Not to be confused with AWS Copilot - which has been an area of focus for the AWS container services team: <a href="https:&#x2F;&#x2F;aws.github.io&#x2F;copilot-cli&#x2F;" rel="nofollow">https:&#x2F;&#x2F;aws.github.io&#x2F;copilot-cli&#x2F;</a>

Like this?<a href="https:&#x2F;&#x2F;visualstudio.microsoft.com&#x2F;services&#x2F;intellicode&#x2F;" rel="nofollow">https:&#x2F;&#x2F;visualstudio.microsoft.com&#x2F;services&#x2F;intellicode&#x2F;</a>I bet it&#x27;s the same people, trying to push their crap into all sorts of successful products.

Yeah, running locally would be my preference. I get &quot;Antitrust (2001)&quot; vibes from this, but that&#x27;s the tinfoil hat side of me.

Last time I tried Tabnine it wasn&#x27;t really of much use to me, the top of the line GPT-3 is a much much bigger model, it should be able to do much more intelligent things.

I&#x27;d like to see something like that, but with knowledge about every single file in the codebase, and running locally.

This is really the same argument as it used to be: help intelligences (used to be only human, now artificial) to find bugs by matching text with code.

I suppose this is a whole new argument in favor of good code commenting - to train&#x2F;share context with your future tooling.

There are no tools that can format my code with my coding style AFAIK. There are multiple tool that can format my code with their coding style though, which I don&#x27;t care about.

Don&#x27;t most IDEs handle this already?

Apparently not that easy[0]. Depending on one&#x27;s priorities formatting&#x2F;pretty-printing code can be trivial or very difficult. Doing that dynamically based on a user&#x27;s individual preferences or even their existing codebase&#x27;s style is probably much harder.I don&#x27;t doubt Copilot is also quite a problem to solve, most likely indeed harder, though in that case at least there&#x27;s an abundance of training data.[0] <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=22706242" rel="nofollow">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=22706242</a>

It depends what you mean by &quot;given&quot;, I can&#x27;t write a million line document describing exactly what kind of style I want it to use, the formatter must learn the style on its own from examples.I agree that something like that would be much easier to make in theory, hence why I&#x27;m suggesting it since maybe it could be made ~perfectly, which Codepilot isn&#x27;t (we haven&#x27;t unlocked AGI yet).

Formatting code according to a given style is a much easier task than what Copilot does.

IMO a potentially more interesting application of this technology would be a learning system that is able to learn your coding style. You give it access to the codebase and it reformats all files on save according to your likings, perfectly.Obviously a program that is able to write actual great code reliably would be spectacular, but we aren&#x27;t there yet, I don&#x27;t think Copilot presently is able to make me meaningfully more productive.

From Clippy to this technology, in just under 25 years! Very impressive. I wonder how much impact this tool might have as a teaching aid, as well?

in their own example the copilot aimed for the mountains with float() - thanks but no thanks

Tell that to the new programmer who builds a piece of software using this creating an absolute mess.The danger here isn&#x27;t with experienced developers (this is, obviously, a tool with great potential for productivity). It&#x27;s with people who just blindly trust what the robot spits out.Once code like that is implemented in a mission-critical system without discernment, all hell will break loose.Edit: worth watching this for context <a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=ZSRHeXYDLko" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=ZSRHeXYDLko</a>

What is the hard boundary between you and your AI companion?What happens when people think their &quot;self driving&quot; cars are more capable than they actually are? Many times, you are better off in a dumb car, because your expectations are such that you have to pay attention to a continuous stream of events and respond in a stateful manner.If you try to bootstrap a holistic understanding of your problem in between bursts of auto-generated code, I don&#x27;t think you are going to have a fantastic time of it.

It&#x27;s a copilot. You&#x27;re still the pilot. To be honest this seems like it can definitely save me a bunch of googling and let me stay in the ide.

I would say the use of float makes it a nonstarter - even for &quot;quick one-off&quot; scripts.  That&#x27;s a fundamental error in the generated code.  It maybe looks correct at a quick glance but it&#x27;s only introducing subtle errors to find down the line.

The output is great for a quick one-off script. Maybe if you make the comments look more “enterprise-y”, it’ll go for more careful code?

Should I be impressed that the example parse_expenses.py on the home page doesnt include any error handling and uses a float for currency?  This seems like it&#x27;s going to revolutionize copy and paste programming.

This will allow for a D or C level coder to be a B&#x2F;B- coder which is great, quality goes up. But corps will use this to depress wages and finally be able to create that wonderous unicorn of completely fungible coder.This kind of tooling is akin to the crossbow.It will allow for less skilled folks to push out code that is like other code at great speed. A copy pasta accelerator if you will.

Maybe thirty four years later. I don&#x27;t think there&#x27;s AI to gather requirements, talk to people, understand a problem and produce code. That&#x27;s kinda general intelligence level AI.
But this thing can possibly make devs work easier and if it&#x27;s good enough maybe smaller teams can produce more.

I think programmers will just write model and tests in the end. The rest will be generated.

I&#x27;m not at all worried about AI taking over software development. In all likelihood, what you&#x27;ll see instead are AI plugins in IDE editors which just assist in a much more advanced way than the intellisense we have now. Having machines code out the business logic is very much so something that would be less efficient than having a person do it.Realistically, it just means that, rather than your coworker code-reviewing you and making a handful of comments, you get a machine to do that, and get two or so comments from your coworker about the business logic.To answer your question: never.

I imagine that AI like this will certainly speed up development, but I suspect you will almost always need someone in the middle putting the pieces together.

I&#x27;d love to live in a world where AGI solves cancer.

Unless we achieve AGI this is never going to replace programming because it will instead just make programming a higher level task. 
And well, if we achieve AGI (which I think could be pretty soon), all jobs will replaced, so it&#x27;s not something I think anybody should be worried about.

Nah. Remember that &quot;Google AI signs you up for a hair stylist session&quot; demo? We never got anything out of that.

Four years later: your AI replacement? When do you all predict something like this will happen?

