Artificial intelligence still has some way to go

Okay I had to try to track that Elizabeth quote down: https://conversableeconomist.blogspot.com/2019/03/the-story-of-william-lee-and-his.html

Just wait till AI CEOs have to deal with AI peons unionizing...
https://www.youtube.com/watch?v=QFfaL3_Mv0s

It seems more likely that companies charging usurious royalties and licensing on monopolized AI models will be the thing that keeps human labor a thing far past the point where anyone would really want to do it.

Okay I had to try to track that Elizabeth quote down: https://conversableeconomist.blogspot.com/2019/03/the-story-of-william-lee-and-his.html

― what you say is true but by no means (lukas), Tuesday, November 28, 2023 5:27 PM (four minutes ago)

ha, thanks, that's actually even better, I didn't love having to hand it to Elizabeth I

I just want to humourlessly point out "AI CEOs" would mean software optimized to maximize shareholder value making decisions about layoffs, plant locations, personnel policies, wages, etc. Please understand I'm not saying human CEOs are good, just that being ruled by an algorithm could be worse!

― rob

ok, well, look at the selection process and expectations of ceos now

i think it would be fair to say that human ceos are people optimized to maximize shareholder value making decisions about layoffs, plant locations, personnel policies, wages, etc

now

who is reaping the benefit from those processes

with human ceos, you can say, ok, sure, it's ceos themselves

having an "ai ceo", to me, that doesn't seem better _or_ worse. capital can't be bothered to exploit workers themselves, so they've programmed an ai to do it for them. as an added bonus, blaming "ai" for their own exploitative actions helps shield them from responsibility for those actions.

if you were really going to "replace" a human ceo with an AI, the profit that would have gone to that ceo ought logically to go to the workers, right? but that doesn't make sense, because... i mean, this stuff ought to be data-driven, right? is there _any_ evidence that CEOs create anything of value? legitimate and serious question, are decisions made by human ceos objectively better than decisions made by a fucking random number generator?

i've come up with a brilliant new ai to replace landlords, you pay it money and it does nothing

AI Christ

i once read absalom, absalom! in one sitting. (weird period in my life emotionally). can ai do that????

― treeship., Tuesday, November 28, 2023 8:44 AM (fourteen hours ago) bookmarkflaglink

lol thats out there, salute

AIbsalom, AIbsalom! is that anything

AbsAIlom

xp i want to lol but come on, it is better. Tho kind of funny on its own terms for being self-conscious and try-hard.

― Ethinically Ambigaus (Bananaman Begins), Tuesday, 28 November 2023 bookmarkflaglink

I like Faulkner's fiction but that's hack work. Automating it isn't the end of creativity.

I can see a scenario where a writer automated this hackwork under their name, leaving them to write what they want to write. This is a positive.

it is hackwork, but it has some charm, some humanity.

one problem for me might be aesthetic. (there are also other, bigger problems). living in a world filled with formulaic advertisements and other kinds of content made by people working under a deadline is not great. it might be similar --- perhaps unrecognizably so --- to live in a world where all that junk was made by robots. but there is an invisible difference, an uncanniness, an eeriness to knowing that the messages that surround you were made by robots and not people.

when i read something online -- like a facebook comment or something --- and, halfway through, i realize that it was written by ai, i feel betrayed. it's hard to describe. the internet right now is populated by bots, so like accounts that pose as human beings and do an ok job impersonating them. that is deeply weird and unsettling if nothing else.

if i believed, as you suggested, that automating all this content production will free people up to be creative, i would accept it as a reasonable price to pay. but i don't think we have the right political economy for that. i don't think workers are going to be the beneficiaries of this wave of automation.

"it is hackwork, but it has some charm, some humanity."

I'd be surprised if Faulkner liked doing it.

People have said for the last 20 years or more that automation would bring the end of jobs and so on.

I've yet to see anything like this. The nearest is offshoring but that's people doing the work somewhere else for lower cost. That can brutally shift if economies go under.

I said this already but humans are already cranking out robotic creative work to serve capitalism. cf Corporate Memphis. If this is what gets automated, who would even notice? Besides the out of work designers obv. But again, most of humanity used to work on farms. And even from a capitalist point of view, Western economies have been stuck in a low productivity growth plateau for a long time. We need a jump start if we have a hope in hell of paying off pension obligations. I for one welcome etc etc

There's a lot of work that needs doing, whether it's care work, community works, so much that isn't done here because of austerity policies.

Nevermind the work that is underpaid. Huge amounts of wage theft.

Plenty out there, but the political and business class don't want to pay, or they deny it needs doing, or they don't see value in it. Or if there is exploitation comes in quickly enough, instead of hiring.

pension obligations? we don’t need an AI for that, nearly all jobs in the US dumped pensions and you get a 401k with bad administration rates

I suppose there will soon be organizations who's sole focus is to verify something was not generated by ai.

