I have been giving some thought as to why Western societies (I cannot speak for others) who put so much store in individuality and self expression and identity are so willing to let an #LLM in the form of an #AI chat bot speak for them.
-
@carbontwelve I think the emotional aspect of this is really quite important to consider. I like the cake mix analogy. Because, baked into it (forgive the pun) are two things: profit and the hegemonic definition of what a cake is.
@Remittancegirl I think if I were doing a PhD in Sociology today (I got so far as a Masters before running out of money) this might make for an interesting subject.
-
I am aware. Just like the idea of the perceptron has been around since the 1960s. The idea of AI has existed since the 1950s. Mind you, there was a famous movie in the 1960s whose whole point was AI that kills you, hardly a novel concept.
What we are discussing is a very large change of architectural capability.
@vruz @Remittancegirl @passenger LLMs aren’t the whole of machine learning. Going back to an earlier example you threw out, if ChatGPT, Claude, and the rest went away tomorrow, we’d still have exactly the same Netflix we do today, because the Netflix recommendation engine isn’t an LLM.
Netflix isn’t built on the large scale theft of people’s work, while all commercial LLMs are. Netflix isn’t being pushed as a tool to replace human creativity, while LLMs are. And while companies like OpenAI would love to group their product with other, less problematic, more useful applications of ML in order to blunt criticism, their product isn’t central to most of the applications you named.
-
That's not what i said. I'm disabled too. If i can find it again, I'll share a great essay by a blind author about the problems of LLMs.
Accessible tools for these sorts of things do exist but are being drowned by LLMs that have major flaws and dangerous impact to the environment.
And we know they are causing harm to many as studies are showing sharp drops in literacy, increase in LLM fueled delusions, and sometimes LLM output is not accurate or outright wrong.
-
@mfeir Here's what haunts me about this: there are two issues here.
Some people do have a struggle expressing their ideas clearly because they lack eloquence. And our society used to, at least, reward eloquence over substance often.
But often people's ideas aren't very clear. And they have not been taught how to hone them, or pin them down. In that case, the machine literally puts words in their mouth and they settle for that.
@Remittancegirl @mfeir I think that our society may be changing on the eloquence vs substance thing, thanks to AI slop.
Now, more and more, flawed prose is seen as a sign of quality, whereas LLM generated text is seen as a sign of not caring about quality.
It's better to express yourself by posting your (human written) prompt, than the more lengthy but less contentful LLM output. The prompt may have more obvious flaws, but it will convey your actual ideas, and more succinctly.
-
That's not what i said. I'm disabled too. If i can find it again, I'll share a great essay by a blind author about the problems of LLMs.
Accessible tools for these sorts of things do exist but are being drowned by LLMs that have major flaws and dangerous impact to the environment.
And we know they are causing harm to many as studies are showing sharp drops in literacy, increase in LLM fueled delusions, and sometimes LLM output is not accurate or outright wrong.
My point is that the answer is not LLMs, but making space and building up tools and techniques that are designed for that community that actually increase literacy, are good for environment, etc. These are possible to make. Some tools already exist but need better word of mouth and/or designs refined, but funding is scarce for non-LLM tools.
-
My hypothesis - and I don't have training in psychology or sociology so I don't know whether this is a trivial hypothesis or is nonsense or what - is that LLMs exploit an existing loophole in human psychology, which is that people confuse doing a thing and giving an order to do the thing.
"King Offa built a great earthworks to defend his kingdom." Really. Did he wield the spade himself?
The chatbot is built so that its interactions feel like being a manager and giving orders, which then similarly makes people feel that they personally did the thing.
@passenger@kolektiva.social @Remittancegirl@mstdn.social
“The chatbot is built so that its interactions feel like being a manager and giving orders, which then similarly makes people feel that they personally did the thing.”
And everyone wants to be a lazy, stupid manager that takes credit for the work of others. That’s the “good” capitalist mindset. Capitalism is the exploitation of workers by the management class. And toadies like to see themselves as overlords. -
I think one reason is that most chat bots are designed to encourage users to believe that they are 'working together'. This is utterly false, but it credits the user with work they aren't doing, which feels good.