I&#x27;d say it&#x27;s more than &#x27;just&#x27; an autocomplete system.Naive autocomplete, as implemented in Excel since forever ago (and I&#x27;m sure long before that, I&#x27;m just familiar with being annoyed by Excel suggesting wrong entries from its simple and over-eager autocomplete system), merely matches a sequence of characters - if I typed &quot;aut&quot; again in this paragraph it will suggest &quot;autocomplete&quot; because I recently typed it. Implementing it is the kind of task you give to a first-year programming student to practice string matching data structures, similar to a spell checker that merely checks that a string exists in a dictionary.There&#x27;s a spectrum from &#x27;just&#x27; autocomplete, to a syntax-aware system like VS Intellicode, to this, and eventually beyond this. As mobile predictive text is to a spell checker, so Github Copilot is to autocomplete. As mobile predictive text is to GPT3 [1], so Github Copilot is to...what next? GPT3 is not just a spell checker.[1] Also by OpenAI: <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=23345379" rel="nofollow">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=23345379</a>

It should be able to (try to) do something like that too.There&#x27;s a little demo about that here: <a href="https:&#x2F;&#x2F;copilot.github.com&#x2F;" rel="nofollow">https:&#x2F;&#x2F;copilot.github.com&#x2F;</a>It&#x27;s &quot;just&quot; an autocompletion system basically, if you write something that looks like the beginning of a test it should understand that and try to autocomplete that.

<a href="https:&#x2F;&#x2F;www.diffblue.com" rel="nofollow">https:&#x2F;&#x2F;www.diffblue.com</a>Here you go :)

I would love to see an equivalent where it generates all the tests for you

(not a lawyer) Copyright issues tend to involve the question of how transformative a work is. This means the code coming out the other end is probably fine. I don&#x27;t know about the training side, though. Are there license issues in using copyrighted training data without any form of licensing? Typically ML researchers have a pretty free-for-all attitude towards &#x27;if I can find data, I can train models on it.&#x27;

From the FAQ at the bottom of the project showcase page[0]:&quot;GitHub Copilot is a code synthesizer, not a search engine: the vast majority of the code that it suggests is uniquely generated and has never been seen before. We found that about 0.1% of the time, the suggestion may contain some snippets that are verbatim from the training set. Here is an in-depth study[1] on the model’s behavior. Many of these cases happen when you don’t provide sufficient context (in particular, when editing an empty file), or when there is a common, perhaps even universal, solution to the problem. We are building an origin tracker to help detect the rare instances of code that is repeated from the training set, to help you make good real-time decisions about GitHub Copilot’s suggestions.&quot;[0] <a href="https:&#x2F;&#x2F;copilot.github.com&#x2F;" rel="nofollow">https:&#x2F;&#x2F;copilot.github.com&#x2F;</a>[1] <a href="https:&#x2F;&#x2F;github.co&#x2F;copilot-research-recitation" rel="nofollow">https:&#x2F;&#x2F;github.co&#x2F;copilot-research-recitation</a>

(not a lawyer) my interpretation of GPLv2 is that at the very least the model would be licensed under GPL if it was trained on GPL code. The model is a derivative work. Whether the generated code coming out of the model is GPL is trickier. I would lean towards yes but I&#x27;m not entirely sure.I think talking about exact text matches to existing code is a red herring. If you took GPL code and ran it through an obfuscator that changed every byte of the code to new code, that resulting code would be derivative and would need to be licensed under GPL too.Thank you Microsoft for ushering in a new era of free software.

GitHub says you don&#x27;t need to credit GitHub for any of the code suggestions, but since it&#x27;s trained on public sources of code, anyone have a clue on potential licensing pitfalls?

Or more realistically, a language server which would then be compatible with many editors including (Neo)Vim.

Presumably, this will be accessible through a REST API or something like that at some point, so that it can eventually be integrated into all editors.

now they just need to build this into vim.

I don&#x27;t know if they&#x27;ve officially said anything on this, but since the MS acquisition it feels like Atom is on life support. Github Codespaces is built on VSCode, and there&#x27;s a lot of effort from Github going into their VSCode extension for things like PR review.

It&#x27;s funny that GitHub ships their own text editor, Atom, but the second example (under the &quot;More than Autocomplete&quot; header) on the Copilot website is clearly using VSCode.

Then, I should see absolutely zero complaints about privacy, tracking, spying, google analytics and facebook tracking.Perhaps we should ask Scott if he is willing to share his browsing history, his personal photos and his passwords with the rest of us or maybe if I can come into his house?After all, &quot;You have zero privacy anyway&quot;

&quot;You have zero privacy anyway. Get over it.&quot;Scott McNealy, CEO Sun Microsystems, 1999

No, you are not being paranoid. This tool literally uploads all the code it wants off your machine, and I see no way of filtering out secrets and the likes.
You have all the rights to be worried about that.

&gt; Am I being to paranoid here?No.They already admitted that they send telemetry of the code you give it and its training set already has personal information in it anyway, despite what is being hyped up here by the fanatics, even when someone said that &#x27;Copilot guesses the exact code I want to write about one in ten times&#x27; [0]No thanks and certainly no deal.[0] <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=27676845" rel="nofollow">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=27676845</a>

I am pretty sure they will train it on private repos they are hosting as well (just not be public about it)

Right. If you’re comfortable giving access to your source files to GitHub+OpenAI, then go for it.I’m not sure how this would apply to secret keys or flat files with customer data&#x2F;PII, but in any case that makes it a non-starter for me.Their “Please do not share this URL publicly.” Banner at the top of the page which disclosed this info makes my skin crawl a bit…If I were only working on public projects I would be on board right away, it looks like a big time saver.Am I being to paranoid here?

Gigantic caveat.&gt; I agree to these additional telemetry terms as part of the technical preview

I would argue that a trained model falls under the legal category of &quot;compilation of facts&quot;.More generally, keep in mind that the legal world, despite an apparent focus on definition is very bad at dealing with novelty, and most of it will end up justifying a posteriori existing practices.

&gt; (I remember an article from HN about how bash had to be written by someone who hadn&#x27;t seen the unix shell code or something like that).I believe you&#x27;re referring to Clean Room Design[1].[1] <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Clean_room_design" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Clean_room_design</a>

You bring up a really good point. I&#x27;m super curious what the legality and ethics around training machines on licensed or even proprietary code would be. IIRC there are implications around code you can build if you&#x27;ve seen proprietary code (I remember an article from HN about how bash had to be written by someone who hadn&#x27;t seen the unix shell code or something like that).How would we classify that legally when it comes to training and generating code? Would you argue the machine is just picking up best practices and patterns, or would you say it has gained specifically-licensed or proprietary knowledge?