Also the age of the "non-ai " content internet is over so ai is going to be trianed on other ai recursively to who knnows what end.

only the wayback machine at internet archive will be "pure"

Well pension obligations are still pretty big for unionized industries but yes point taken. More broadly though productivity growth is how we get rising living standards. AI seems like a very good bet to accomplishing this.

Re: care work, social work etc - AIs will never do this, can't do it, and yes that work needs to start being paid commensurate with its social value.

Robocare for the elderly has been a goal as long as self-driving cars and frankly I think we'll make better adoption with the former than the latter. People really loved their AIBOs! Also, given how quickly people took to even the most limited ELIZA routines decades ago, there's an argument to be made that a lot of this kind of care work would be preferable if done by machine than another potentially judgy human (or one you could be ashamed pooping in front of or confessing your worst impulses).

there are also examples of it going haywire and giving very inappropriate messages to patients struggling with psychological issues.

https://www.npr.org/sections/health-shots/2023/06/08/1180838096/an-eating-disorders-chatbot-offered-dieting-advice-raising-fears-about-ai-in-hea

i dont think ai will ever do these complicated human jobs, technology never maps one to one with human behaviors, itll do little tasks where its very clear what the paramaters are, just like all other computer programs, ai is just computer programs, with a compelling misleading name

i hope so

that is my impression too, but i am also worried about it being forced into places it doesn't belong, causing havoc and destroying jobs

im sure thatll happen, and is in fact already happening see sport illustrated story, but thats not ais fault that just business

But does that mean there won't be another magazine that will hire human staff and be a good read and sustainable? AI doesn't stop that either.

i blame drew ortiz

yeah its not ai that stops that its private equity et al, if you want to talk at something truly disastrous talk about the total financialization of our whole sitch

its the money guys who are really enthused about ai, its not hard to see why particularly considering they view whatever content product item they churn out as being basically interchangeable, they dont know whats good, they dont even acknowledge that theres such thing as good, the saving grace is that theyre wrong no one wants this computer slop, a computer will never write even the shittiest x man movie

likewise just like crypto a lot of the ai hype is just driven by big investors needing somewhere to put their money and then hopefully get out before the last song plays

in this world where journalists breathlessly report on valuations when you invest free advertising comes with it, once unprofitable companies started going public vcs realized you can make a lot of money on hype alone

that is my impression too, but i am also worried about it being forced into places it doesn't belong, causing havoc and destroying jobs

― treeship., Wednesday, November 29, 2023 10:00 AM (eighteen minutes ago) bookmarkflaglink

this is just the market. whoops the Q3 profits weren't as great as we hoped, better lay off 1/4 of our workers before the next quarter starts so we can show a decrease in costs. AI is just another lever to pull

lag∞n got most of that while I was typing, once again

:)

ai is just computer programs, with a compelling misleading name

― lag∞n, Wednesday, November 29, 2023 7:59 AM

Yeah seriously

there's a cynical optimism to be had that all this AI funding gives rise to an "actually profit-maximizing regime" that eliminates the parasite rentier/investor class.

also, I'm a fan of the AI-gone-wrong computer slop -- unleash it on the next marvel phase, I say!

read a thing that said that openai is full of people who really believe theyre on their way to making actual computer intelligence, funny how these extremely clever computer people believe this very silly thing, calling it religious or a cult seems accurate

There's frankly more overlap between the two groups than anyone should find comfortable...

need to pull some sort of mechanical turk scheme on them where the sentient computer tells them to transfer cash to my account

In the sense that people thought SBF was a sentient computer, that would totally work!

To be fair to AI hypers, actual human intelligence is way overrated by humans (we get outsmarted by bears all the time!), so the gulf between the two is probably not as big as people want it to be.

Bears?

the thing is computer intelligence isnt intelligence

From a practical POV, most of human intelligence isn't intelligence either though... (you don't have to fully buy into the stochastic parrot argument to concede that)

re: bears
https://static.wikia.nocookie.net/yogibear/images/8/88/Yogi_Bear_-_Character_Profile_Image.png/revision/latest?cb=20220627024042

Fair enough, he was smarter than average...

human intelligence is intelligence i mean thats where the idea came from, the question is does what the machines are doing resemble that, and the answer is no

If you flip the question around to does what we do resemble machines, unfortunately the answer is yes (though you could say that's the end result of capitalism)

i dont think humans resemble machines, were really bad machines if so, but were good humans

It's true AI is just computer programs but tbf it feels like computers might finally be realising a lot of their potential with LLMs. Like how many times have we all in our jobs or day to day lives had to do some kind of numbing, tedious task and thought "wait isn't this what computers are for, why isn't my computer helping me do this?" It's like we've been stuck in Zork world for decades, where you have to use exactly the right syntax. They're incapable of even the simplest inferences. There have been marginal improvements around the edges but not since the GUI has there been a kind of quantum leap in usability. I do think it's exciting. It's not "intelligence" but it's just like computers finally getting good

im gonna wait til theyre actual good before i get too excited

When did an LLM last help you? I'm curious xp

llms for general purpose talking to your computer stuff are worse than useless currently, theyre actively harmful, pure enshitification