@joepie91 "AI" was always supposed to automate away the tedious stuff.

What we are doing now is automating away art, as well as putting folks out of their jobs before realizing we still needed them (and putting them in misery because not having job under capitalism = pretty fucking bad because no income leads to poverty eventually)

@raito The thing is, I don't see why point 1 would be in any way desirable, and point 2 doesn't seem to be what people actually are doing - instead they're trying to 1:1 emulate how human brains work, warts and all.

@joepie91 I don't think it's desirable, but rationality did not prevent certain undesirable things to be pursued in the history of humankind, right?

as for point 2, the overall design started by doing biomimicry and all that stuff because human has some insights that are not well understood and mimicking is good for obtaining the results without understanding the mechanisms, belief is "more progress" will give cheaper human-like thinking

@joepie91 (sorry, should have had that the last paragraph is not my opinion but my understanding of what people believes)

@raito This seems like it just circles back to "labour exploitation" again, honestly.

@joepie91 well, the implicit is moving the labour exploitation from genuine humans to artificial humans I'd assume?

as with military drones: moving losses from genuine humans to artificial "humans"

@raito It doesn't, though; unless there is a way for people to live without a job, all it will do is artificially increase competition for the same jobs, putting human workers into a worse negotiating position (= worse pay, worse work environment).

Military drones don't move the losses to artificial 'humans' either; the casualties on the other end are still just as human, and the scale of them increases as automation is applied.

@joepie91 well that's the inherent contradiction in what capitalistic-aligned folks have been pushing

to justify pushing for innovations and funding them, you say "oh but look life is going to be easier"

but then once lines of work have been successfully automated away or extinguished, the people who were in there were told to gtfo, no sharing of the generated wealth

in the countries where they decided to do it, there consistently were voices to remove social security again

@joepie91 universal basic income and all that stuff is really all the same

there's less and less work available for the society, but at the same time, everyone wants people to work more and more

At the same time, we do everything we can to automate work but we refuse to abandon the accomplishment of lifehood uniquely via the lens of "work", "career" etc.

Ultimately, with this trajectory, one can wonder whether the end result will be a net reduction of the size of humankind

@joepie91 finally, on the last paragraph, you are spot on to me

this is the blind spot for all these enthusiasts of making war casualtyless on the side of the Right™ and whatever for the side of the Wrong™

but also war has became professional, human losses are not acceptable anymore in public opinion, therefore… the march towards automation and artificial assets that does war by proxy seems unstoppable

@virtulis But we already *have* less squishy and less biased systems; computers can do that just fine, given careful application.

So why do we need "AI"?

Especially considering that trying to recreate sentience is more likely to *reintroduce* bias, because biases, however frustrating, do actually play a role in survival and development.

@joepie91 well, exactly that. We could attempt engineering a "good" bias. I don't think we even need "laws of robotics", just basic empathy, which is way easier than sentience (see dogs).

By less squishy I imagine:

Not requiring oxygen or human energy sources, or otherwise more environment-resilient. On Earth easily solvable with remote control, but as latency grows some*one* better suited for it might be worth considering.
Serializable and/or modifiable. Again, I think engineering a "re-printable" sentience from scratch is probably easier than the magical instant human backup and cloning trope. So some*one* that can literally "unsee" things without years of therapy. Quite abuse-prone though.

Yeah I think that's about all I got.

@joepie91 tangent: I think it's interesting to think in terms of probabilities disregarding time sometimes.

Like, objectively and realistically speaking, artificial sentience, faster-than-light travel/comms, cryosleep/cloning/etc all seem equally fantastic and unlikely to ever happen (let alone our lifetime).

But also objectively, one of these is not like the others, because sentience exists and is not magic, so the question is only of making some more of the same. So idk, I guess it's on the very fringe of something worth talking about sometimes?

But definitely not with those people.

@fionafokus you mean like... like rokos basilisk but every team believes in a different basilisk they work to materialize before the others get theirs?

@joepie91
I think that crucial level of introspection requires (or itself actually is) a form of general intelligence.

@silvermoon82 I feel like this elides the much more important question: why do we need this? Why aren't we instead building better (technical and social) structures for the experts we *already have* and who *already* want to help others, to provide that kind of assistance?

@joepie91
"Why do we need this" vs "why do we need to make this" are two very different and important questions.
I think humans have a certain drive to be tool makers; even if the tool is unnecessary or harmful, the act of making it satisfies something human.

@sofia Is all of this actually about *artificial* intelligence, though? Because it reads more to me like it is about augmentation of natural intelligence.

(Aside, "AGI" is a term that didn't exist until pretty recently, and it's what "AI" used to mean originally, also in modern AI research)

@joepie91 sure, it doesn't really matter how you get to it (though getting there in more than one way is probably desirable, because of robustness and diversity).

i feel like many of these things seem intuitively easier to implement in software, but i woudn't complain if we had, like, brain eingineeers working on superintelligence either. copying them for backups and sharing seems among the rather more tricky aspects.

@joepie91 i think in the early days AI folks tended to massively underestimate the complexity of seemingly easy thought processes. what their shiny new toys coud do just seemed way more impressive than recognizing a picture of a cat, etc.

the term "strong AI" and superintelligence are older. still i think if you asked AI folks back then they would say the things they make are AI, nor precursors or attempts at AI.