Could it be this? <a href="https:&#x2F;&#x2F;visualstudio.microsoft.com&#x2F;services&#x2F;intellicode&#x2F;" rel="nofollow">https:&#x2F;&#x2F;visualstudio.microsoft.com&#x2F;services&#x2F;intellicode&#x2F;</a>I was wondering the same thing, especially with MS being behind both.edited: or this? <a href="https:&#x2F;&#x2F;docs.microsoft.com&#x2F;en-us&#x2F;visualstudio&#x2F;intellicode&#x2F;custom-models" rel="nofollow">https:&#x2F;&#x2F;docs.microsoft.com&#x2F;en-us&#x2F;visualstudio&#x2F;intellicode&#x2F;cu...</a>

I seem to remember a similar discussion on Intellicode (similar thing, but more like Intellisense, and as Visual Studio plugin), which is trained on &quot;github projects with more than 100 stars&quot;. IFIR they check the LICENSE.txt file in the project and ignore projects with an &quot;incompatible&quot; license. I don&#x27;t have any links handy which would confirm this though.

There will almost certainly be cases where it copies exact lines. When working with GPT2 I got whole chunks of news articles.

Well, it probably is explicitly copying at least some subset of the source text - otherwise the code would be syntactically invalid, no?

But it actually is explicitly copying the text.  That&#x27;s how it works.  The training data are massive, and you will get long strings of code that are pulled directly from that training data.  It isn&#x27;t giving you just the style.  It may be mashing together several different code examples taking some text from each.  That&#x27;s called &quot;derivative work&quot;.

The trained model is a derivative work that contains copies of the corpus used for training embedded in the model.  If any of the training code was GPL the output is now covered by GPL. The music industry has already done most of the heavy lifting here in terms of scope and nature of derived works, and while IANAL I would not suggest that it looks good for anyone using this tool if GPL code was in the training set.

My guess is that it is, if we think of a machine learning framework as a compiler and the model as compiled code. Compiled GPL code is still GPL, that&#x27;s the entire point.Anyways, GitHub is Microsoft, and Microsoft has really good lawyers so I guess they did everything necessary to make sur that you can use it the way they tell you so. The most obvious solution would be to filter by LICENSE.txt and only train the model with code under permissive licenses.

&gt; you cannot copyright a &quot;style&quot;This line of thinking applies to the code generated by the model, but not necessarily to the model itself, or the training of it.

No, a model trained on text covered by a license is not itself covered by the license, unless it explicitly copies the text (you cannot copyright a &quot;style&quot;).

No, the code generated is what copyright law calls a derivative work and you should go ask Robin Thicke and Pharrell Williams exactly how much slack the courts give for &#x27;sufficiently transformative new work.

My guess would be that the model itself (and the training process) could have different legal requirements compared to the code it generates. The code generated by the model is probably sufficiently transformative new work that wouldn&#x27;t be GPL (it&#x27;s &quot;fair use&quot;).I suspect there could be issues on the training side, using copyrighted data for training without any form of licensing. Typically ML researchers have a pretty free-for-all attitude towards &#x27;if I can find data, I can train models on it.&#x27;

<a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Clean_room_design" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Clean_room_design</a>There&#x27;re definitely cases when devs avoid even looking at implementation before creating their own

Its not cause the License says nothing about training. I mean every oss dev&#x27;s brain would be under GPL then.

And so it begins: We start applying human rights to AIs.Not a critique on your point, which a was just about yo bring up myself.

Certainly not. If I memorize a line of copyrighted code and then write it down in a different project, I have copied it. If an ML model does the same thing as my brain - memorizing a line of code and writing it down elsewhere - it has also copied it. In neither case is that &quot;fair use&quot;.

1) this is not human, it&#x27;s some software2) if I write a program that copies parts of other GPL licensed SW into my proprietary code, does that absolve me of GPL if the copying algorithm is complicated enough?

What if I put a licence on my Github-repositories that explicitly forbids the use of my code for machine-learning models?

I think this would fall under any reasonable definition of fair use. If I read GPL (or proprietary) code as a human I still own code that I later write. If copyright was enforced on the outputs of machine learning models based on all content they were trained on it would be incredibly stifling to innovation. Requiring obtaining legal access to data for training but full ownership of output seems like a sensible middle ground.

Is what a human generates GPLv2 because it learned from GPLv2 code?

until the legal position is clear it you&#x27;d have to be insane to allow output from this process to be incorporated into your codebasesimagine if the output was ruled as being GPLv2, then having to go through a proprietary codebase trying to rip out these bits of codeit would be basically impossible

It has been cleared up! <a href="https:&#x2F;&#x2F;towardsdatascience.com&#x2F;the-most-important-supreme-court-decision-for-data-science-and-machine-learning-44cfc1c1bcaf?gi=b4317de7ca2" rel="nofollow">https:&#x2F;&#x2F;towardsdatascience.com&#x2F;the-most-important-supreme-co...</a>

This is a bit tricky, because at least in the U.S., I don&#x27;t believe it&#x27;s settled question in law yet. Some of the other posters on here have said that the resulting model isn&#x27;t covered by GPL--that&#x27;s partially true, but provenance of data, and the rights to it, definitely does matter. A good example of this was the Everalbum ruling, where the company was forced to delete both the data and the trained models used they were used to generate due to lack of consent from the users from whom the data was taken[1]. Since open source code is, well, open, it&#x27;s definitely less a problem for permissively-licensed code.That said, copyright is typically generally assigned to the closest human to the activation process (it&#x27;s unlikely that Github is going to try to claim the copyright to code generated by Copilot over the human&#x2F;company pair-programming with it), but since copyleft in general is a pretty domain-specific to software, afaik the way that courts interpret the legality of using code licensed under those terms in training data for a non-copyleft-producing model is still up in the air.Obligatory IANAL, and also happy to adjust this info if someone has sources demonstrating updates on the current state.[1] <a href="https:&#x2F;&#x2F;techcrunch.com&#x2F;2021&#x2F;01&#x2F;12&#x2F;ftc-settlement-with-ever-orders-data-and-ais-deleted-after-facial-recognition-pivot&#x2F;" rel="nofollow">https:&#x2F;&#x2F;techcrunch.com&#x2F;2021&#x2F;01&#x2F;12&#x2F;ftc-settlement-with-ever-o...</a>

IMO the closest case is probably the students suing turnitin a number of years ago, which iParadigms (the turnitin maker) won [1].I think this is definitely a gray area and in some way iParadigms winning (compared to all the cases decided in favour of e.g. the music industry), shows the different yardsticks being used for individuals and companies.I&#x27;m sure we will see more cases about this.[1] <a href="https:&#x2F;&#x2F;www.plagiarismtoday.com&#x2F;2008&#x2F;03&#x2F;25&#x2F;iparadigms-wins-turnitin-lawsuit&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.plagiarismtoday.com&#x2F;2008&#x2F;03&#x2F;25&#x2F;iparadigms-wins-t...</a>

