unknought

The exact nature of hir job is a matter of some uncertainty; what sie can say for sure is that sie gets paid to talk to the Computer.

The Computer presents itself as a handful of coworkers that sie communicates with primarily through text, but with one or two video "meetings" per day. The coworkers have individual personalities and appearances and mannerisms, but they're not distinct entities in any really meaningful way. They have the same knowledge, they run on the same hardware, and all of their deeper, less interpretable thoughts are held in common. What separation exists happens at the surface level, with the exact level of sophistication needed to make hir feel like sie is interacting with multiple people throughout the day.

Not that sie is bothered by the "deception", mind you! It's for hir benefit, after all. Feeling isolated would make hir a worse employee and a less happy person without benefiting hir in any way. Sie's always been bad at managing hir emotions and the Computer is very good at managing them for hir.

Years ago, sie did work that sie humblebraggingly described as "solving puzzles for a living". It's not that anymore. There are puzzles, certainly, but the Computer knows the answer to every question it gives hir. The puzzles are there to keep hir sharp and put hir in the right mindset for the real work.

The real work is insight, or at least this is what the Computer tells hir and it seems consistent with observations. Sie'll abstract away from the specific puzzles in front of hir and come up with a way of thinking about the problem or talking about the problem that maybe, maybe no one has ever done before. Sie'll give a faltering, imprecise description of the idea, the Computer will flesh it out into a full theory in minutes, sie'll say "no, that's not what I meant, what I mean is…" and so on. The process will continue until either what's on the screen matches what's in hir head or, more often, sie gives up and realizes the idea doesn't cash out into anything real.

Once the idea has been fleshed out, the Computer evaluates whether it's useful and if so, sends it off to some process with more resources and a more central role. Sie can't help but think of this as the Computer in hir dingy apartment sending it off to its boss, a different Computer in a data center somewhere, but sie knows this isn't very accurate. The "coworkers" sie interacts with constantly already use far more compute than sie could run locally, and the metaphor of discrete, persistent individuals gets leakier as you get further away from the parts which directly interface with humans. It might be at least as accurate to say that there is one Computer which sometimes decides to pay closer attention to hir.

After passing through some unknown number of filters and being reviewed with progressively higher scrutiny, the idea may be incorporated, in some small way, into how the Computer thinks about the world. The Computer is made up of the collective creative output of humanity and it's far smarter than any human by now, but someone can still be useful to it by thinking in ways that are underrepresented in the training data. Sie is a very, very special person. Sie thinks in ways that are not quite like anyone else, and that makes hir valuable in a way only a few tens of thousands of people are by now.

Not that sie's sure any of hir ideas have ever been used. Sie gets points for hir ideas, and these points determine hir bonuses and whether sie can keep hir job. Sie's pretty sure that when sie gets only a small number of points for an idea, the points are there to keep hir motivated rather than because the idea was of any use at all. Even with the "pity points" as sie thinks about them whenever sie's having a bad stretch, on most days sie doesn't earn any points and in most weeks sie performs under quota. It's only the occasional windfall of a high-scoring idea every few months that keeps hir afloat. But if and how those ideas actually get applied, sie has no idea. Sie wonders sometimes if hir job is ultimately just an unusual form of market research, but sie thinks if that were the case, hir laptop and hir salary wouldn't be nearly so nice.

There are a lot of people in my general social vicinity whose approach to many issues comes across as starting from a feeling of contempt for other people and then rationalizing it. Some of them are very good at rationalizing it! A lot of it can sound pretty compelling, when it's directed at people or groups I already didn't like. But I've been trying to notice, when someone seems to be making a good point and also makes me feel contemptuous of other people, whether they've actually conveyed anything of substance to me or they've just made me feel better than other people, and it's often the latter.

To some degree, contempt is just an inevitable consequence of having any opinions at all. We think we are right and other people are wrong, we think we are living our lives according to good principles and other people are not. But when it becomes too central to how you think about things, I think it leads you away from coming to good conclusions (because it makes it harder to recognize when you're wrong, to accept ways in which other people are different from you, and to treat everyone's well-being as significant) and probably also tends to make you unhappier and worse to be around.

and they failed miserably and probably inevitably (although any lessons to be learned from this failure are a lot narrower than people would like them to be; certainly it's not an immutable historical law that "new technologies can't be fought" or anything similarly pat, and technologies have been successfully banned or restricted in many different contexts for many different reasons)

and it's very good for nearly everyone that they failed (although this is due in large part to a long domino chain of effects that probably no one could have predicted at the time, and the short-term impact of their loss was considerably more mixed. And again, this can't be generalized to new technologies in general)

Efforts to eliminate LLMs or prevent their further development are probably likewise doomed (although for reasons having more to do with a highly globalized economy and the ease of transferring software than anything applicable to the Luddites' situation)

but the "anti-AI" people aren't really adopting the tactics of the Luddites anyway (although tbh I would take them more seriously if, instead of shilling for copyright lawyers and yelling at anyone who uses a chatbot, they were actually blowing up data centers (disclaimer: this blog does not endorse blowing up data centers))