Another could be that, despite all the outlets for personal expression, it is very easy to feel like one's voice isn't heard and doesn't matter. In that case, what does it matter if one's voice is being usurped.
Have you got any theories?
@Remittancegirl My theory is that it takes effort to think and write, whereas telling an LLM to think and write for you takes a lot less effort.
Sadly, most writing tasks are pretty useless. Write a summary that no one will read. Write what you did this week and what your plan is for next week, that no one will read.
When no one's going to read it anyway, at some point all you'll care about is making it long enough.
-
@vruz @Remittancegirl @passenger LLMs aren’t the whole of machine learning. Going back to an earlier example you threw out, if ChatGPT, Claude, and the rest went away tomorrow, we’d still have exactly the same Netflix we do today, because the Netflix recommendation engine isn’t an LLM.
Netflix isn’t built on the large scale theft of people’s work, while all commercial LLMs are. Netflix isn’t being pushed as a tool to replace human creativity, while LLMs are. And while companies like OpenAI would love to group their product with other, less problematic, more useful applications of ML in order to blunt criticism, their product isn’t central to most of the applications you named.
"You threw out" tells me that you're not very receptive to any exchange of ideas. Which is precisely the irreflexive and blunt kind of discussion I don't believe has any merit to engage with.
Also "theft" and other morals encoded. What a mess.
-
Yeah, absolutely. It's the willingness to see other humans as nothing more than tools, and therefore to equate King Offa using the peasants, versus the peasants using spades.
@passenger @uastronomer @Remittancegirl i think this is why society tolerates billionaires to the extent it does
-
My hypothesis - and I don't have training in psychology or sociology so I don't know whether this is a trivial hypothesis or is nonsense or what - is that LLMs exploit an existing loophole in human psychology, which is that people confuse doing a thing and giving an order to do the thing.
"King Offa built a great earthworks to defend his kingdom." Really. Did he wield the spade himself?
The chatbot is built so that its interactions feel like being a manager and giving orders, which then similarly makes people feel that they personally did the thing.
Probably true. But with independent AI agents, people won't even be giving commands.
-
@Remittancegirl For some people, expressing their ideas clearly is a real struggle. Learning to write well is a very tall order for them. I feel very comfortable expressing myself and would only use an LLM for the odd alternative suggestion. Meanwhile, if I have to heave some heavy furniture anywhere, I'll want a little help from my more muscular friends. Even better if there were a machine I could ask rather than interrupting their lives. That's the mindset at work here I think.
@mfeir @Remittancegirl I’ve definitely got sympathy for that, but at the same time, I think the substitution of machines for community is ultimately somewhat corrosive. I intentionally make a point of reaching out to people I know with particular skills at work when I’m tackling a problem that could benefit from their input, not just as a way to get my work done faster but as a way of acknowledging their ability and building a relationship. Going for Copilot might get the work done (or maybe not, given the chances of getting incorrect output), but I feel like I’d lose something in terms of relationships. Especially in the current moment, that feels like a significant problem.
-
If we want people to use their own voices, we need to understand what they're getting out of allowing #AI to speak for them (which in essence is always AI speaking pretending to have one's best interests at heart which is a laughable falsehood)
It also occurs to me that correcting someone when they do speak for themselves - inexpertly or haltingly or confusingly - is a fairly large wound to their self-esteem.
That fear of a lack of eloquence, I think, also drives people to AI.
@Remittancegirl I think a lot of it, as @isaackuo put it, is that people use LLMs for stuff they don't care about. "I don't care about making a menu to display, I just want to show the menu." "I don't care about this email." "I don't care about this interaction. I just want to go out for a drink."
I myself battle constantly with my urge to do things fast, quick, to cut corners, to Learn Japanese Fast. I've learned that you really, really can't. But the urge is there.
-
@Remittancegirl I think a lot of it, as @isaackuo put it, is that people use LLMs for stuff they don't care about. "I don't care about making a menu to display, I just want to show the menu." "I don't care about this email." "I don't care about this interaction. I just want to go out for a drink."