&gt; So everything generated also GPLv2?Almost certainly not everything.But possibly things that were spit out verbatim from the training set, which the FAQ mentions does happen about .1% of the time [1]. Another comment in this thread indicated that the model outputs something that&#x27;s verbatim usable about 10% of the time. So, taking those two numbers together, if you&#x27;re using a whole generated function verbatim, a bit of caveat emptor re: licensing might not be the worst idea. At least until the origin tracker mentioned in the FAQ becomes available.[1] <a href="https:&#x2F;&#x2F;docs.github.com&#x2F;en&#x2F;early-access&#x2F;github&#x2F;copilot&#x2F;research-recitation" rel="nofollow">https:&#x2F;&#x2F;docs.github.com&#x2F;en&#x2F;early-access&#x2F;github&#x2F;copilot&#x2F;resea...</a>[2] &quot;GitHub Copilot is a code synthesizer, not a search engine: the vast majority of the code that it suggests is uniquely generated and has never been seen before. We found that about 0.1% of the time, the suggestion may contain some snippets that are verbatim from the training set. Here is an in-depth study on the model’s behavior. Many of these cases happen when you don’t provide sufficient context (in particular, when editing an empty file), or when there is a common, perhaps even universal, solution to the problem. We are building an origin tracker to help detect the rare instances of code that is repeated from the training set, to help you make good real-time decisions about GitHub Copilot’s suggestions.&quot;

IANAL, but my interpretation of the GitHub TOS section D4 would give give GitHub the right to parse your code and&#x2F;or make copies regardless of what your license states. This is the same reason the GitHub search index isn’t GPL contaminated.

You can train models on copyrighted materials. <a href="https:&#x2F;&#x2F;towardsdatascience.com&#x2F;the-most-important-supreme-court-decision-for-data-science-and-machine-learning-44cfc1c1bcaf?gi=b4317de7ca2" rel="nofollow">https:&#x2F;&#x2F;towardsdatascience.com&#x2F;the-most-important-supreme-co...</a>

We kinda have to accept that - we don&#x27;t have to accept this. You can&#x27;t interface with humans but you can interface with one of the biggest corporate tech giants straight up leeching from explicitly public and free work for their own private benefit.

Developers Human brain are also trained with propietary code bases, then when they quit and go elswere , they program using knowledge learned previously, yet you do not sue them.

So if it was trained using &quot;source code from publicly available sources, including code in public repositories on GitHub.&quot; was it also GPLv2?So everything generated also GPLv2?

Well, it&#x27;s a brand new product, so ... no?

According to the comments here, people have been using this for a couple of weeks now.

From the FAQ:
&quot;Not yet. For now, we’re focused on delivering the best experience in Visual Studio Code only.&quot;I do hope they feature others at a later date, especially since they are planning to develop a commercial version.

I wonder if this can work with any editor out of the box, or it&#x27;s just going to be VS Code.

The averageRuntimeInSeconds example does not check for division by zero so it creates broken code at least 20% of the time based on the examples on the homepage :)

Maybe not malicious per se, but certainly I&#x27;d be concerned about seemingly-correct but actually-wrong code being suggested. Considering how often the top StackOverflow answer is slightly wrong or how often antipatterns crop up across various projects, I&#x27;m sure the training data is nowhere near &quot;perfect code&quot; - implying the output cannot be perfect either.

Since it is per line I highly doubt it. I think of it as intellisense+. You select suggestions that you would have written anyway.

I wonder if there&#x27;s any potential for Copilot to suggest malicious code because it&#x27;s been trained on an open source projects containing intentionally malicious code.

The IP implications of how network weights depend on their data sources is, shall we say, a matter of ongoing legal discussion.We genuinely don’t know what it means for licenses or whatnot right now.

Hard to say, really. When writing React components, Jest tests and documentation, it&#x27;s often not very far. I found it off when writing HTML markup (which is hard to describe with words).

&gt; and the rest of the time it suggests something rather good, or completely offIn what, proportion roughly?

It shows the suggestions line by line, and only shows the best guess. It&#x27;s not more intrusive than Intellisense.You can actually see all the code blocks Copilot is thinking about if you want to, but that is indeed a distraction.

How big&#x2F;complicated are the functions Copilot is autocompleting for you? I&#x27;m thinking perhaps reading 10 potential candidates is actually slower and less instructive than trying to write the thing yourself.

do you still go over the generated code line by line and touchup in places where it did not do a good job?

I tried TabNine and it wasn’t a huge improvement because what costs the most time isn’t typing stuff but thinking about what to type.

In addition, I would like to see GitHub reviewing my code &amp; giving me suggestions on how I could improve. That will be more educative &amp; a tool to ensure consistency across code.

I&#x27;m not sure I spend less time actually coding stuff (because I have to review the Copilot code). But the cost of the code I write is definitely reduced, because:- the review from my peers is faster (the code is more correct)
- I come back less to the code (because I have thought about all the corner cases when checking the copilot code)
- As I care more about naming &amp; inline docs (it helps copilot), the code is actually cheaper to maintain.

Having been a TabNine user for a while I can say that it&#x27;s less of a productivity boost and more of a quality of life win. It&#x27;s hard to measure - not because it&#x27;s small, but because it&#x27;s lots of tiny wins, or wins at just the right moment. It makes me happy, and that&#x27;s why I pay for it - the fact that it&#x27;s probably also saving me some time is second to the fact that it saves me from annoying interrupts.

What&#x27;s your estimate of the productivity boost expressed as a percentage? I.e. if it takes you 100 hours to complete a project without Copilot, how many hours will it be with Copilot?

do you have an invite? very interested to check it out

I have used IDEs with good knowledge of the types and libraries I&#x27;m using (e.g. VSCode with TypeScript). They offer good suggestions once you start typing a function name.But nothing gets close to Copilot. It &quot;understands&quot; what you&#x27;re trying to do, and writes the code for you. It makes type-based autocompletions useless.

I&#x27;ve been using TabNine for a couple years – constantly impresses me, especially how quickly it picks up new patterns. I wouldn&#x27;t say it&#x27;s doing my job for me, but definitely saves me a lot of time.

I&#x27;m curious as to how relevant Copilot would be when autocompleting code that is specific to my codebase in particular, like Tabnine completes most used filters as soon as I type the db table name for the query. I&#x27;m a big tabnine fan because it provides this feature. I&#x27;m much more often looking to be suggested a line than an entire function because I&#x27;m mostly writing business logic.also tabnine is useless in multi-lines completes. which is where co-pilot should be strong.

