this what we want?

* Why is the NSA recording everything we do? Is that consistent with being democratic and free?

* Is it ok that there are so many homeless and poor people in the US? That so many do not have any health care? That so many work many jobs? What is lost by people having no free time? What about the citizen engagement Alexis de Tocqueville thought found so amazing in the US?

* Why does the US healthcare system cost more and have worse outcomes than pretty much any other country's? What are we going to do to fix it?

* Why is one job not sufficient to bring up a family? Is this ok, or is it related to the demographic collapse in the West?

* Why is it acceptable to export technological know-how to other countries to make a small segment of the US population rich, while depriving the rest of the US of that very know-how and the ability to work in those domains? Does it matter that the loss of manufacturing capacity and know-how makes us militarily and economically more vulnerable?

* Are so many migrants coming here because they can no longer make ends meet in their own countries due to our corporations' behavior over there?

* What are the health effects of all the chemicals and micro-plastics in our environment? How will we change regulation about that?

* How corrupt are regulators? How come the revolving door between regulators and the regulated is tolerated?

* why did the people responsible for the last economic crashes, such as the subprime mortgage scandal, not go to prison?

* Why, in a democracy, has it been demonstrated that only the very wealthy have an influence on what laws are made? (Princeton study)

* Why is the amount of money political parties can spend not capped? Why can't the US just run an election properly like other democratic countries do?

* who were the powerful people Epstein trafficked women to? (other than one Queen's son)

* why is Israel so special that companies in the US are forbidden from not selling to it?

* why are there still laws on the books about it being ok to discriminate against communists? do we believe in freedom of thought and association, or not?

* what laws are still on the book after 9/11 which deprived people in the US of habeas corpus and other rights? Why have they not been removed?

* was the US right to attack Iraq, Afghanistan, etc? How can it blame China's treatment of Uyghurs if it ran Guantanamo?

* why does the US get to impose sanctions violating international law? or start wars violating international law?

* why were the people who tortured others at US black sites not sent to prison? Do we want to be people who are seen by the world as torturers?

* Do we really have a free media, if Assange, a non-US citizen, is to be extradited for "treason" to the US for publishing stuff the NY Times would have published at the time of the Pentagon Papers? Would the NY Times no longer publish such material, and if so, why not?

* If the media is so free, why aren't there programs on it about the good things Russia does, as well as the bad? Russians make good stuff: classical music, literature, science, national parks, etc. As the other country with lots of nukes, and with a strong role in the world economy, they matter. And the same observation goes for any other nation our elites don't like, of which there are many.

* What role, if any, did our politicians' actions have on the eruption of war in Ukraine? Were our politicians clever to say that Ukraine and Georgia would join NATO?

* Why do we have inflation? Was it due to our government's decisions about COVID? or due to our sanctions on Russia? or is it really all "Putin's fault?"

* Why are so many non-Western countries turning away from Western leadership of the world? Do we want a new iron curtain between the West and the rest? What will the dollar losing its dominance in the international trade of commodities do to our standards of living? How hard will it be for the US to adapt to a world with gas prices 5x as high? How will the US do when it costs more to import food, at the same time as the Ogallala aquifer collapses, and as the amount of fertile soil left is depleted?

* Why is the US education system failing versus the rest of the world's?

* Why do US journalists never ask politicians hard questions? Politicians are our representatives, yet they seem quite happy not to represent us. Many of them need to be grilled mercilessly.

But sure, there's a "wide" spectrum of opinions in US media, for a particularly low value of "wide". Most people have little free time to read up on these things. Therefore the media plays an essential role in shaping the range of debate. This is a country with a small range of debate. Most people might not notice it, but they would notice a difference, if they were exposed to a much larger range.

You might object that many of my examples have an international flavor to them, but our standard of living is inextricably entwined with our dominant role in the world. Not reporting it doesn't change it. Britain had a similar golden age during its empire. Then it lost it all, and some Brits have yet to recover from the shock that they weren't all that special after all.

Leave a Comment