[superb-owl]

There's actually a really interesting question here - could it be possible to "pirate" a backend with sufficiently clever AI?

At the end of the day, you're just trying to model a black-box function, mapping inputs to outputs. And most of that is CRUD with some basic access control on top. There are definitely complications (e.g. 3rd party integrations, a properly designed/named database schema), but you might be able to get 80% of the way there in an automated way...

[laumars]

People reverse engineer back end servers for online games all the time. You don't need AI. Just a really dedicated following and a lot of free time.

It's not piracy doing this though. Technically you might still be in breach of some intellectual property but since it's usually discontinued services a lot of games publishers turn a blind eye.

[tmp_anon_22]

> You don't need AI

I grew up in the private server scene for a popular MMO and you're absolutely right. It was a whole lot of teenagers with energy drinks grinding through reverse-engineering minutia that adults would gawk at and make excuses to avoid doing.

That said there was a lot of automation, scripts, and other tooling, to make it easier. The best were able to i.e. update a private server automatically when the base game updated. We were doing automation at a higher level then F500 companies were at the time (mid 2000s) and we were just kids.

[kelnos]

> It was a whole lot of teenagers with energy drinks grinding through reverse-engineering minutia that adults would gawk at and make excuses to avoid doing.

Right. I don't think the question is if you need AI (obviously you don't), but if AI could do all the annoying, tedious bits for us, and speed up the process.

[laumars]

I doubt it. You'd probably end up spending more time training the AI than you would using it.

To be clear, people do this stuff as a passion project so nothing stopping someone from investing the time in training a ML model to assist here if that’s something that sounds like fun to them. they wanted to take on. So from a technical standpoint one “could” use AI. But I’d expect you’d first have to train the AI to play game before you can even think about training it to read the network packets. And the former is a far more daunting problem than the latter.

Sometimes problems are better solved with human intelligence, a lot of automation, and patience.

[CRConrad]

That kind of is "artificial intelligence" -- the energy drinks artificially (and temporarily) bump up the teenagers' IQ.

[kmeisthax]

"Sufficiently clever AI" would, in this case, be the person writing a reverse-engineered work alike app.

In some of the cases mentioned (e.g. Spotify, Chegg, etc) you can't really do this, because the actual value in the app is just the copyrighted material being purchased. Reverse-engineering is protected under US law for a variety of reasons, mostly that you can't copyright basic functionality (that's for patent law) and that copyright shouldn't extend to interfaces[0].

AI trying to reverse-engineer all of music or art or writing already exists. They're called MuseNet, DALL-E, and GPT-3 respectively. While you can sort of trick them into regurgitating training data in a way that would make their use to create novel works legally perilous, it's still kind of difficult to get them to generate exact copies in a way that would be useful for "pirating" all of Spotify.

[0] SCOTUS tried very very hard in the Google v. Oracle decision not to actually say this. However, the actual ruling has a similar effect.

[derangedHorse]

That would be considered a remake though and would essentially be a competing product with the same api (since it’s highly unlikely the remake would be written the same as the original). Whether copying the api is infringing on anything seems to be uncertain as the outcome of the Google v Oracle case seemed to only set a light precedent