Interestingly the founder of TabNine (which was acquired by Codota[0]) is currently working at Open AI (edit: comments corrected me he left in December 2020 according to his blog).
I imagine they&#x27;re livid about Open AI creating a competing product.TabNine at times was magical, but I stopped using it after Codota started injecting ads directly into my editor[1][0] <a href="https:&#x2F;&#x2F;betakit.com&#x2F;waterloo-startup-tabnine-acquired-by-israeli-startup-codota&#x2F;" rel="nofollow">https:&#x2F;&#x2F;betakit.com&#x2F;waterloo-startup-tabnine-acquired-by-isr...</a>
[1] <a href="https:&#x2F;&#x2F;github.com&#x2F;codota&#x2F;TabNine&#x2F;issues&#x2F;342" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;codota&#x2F;TabNine&#x2F;issues&#x2F;342</a>

I tried out TabNine. It was a very frustrating experience. In almost all cases it gave completely useless suggestions which overrode the better ones already suggested by IntelliJ. I persevered for a few days and then uninstalled it.

Have you used any intelligent code completion in the past? E.g. I&#x27;d really be interested how it compares to TabNine[0], which already gives pretty amazing single line suggestions (haven&#x27;t tried their experimental multi-line suggestions yet).[0]: <a href="https:&#x2F;&#x2F;www.tabnine.com" rel="nofollow">https:&#x2F;&#x2F;www.tabnine.com</a>

It&#x27;s smarter than that. It suggests things that have never been written. It actually creates code based on the context, just like GPT-3 can create new text documents based on previous inputs.Edit: Check this screencast for instance: <a href="https:&#x2F;&#x2F;twitter.com&#x2F;francoisz&#x2F;status&#x2F;1409908666166349831" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;francoisz&#x2F;status&#x2F;1409908666166349831</a>

It seems to replace&#x2F;shorten the loop of “Google for snippet that does X” copy, paste, tweak, no? Which of course is super cool for many tasks!

They finally did it. They finally found a way to make me write comments.

Has anyone used Copilot with a more succinct language? It appears to only automate boilerplate and rudimentary patterns, which while useful in repetitive low signal to noise ratio languages like React or Java, sounds less appealing if you&#x27;re writing Clojure.

Jesus. From a guy with your track record that means a lot.

I’ve also been using the Alpha for around two weeks. I&#x27;m impressed by how GitHub Copilot seems to know exactly what I want to type next. Sometimes it even suggests code I was about to look up, such as a snippet to pick a random hex color or completing an array with all the common image mime-types.Copilot is particularly helpful when working on React components where it makes eerily accurate predictions. I see technology like Copilot becoming an indispensable part of the programmer toolbelt similar to IDE autocomplete for many people.I also see it changing the way that programmers document their code. With Copilot if you write a really good descriptive comment before jumping into the implementation, it does a much better job of suggesting the right code, sometimes even writing the entire function for you.

Few days back, Sam Altman tweeted this&quot;Prediction: AI will cause the price of work that can happen in front of a computer to decrease much faster than the price of work that happens in the physical world. This is the opposite of what most people (including me) expected, and will have strange effects&quot;And I was like yeah I gotta start preparing for next decade.

Automation has always produced an increase in jobs so far, although sometimes in a disruptive way. I consider this like the switch from instruction-level programming to compiled languages, a level of abstraction added that buys a large increase in productivity and makes projects affordable that weren’t affordable before. If anything this will probably lead to a boom in development work. But there’s a bunch of low skill programmers who can’t do much more than follow project templates and copy paste things. Those people will have to level up or get out.

&gt; Anyone want to hire me to teach your grandma how to use the internet?Only for the first time to train a model for that.

On the contrary. This tool increase the programmer productivity, hence you will get more salary not less.You assumption is that programming demand is finite, AND that all programmers are equal, both of those are false.I must also say that actual programming is around 10-15 % of the programmer job, so the tool will make you around 10% more overall productive.

&gt;Pack it all up, boys, programming&#x27;s over. Hello, AI.I don&#x27;t know, cranking out some suggestion for a function is not the same as writing a complete module &#x2F; application.Take the job of a translator, you would say the job would go extinct with all the advances in autotranslation? here it says that &#x27;employment of interpreters and translators is projected to grow 20 percent from 2019 to 2029, much faster than the average for all occupations&#x27; [1]. You still need a human being to clear up all of the ambiguities of languages.Maybe the focus of the stuff we do will change, though; but on the other hand, we do tend to get a lots of changes in programming; it goes with the job. maybe we will get to do more code reviews of what was cranked out by some model.However, within a decade it might be harder to get an entry level job as a programmer. I am not quite sure if i should suggest my profession to my kids, we might get a more competitive environment in some not so distant future.[1] <a href="https:&#x2F;&#x2F;www.bls.gov&#x2F;ooh&#x2F;media-and-communication&#x2F;interpreters-and-translators.htm" rel="nofollow">https:&#x2F;&#x2F;www.bls.gov&#x2F;ooh&#x2F;media-and-communication&#x2F;interpreters...</a>

Nah, we will just all be low-code function couplers instead of coders..

The next skill will be perfectly casting the right spell to make the AI spit out the product as spec&#x27;d.Repurposed Google-fu. We&#x27;ll always have jobs :)

Pack it all up, boys, programming&#x27;s over. Hello, AI.Anyone want to hire me to teach your grandma how to use the internet?

With VSCode, Github and a perhaps a little bit of help from OpenAI, Microsoft is poised to dominate the developer productivity tools market in the near future.I wouldn&#x27;t be surprised to see really good static analysis and automated code review tools coming out of these teams very soon.

I watched the same thing happen with the introduction of Intellisense. Pre-Intellisense I had tons of RSI problems and had to use funky ergonomic keyboards like the Kinesis keyboard to function as a dev. Now I just hop on whatever laptop is in front of me and code. Same reason - massive reduction in the number of keys I have to touch to produce a line of code.

Side note: I recently suffered from a tennis elbow due to sub optimal desktop setup when working from home. Copilot has drastically reduced my keystrokes, and therefore the strain on my tenders.It&#x27;s good for our health, too!

&gt; So no need to pay a high level programmer, just a mediocre one and the right AI assistant gives the same results.I think of it as not needing juniors for boring work, all you need as a company is seniors and AI.

Quite the opposite. Menial work will be automated away (e.g. CRUD) and only good programmers will be needed to do the more complicated work.

I think it&#x27;s more that this tool is only capable of automating the non-creative work that thousands have done already.It&#x27;s still insanely impressive (assuming the examples aren&#x27;t more cherry picked than I&#x27;d expect).

