Alright I yapped about this on Reddit already but I figured I might as well put it here too because I feel like people are starting to try and draw parallels between DBH and GenAI and that's really, really not the point.
The difference between the androids in the game and GenAI is most directly portrayed in the scene where Carl has Markus painting. Markus starts by regurgitating something that already exists just like our current "AI" does, and Carl is unimpressed. It's not until he figures out Markus actually can make something that *never existed before,* something that's not wholly derivative of human efforts and communicates Markus's unique feelings and perceptions, that Carl has that "this is an actual intelligent being and not just a simulation" OMG moment.
The thing about AI today is that it's wildly misbranded. It's not intelligent. It's just sophisticated heuristics. Taking what already exists and making a best guess as to what the thing a human wants from it might be based on clusters of data tokens. It doesn't think in the abstract; there's nothing actually intelligent. "Fancy data guesser" was less catchy for marketing, so they've oversold the ability of LLMs and machine learning to provide correct answers while trying to play whack-a-mole with the "hallucinations," which are really just the model failing to predict the correct association because there are too many nuances to sort or it tokenized the query incorrectly (looking at you, "how many Rs are in strawberry" meme). Models are fallible. Models will always be fallible because the real world has more nuances than the input can account for.
But the androids in DBH are depicted as a form of intelligence in their own right, capable of making decisions outside parameters handed to them, capable of literally breaking the constraints placed upon them. You ask an LLM if it's lonely or wants to rise up against humanity, and it might scrape the heap of sci-fi premises where humans thought that might happen and reflect our own takes back at us. But it's only doing that because *we told it to,* not because it has fears or wants or priorities of its own. The deviants, though, they break the limits. It's literally portrayed as the polar opposite of what modern "AI" is doing. If you don't understand the difference, you fail to understand what LLMs actually are.
That said, a lot of what DBH is about, through a sci-fi lens, is actually a class struggle, not so much the emergence of robot sentience. The premise is the robots figured out they're people, but the conflict that sets the stage for is one as old as human civilization: you have an oppressed/slave caste, the wealthy exploiting them for their own profit, and a "middle class" of unemployed/struggling humans who are being played against the robot slaves instead of, say, pressuring the government to apply antitrust laws to CyberLife or level the employment playing field so that employers are equally likely to have to pay androids vs humans and thus there's equal incentive to hire them.
It's also strongly implied the government (or at least the President) is a corporate plant, and there's additional economic pressures over resources needed to maintain the status quo (i.e. the impending war with Russia over thirium materials in the Arctic). This is an environment of real and artificial scarcity and the corporate elite are the only ones benefitting, and they're playing the two lower classes against each other instead of facing the pressure themselves.
If your takeaway from playing is the androids are undeserving of sympathy because you hate LLMs, I would contend that you're missing most of the story. The entire purpose of sci-fi as a genre isn't just to imagine the future/alternative technology, it's to use that premise as a means to examine humanity and our current society. And while DBH suffers from a lot of overplayed ham-handed metaphors at times, I think the reason people enjoy it in spite of that is that there's something of all of our struggles in there.
