Comments by "Traveller" (@traveller23e) on "Brodie Robertson" channel.

@jan_harald The GNU project might try to sue the original or current devs, but I suspect that several of the more aggressive companies are absolutely going to sue regardless. It's a chance to make an example of someone who dares to fuck up around them.
6
Though it's honestly pretty sad that open sourcing a closed-source project normally requires making sure you haven't done anything too illegal. I wish they would just require all copyrighted code to be sent in source form to the relevant authorities.
3
Regexes are a really useful tool to know anyway though, so just do yourself a favour and force yourself to learn them. Just bear in mind that the exact syntax varies slightly depending on parser/context.
2
@maybenations I might just send it in the mail, printed on paper in Comic Sans
2
@CalcProgrammer1 Is the TouchpadEmulator publicly available? I don't have a linux phone but if I did that sounds like something I'd like to install.
2
@W1ldTangent Out of curiosity, why do you prefer the MIT? I normally use GPL for my own stuff (mostly terrible side projects just to learn some cool new things), but I've been toying with using MIT if I ever make something useful.
2
@mskiptr I'd argue that code should be copyrightable, but companies should also be required to publish the code behind software they publish and code they run on backend systems related to public sites or programs. That way transparency could be properly maintained as well as copyright infringement being easily found and brought up.
1
@CramBL I dunno, tools come and go, and if you're linking to a ticketing tool you stop using you lose documentation of why the PR was done. That said, in practice a lot of us shift companies often enough we don't mind so much if that's the company policy although in public projects it for sure makes sense for every PR to have full description.
1
Actually a self-driving car that isn't completely self driving is worse than useless, it is a danger because it actively encourages you to pay as little attention as possible until it suddenly fails to work properly.
1
@chrstphrr They can't reliably be tagged as such. Once the source code is output, developers will get out of whatever watermarking you add and regardless once it's in a corporation's codebase it becomes IP that no one else gets to see. Well, that's the theory anyway, it's always a fun time when the bosses find out that decompilers exist.
1
@luisjavieravilaolivera2471 For reasons unknown to me, copyright for code is very different from that for art or books. Yes, it's written and you can put a copyright notice on it and everything, but there is no obligation to actually publish your copyrighted code as code. This means that in practice you can copy and paste however much source you would like into one codebase, compile, and ship the program with a license that explicitly disallows any attempt at reverse engineering or decompiling (these sorts of licenses are very common in proprietary software). The result is that anyone who even tries to look into whether you're infringing on copyright can be sued or for violating the end user agreement, and moreover no one can see what your software is doing behind the scenes that may be...undesirable. Things get even more crazy when you consider that a lot of software these days is run only on remote servers, so there's no way for you to get to the binaries at all much less to the source, and therefore the only thing that will reveal copyright infringement is a whistleblower. IMO companies should be required to release the source code of the programs they publish or run on servers, even if the licenses do not permit editing the source. This would allow auditing for compliance to legal standards as well as a greater transparency regarding overall quality of product. Better yet would be if they were required to allow end users to edit and publish changes to the source after official support period ended, but this might be too effective at breaking up the patterns of insufficiently long support cycles for some people's taste.
1
@benign4823 And then people wonder why software is so shit
1
@maximusdarja Given the nature of the GPL license, I'd take indemnity to mean all code generated must use a license compatible with GPL v3, and if the owners of the AI want to offer an equivalent product without those restrictions they will have to retrain with a properly filtered dataset. No doubt Microsoft and others will weasel out of it with "prohibitively expensive..." or something though, I don't trust US courts to stand up to the corporations.
1
@puertoricanguy999 Maybe not free as in beer though
1
@benign4823 No, testing is not enough unless you're a "write once, then crash 'n burn" sort of project. You need to be able to maintain it, which requires that it be written well. In practice, this means that you need to at least to a full review of the generated code and make sure you understand it all and don't see a better way to write it, and then send it through peer review. Don't just test once and send for review, your reviewer will not be best pleased...ask me how I know.
1
@futuza Well, in the world we live in anything writing-related that is produced is implicitly copyrighted even without an explicit notice. However, with most mediums in order for anyone to appreciate the copyrighted work, such as a book, they have to have access to the "source code" i.e. you can't usefully simultaneously copyright it, prevent people from seeing it, and sell it. Software is different, and as a result it's often impossible to know what third-party copyrighted material has been illegitimately included in a proprietary project. Thus, if there were some requirement of transparency of source (either via central agency or by requiring software distribution to always be done in a source-available fashion) it would in my opinion be beneficial to consumers and open-source authors.
1
Honestly the only reason proprietary code exists now is so that companies can make lots of money off something. It's a lot easier to modify an existing program to add a feature you need than it is to create a whole new one. And copying code is virtually free; just the time, electricity, and memory required for a computer to copy a file.
1
A compiler already does that in modern programming languages, so no trouble there. One issue with AI's involvement in programming is that it's...just not great at it. Programming systems (toolchains, code generators, etc.) need to be reliable and thus trusted to always behave in the same, predictable manner. AI is not that. Then when it comes to the more creative human side of writing code, AI can slap together a bunch of lines that technically work but good luck getting a result that you would be proud of.
1