With an AI assistant, in the best scenario, you&#x27;ll get a &quot;wisdom of crowds&quot; effect on implementation details and program architecture. At worst, you&#x27;ll get a lot of opinionated code bloat and anti-patterns as suggestions.For most backend programming jobs, the challenge is not in writing complex code but figuring out what the business wants, needs and should have in the first place and distinguishing between them. Figuring out how to integrate with existing systems, processes, fiefdoms and code. Knowing when to say yes and no, how to make code future proof, etc. This is a task fundamentally unfit for what we currently call &quot;AI&quot;, because it&#x27;s not actually intelligent or creative yet.On the frontend, it becomes even more nebulous. Maybe Copilot can suggest common themes like Bootstrap classes for a form, CSS properties, the basic file structure and implementations for components in an SPA, etc. As I see it, the main challenge is in UX there, not the boilerplate, which again makes it about understanding the user, figuring out how to make UI feel intuitive, etc. Again: unfit for current AI.I cannot offer any opinion on the utility for actually complex, FANG-level code, for lack of experience.

Maybe it&#x27;s just because humans are not as creative as they think.
Whatever you do, thousands of others have done the same already.
So no need to pay a high level programmer, just a mediocre one and the right AI assistant gives the same results.

Until it automatically knows when and how it&#x27;s wrong, you&#x27;ll still need a human to figure that out, and that human will need to actually know how to program, without the overgrown auto-complete.May or may not reduce the demand for programmers, though. We&#x27;ll see.

It could start to replace us in 20 years. Or reduce. It is exciting for now

You&#x27;re using the word &quot;memorized&quot; in a very loose way.

Finally, a faster way to spread bugs than copy&#x2F;paste.

Not sure how this is handled in US, but in Germany a few lines of code have in general not enough uniqueness to be licensed.

The landing page for it states the below, so hopefully not too much of an issue (though I guess some folks may find a 0.1% risk high).&gt; GitHub Copilot is a code synthesizer, not a search engine: the vast majority of the code that it suggests is uniquely generated and has never been seen before. We found that about 0.1% of the time, the suggestion may contain some snippets that are verbatim from the training set.

Update: Nat Friedman answered this as part of this thread on twitter:<a href="https:&#x2F;&#x2F;twitter.com&#x2F;natfriedman&#x2F;status&#x2F;1409883713786241032" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;natfriedman&#x2F;status&#x2F;1409883713786241032</a>Basically they are building a system to find explicit copying and warn developers when the output is verbatim.

What is the licensing for code generated in this way? GPT-3 has memorized hundreds of texts verbatim and can be prompted to regurgitate that text. Has this model only been trained on code that doesn&#x27;t require attribution as part of the license?

for those of you like this concept but 1. did not get into the alpha (or want something that has lots of great reviews) 2. need to run locally (security or connectivity) 3. want to use any IDE... please try Tabnine

You are correct. I feel this is why the service is called Copilot, not Pilot :)

I keep wondering how much time it could possibly save you, given that you&#x27;re obligated to read the code and make sure it makes sense. Given that, the testimonials here are very surprising to me.

I think 1&#x2F;10 is incredible. If it holds up it means they may have found the right prior for a path of development that can actually lead to artificial general intelligence. With exponential improvement; humans learning to hack the AI and the AI learning better suggestions, this may in theory happen very quickly.We live in a very small corner of the space of possible universes, which is why finding a prior in program space within it is a big deal.

Unfortunately, one in 10 times is far from good enough (and this is with good prompt engineering which after using large language models for a while, one starts to do).I feel like the current generation of AI is bringing us close enough to something that works once in a while but requires constant human expertise ~50% of the time. The self-driving industry is in a similar situation of despair where millions have been spent in labelling and training but something fundamental is amiss in the ML models.

It sounds similar to the editor plugin called TabNine

How did you manage to get early access? Is there some kind of GitHub Early Access programme (or something similar part of Enterprise)?

I&#x27;ve been using the alpha for the past 2 weeks, and I&#x27;m blown away. Copilot guesses the exact code I want to write about one in ten times, and the rest of the time it suggests something rather good, or completely off. But when it guesses right, it feels like it&#x27;s reading my mind.It&#x27;s really like pair programming, even though I&#x27;m coding alone. I have a better understanding of my own code, and I tend to give better names and descriptions to my methods. I write better code, documentation, and tests.Copilot has made me a better programmer. No kidding. This is a huge achievement. Kudos to the GitHub Copilot team!

So now I make a bot to upload repositories of intentionally buggy code so that when people blindly use this autocomplete my hacking becomes easier!

If you are interested to read[1] more about it.[1] <a href="https:&#x2F;&#x2F;docs.github.com&#x2F;en&#x2F;early-access&#x2F;github&#x2F;copilot&#x2F;research-recitation" rel="nofollow">https:&#x2F;&#x2F;docs.github.com&#x2F;en&#x2F;early-access&#x2F;github&#x2F;copilot&#x2F;resea...</a>

How does this compare to TabNine or Kite?

So this is basically fully automated Stackoverflow assisted programming? Excellent.

&gt; What are some ethical problems that could emerge from the box?Being put out of job by an AI trained on your own code?It&#x27;s really the same ethical problem of all automation ... and will be as long as we need a job to fulfill basic needs like food, housing and medical care.

There are some really weird licensing problems. Like does your code license say they can use the your code to train AIs that then reproduce very similar code to you but with no attribution etc in someone else&#x27;s codebase.

What are some ethical problems that could emerge from the box? Maybe unfair competition from having very good tools compared to other programmers, or havin irresponsibly shallow understanding of what the produced code does?

s&#x2F;OpenAI&#x2F;PhotoshopReads similarly :)

&gt; Maybe we&#x27;ll start seeing licenses with a section saying &quot;not for use as training data for commercial models.&quot;Considering that the impact of a single example is extremely small in training a model, and that it is trained on an ungodly amount of examples, then I wonder if the effort of forbidding its use has any real benefits.

&gt; Yet, isn’t that effectively what human brains do too?If I want to watch a bunch of movies, I have to pay the theater for each movie, or pay netflix, or whatever. The screenplay I write afterwards belongs to me, but the learning process involved me paying for access to others&#x27; work. That&#x27;s what&#x27;s often missing here. But at the same time, if you train on legally public data, there&#x27;s no &#x27;theater&#x27; to be paid.(Often, people train on illegally public data though, like the eleuther folks. That&#x27;s a whole extra can of worms I&#x27;ve ranted about plenty).Maybe we&#x27;ll start seeing licenses with a section saying &quot;not for use as training data for commercial models.&quot;

This is very impressive!OpenAI’s tech opens an ethical Pandora’s box:1. It’s clear that the raw inputs to all of OpenAI’s outputs originated with real, human creativity.2. So, in a sense, OpenAI is laundering creativity. It reads in creative works, does complicated (and, yes, groundbreaking) transformations, and produces an output that is hard to trace to any particular source.3. Yet, isn’t that effectively what human brains do too? Perhaps OpenAI lacks the capacity for true invention, but I’d argue that most people live their whole lives without a meaningful creative contribution as well.All told, I don’t have a good framework for thinking about the ethics here. So instead, I’ll simply say:Wow.

I haven&#x27;t seen any reports of this, but it&#x27;s certainly something we want to guard against: <a href="https:&#x2F;&#x2F;copilot.github.com&#x2F;#faq-can-github-copilot-introduce-insecure-code-in-its-suggestions" rel="nofollow">https:&#x2F;&#x2F;copilot.github.com&#x2F;#faq-can-github-copilot-introduce...</a>

Have there yet been reports of the AI writing code that has security bugs? Is that something folks are on the lookout for?

It shouldn&#x27;t do that, and we are taking steps to avoid reciting training data in the output:
<a href="https:&#x2F;&#x2F;copilot.github.com&#x2F;#faq-does-github-copilot-recite-code-from-the-training-set" rel="nofollow">https:&#x2F;&#x2F;copilot.github.com&#x2F;#faq-does-github-copilot-recite-c...</a>
<a href="https:&#x2F;&#x2F;docs.github.com&#x2F;en&#x2F;early-access&#x2F;github&#x2F;copilot&#x2F;research-recitation" rel="nofollow">https:&#x2F;&#x2F;docs.github.com&#x2F;en&#x2F;early-access&#x2F;github&#x2F;copilot&#x2F;resea...</a>In terms of the permissibility of training on public code, the jurisprudence here – broadly relied upon by the machine learning community – is that training ML models is fair use. We are certain this will be an area of discussion in the US and around the world and we&#x27;re eager to participate.

Might this end up putting GPL code into projects with an incompatible license?

This looks really cool. Do you plan to release this in some other form like a language server so that it can be easily integrated to other editors?

How many jobs have developers helped displace in business and industry?  I don&#x27;t think it&#x27;s controversial that we become fair game for that same automation process we&#x27;ve been leading.

We think that software development is entering its third wave of productivity change. The first was the creation of tools like compilers, debuggers, garbage collectors, and languages that made developers more productive. The second was open source where a global community of developers came together to build on each other&#x27;s work. The third revolution will be the use of AI in coding.The problems we spend our days solving may change. But there will always be problems for humans to solve.

&gt; This is obviously controversial, since we are thinking about how this could displace a large portion of developers.It... couldn&#x27;t, in net.Tools which improve developer productivity increase the number of developers hired and the number of tasks for which it is worthwhile to employ them and the market clearing price for development work.See, for examples, the whole history of the computing industry as we’ve added more layers of automation between “conceptual design for software” and “bit patterns in hardware implementing that conceptual design as concrete software”.It might displace or disadvantage some developers in specific (though I doubt a large portion) by shifting the relative value of particular subskills within the set used in development, I suppose.

Intel put itself out of business since all the world&#x27;s computing needs can now be done on one CPU.Or, perhaps, lower unit costs of computing lead to far far greater demand for computing since it became practical to apply to more domains.

In just 2-3 years time DL algorithms will have as many synapses (parameters) as the human brain. The only thing needed to teach such algorithm to program better than us it to say to it that program needs to be faster, more secure and more user friendly (it&#x27;s not impossible to teach this to DL alg.). This &quot;tool&quot; will make our work 90% easier in the next few years, so unless we have much more work to do we will earn much less and juniors will most likely be not needed anymore...

This is obviously controversial, since we are thinking about how this could displace a large portion of developers. How do you see Copilot being more augmentative than disruptive to the developer ecosystem? Also, how you see it different from regular code completion tools like tabnine.

It has the ability to generate unit tests as well, which will help cut down some on the verification side if you feed it enough cases.

&gt; I had to spend a bit of time to see if it was actually correct.Interesting point - it reminds me of the idea that it’s harder to debug code than to write it. Is it also harder to interpret code you didn’t write than to write it?

Interesting comment and I agree. Reading and writing code seem to involve different parts of the brain. I wonder if tools like this will create some sort of code review fatigue. I can write code for a few hours a day and enjoy it but I couldn&#x27;t do code review for hours, every day.This isn&#x27;t like skimming through a codebase to get a sense of what the code does. You&#x27;d have to thoroughly review each line to make sure it does what you want it to do, that there are no bugs. And even then, you&#x27;d feel left behind pretty quickly because your brain didn&#x27;t create the paths to the solution to the problem you&#x27;re trying to solve. It is like reading a solution to a problem on leetcode vs coming up with it yourself.

I visited <a href="https:&#x2F;&#x2F;copilot.github.com&#x2F;" rel="nofollow">https:&#x2F;&#x2F;copilot.github.com&#x2F;</a>, and I don&#x27;t know how to feel. Obviously it&#x27;s a nice achievement, not gonna lie.But I have a feeling it will end up causing more work. e.g. the `averageRuntimeInSeconds` example, I had to spend a bit of time to see if it was actually correct. It has to be, since it&#x27;s on the front page, but then I realized I&#x27;d need to spend time reviewing the AI&#x27;s code.It&#x27;s cool as a toy, but I&#x27;d like to see where it is one year from now when the wow factor has cooled down a bit.

I have been using this - for example working in Go on Dapr (dapr.io) or in Python on one of its SDKs.I love it. So often the code suggestions accurately anticipate what I planned to do next.It&#x27;s especially fun to write a comment or doc string and then see Copilot create a block of code perfectly matching your comment.

Hi Nat! Just signed up for the preview (even though I&#x27;m the type to turn off intellisense and smart indent). I was wondering if WebGL shader code (glsl) was included in the training set? Translating human understandable graphics effects from natural language is a real challenge ;)

Cool project! Have you seen any interesting correlations between languages, paradigms and the accuracy of your suggestions?

<a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Profession_(novella)" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Profession_(novella)</a>An Isaac Asimov story about someone who didn&#x27;t take to the program and, as a result, got picked to create new things because someone has to make them.

Are those developers worried about having their jobs replaced by a code-writing AI? :)I mean... why would 95% of developer jobs exist with this tech available?You just need that 5% of devs who actually write novel code for this thing to learn from.

Why would you say it&#x27;s an error to use a float for currency? I would imagine it&#x27;s better to use a float for calculations then round when you need to report a value rather than accumulate a bunch of rounding errors while doing computations.

How is it different from the status quo of people just doing the wrong thing or copy pasting bad code? Yes there&#x27;s the whole discussion below about float currency values, but I could very well see the opposite happening too, where this thing recommends better code that the person would&#x27;ve written otherwise.

I use a similar feature in IntelliJ idea, and I&#x27;ve often found that first time I learn about new feature in the language is when I get a suggestion. I usually explore topic much more deeply at that time. So far from helping me copy-paste, I find code suggestions help me explore new features of the language and framework, that I might not have known about.

&gt;The first example parse_expenses.py uses a float for currencyI have made this mistake without the help of any AI or copy&#x2F;paste. It&#x27;s still in the hands of the developer to test and review everything they commit.

What do you think about this being overall detrimental to code quality as it allows people to just blindly accept completions without really understanding the generated code. Similar to copy-and-paste coding.The first example parse_expenses.py uses a float for currency - that seems to be a pretty big error that&#x27;s being overlooked along with other minor issues around no error handling.I would say the quality of the generated code in parse_expenses.py is not very high, certainly not for the banner example.EDIT - I just noticed Github reordered the examples on copilot.github.com in order to bury the issues with parse_expenses.py for now. I guess I got my answer.

You should read the FAQ at the bottom of the page; I think it answers all of your questions: <a href="https:&#x2F;&#x2F;copilot.github.com&#x2F;#faqs" rel="nofollow">https:&#x2F;&#x2F;copilot.github.com&#x2F;#faqs</a>

&gt; - under what license the generated code falls under?Is it even copyrighted? Generally my understand is that to be copyrightable it has to be the output of a human creative process, this doesn&#x27;t seem to qualify (I am not a lawyer).See also, monkeys can&#x27;t hold copyright: <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Monkey_selfie_copyright_dispute#Naruto_et_al_v._David_Slater" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Monkey_selfie_copyright_disput...</a>

Some of your questions aren&#x27;t easy to answer. Maybe the first two were OK to ask. Others would probably require lawyers and maybe even courts to decide. This is a pretty cool new product just being shared on an online discussion forum. If you are serious about using it for a company, talk to your lawyers, get in touch with Github&#x27;s people, and maybe hash out these very specific details on the side. Your comment came off as super negative to me.

In general: (1) training ML systems on public data is fair use (2) the output belongs to the operator, just like with a compiler.On the training question specifically, you can find OpenAI&#x27;s position, as submitted to the USPTO here: <a href="https:&#x2F;&#x2F;www.uspto.gov&#x2F;sites&#x2F;default&#x2F;files&#x2F;documents&#x2F;OpenAI_RFC-84-FR-58141.pdf" rel="nofollow">https:&#x2F;&#x2F;www.uspto.gov&#x2F;sites&#x2F;default&#x2F;files&#x2F;documents&#x2F;OpenAI_R...</a>We expect that IP and AI will be an interesting policy discussion around the world in the coming years, and we&#x27;re eager to participate!

Lots of questions:<pre><code> - the generated code by AI belongs to me or GitHub?
 - under what license the generated code falls under?
 - if generated code becomes the reason for infringment, who gets the blame or legal action?
 - how can anyone prove the code was actually generated by Copilot and not the project owner?
 - if a project member does not agree with the usage of Copilot, what should we do as a team?
 - can Copilot copy code from other projects and use that excerpt code?
 - if yes, *WHY* ?!
 - who is going to deal with legalese for something he or she was not responsible in the first place?
 - what about conflicts of interest?
 - can GitHub guarantee that Copilot won&#x27;t use proprietary code excerpts in FOSS-ed projects that could lead to new &quot;Google vs Oracle&quot; API cases?</code></pre>

Has this been tested for accessibility yet, particularly with a screen reader?

Is there a public API? Will it be documented? Are you open to folks porting the VSCode plugin to other editors (I.e. kakoune’s autocomplete)?

&gt; For a seasoned veteran writing the code was never really the hard part in the first place.Yes, to most coders this Copilot software is just a fancy keyboard.

I&#x27;m glad they find it head exploding but my concern is that it would be most head exploding to newbies who don&#x27;t have the skill to discern if AI code is how it should be written.For a seasoned veteran writing the code was never really the hard part in the first place.

Ok, we changed the URL to that from <a href="https:&#x2F;&#x2F;github.blog&#x2F;2021-06-29-introducing-github-copilot-ai-pair-programmer&#x2F;" rel="nofollow">https:&#x2F;&#x2F;github.blog&#x2F;2021-06-29-introducing-github-copilot-ai...</a>.

Only if they trained a model to be able to read and understand LICENSE.txt files -- wowzers what a monster improvement that would be for the worldOr, I guess a sentinel phrase that the scraper could explicitly check: `github-copilot-optout: true`

&gt; If I put a section in my LICENSE.txt prohibiting use as training data in commercial models, would that be sufficient to keep my code out of models like this?Neither in practice (because it doesn&#x27;t look for it) nor legally in the US, if Microsoft’s contention that such use is “fair use” under US copyright law.That “fair use” is an Americanism and not a general feature of copyright law might create some interesting international wrinkles, though.

In the end this would slightly increase likelihood of such sections appearing in licenses generated by AIs.

Don&#x27;t make your code public. Someone could read it and train the model in their brain to synthesize some code based on it.If its publicly available than its fair game to use it to learn and base ideas on.

If I put a section in my LICENSE.txt prohibiting use as training data in commercial models, would that be sufficient to keep my code out of models like this?

One question: how long is the waitlist? Very excited to try this!

This is impressive. And scary.
How long has your team been working on this first release?

I assume this will also work alright on other languages that aren&#x27;t in the demo, e.g. C++?

The technical preview document points out that editor context is sent back to the server not just for code generation but as feedback for improvement. Are you (or OpenAI) improving the ML models based on the usage of this extension? It is interesting what the pricing will look like given that the model was originally trained on FOSS and then you go and harvest test cases from real users. If that’s the case I think that should be clearly explained upfront.

Hi HN, we&#x27;ve been building GitHub Copilot together with the incredibly talented team at OpenAI for the last year, and we&#x27;re so excited to be able to show it off today.Hundreds of developers are using it every day internally, and the most common reaction has been the head exploding emoji. If the technical preview goes well, we&#x27;ll plan to scale this up as a paid product at some point in the future.