I myself battle constantly with my urge to do things fast, quick, to cut corners, to Learn Japanese Fast. I've learned that you really, really can't. But the urge is there.
@Remittancegirl Duolingo lets you think you spent a little money and you learned a lot of language, where in reality you probably learned how to play Duolingo. YouTube Shorts make you smile, and smile, and maybe laugh, and smile, and more and more they're empty robotic voices telling you DID YOU KNOW. And LLMs give you instant satisfaction when you wanted something. If you don't have it clear in your mind that the road matters, you just want to get to the destination.
-
I have been giving some thought as to why Western societies (I cannot speak for others) who put so much store in individuality and self expression and identity are so willing to let an #LLM in the form of an #AI chat bot speak for them. Because essentially, when you ask AI to write something for you, that is what you're doing.
Why are we so eager to have our own voices usurped? 1/
@Remittancegirl I think our skill of using writing as a form of self expression has been eroded by technology. Word processors sped up writing but added spell checkers and then grammar editing, then suggesting entire phrases all before we even got AI. “The computer does it better so just let it.”
Plus, email and texting has condensed the expression of our language and humanness to memes and reduced our emotions to emojis.
The lack of language and words to write means the lack of words and language to speak and express ourselves through the spoken word as well.
-
@Remittancegirl Duolingo lets you think you spent a little money and you learned a lot of language, where in reality you probably learned how to play Duolingo. YouTube Shorts make you smile, and smile, and maybe laugh, and smile, and more and more they're empty robotic voices telling you DID YOU KNOW. And LLMs give you instant satisfaction when you wanted something. If you don't have it clear in your mind that the road matters, you just want to get to the destination.
@Remittancegirl Coders have never cared about tests, or documentation, or scaffolding, or most apps we're asked to write. Managers have seldom cared about code. So when a tool says "I can document and test for you" and "I can build the app in minutes" some people jump to it. Nevermind that the tests are mockups, the docs are wrong, and the app is at the first 80% done and unreadable. You done it.
-
@Remittancegirl Coders have never cared about tests, or documentation, or scaffolding, or most apps we're asked to write. Managers have seldom cared about code. So when a tool says "I can document and test for you" and "I can build the app in minutes" some people jump to it. Nevermind that the tests are mockups, the docs are wrong, and the app is at the first 80% done and unreadable. You done it.
@Remittancegirl I should note that another problem is that Things cost Money and that we usually have less and less money. I used an app to learn Japanese because actual lessons with a real teacher would have cost a lot more and I didn't have that. Would that have been better? 3000% yes. But I couldn't do it.
Is it actually real savings, in the case of LLMs? I don't know. It sounds an awful lot like "they gift you the first dose, then..."
-
@vruz The problem is that I don't buy that they are intellectual aides. I think they masquerade as intellectual aids.
The artificial limb has no agenda. It isn't owned by oligarchs who want you to walk in a certain direction.
@Remittancegirl @vruz @passenger I suspect there is some “bias” from people (like us? I dunno about you, but def applies to me) who are good with words, and enjoy expressing themselves. And/or for whom self-expression is important.
If this is not part of your identity, you really don’t care if they’re not your words. If you feel like you’re bad at words, you want the reassurance of having something whose “job” is words do it for you. And you hated writing that email, anyway. (Plus, if you’re bad at words you won’t notice that what the LLM wrote is bad!)
-
@mfeir Here's what haunts me about this: there are two issues here.
Some people do have a struggle expressing their ideas clearly because they lack eloquence. And our society used to, at least, reward eloquence over substance often.
But often people's ideas aren't very clear. And they have not been taught how to hone them, or pin them down. In that case, the machine literally puts words in their mouth and they settle for that.
@Remittancegirl @mfeir this whole thread is fascinating! I very much appreciate the breadth of perspective, the humility and the honest curiosity!
I'd like to add that sometimes perhaps we prefer to not say precisely what we mean even if we could.
Eg "come up with a reason to not visit my family over the holidays that isn't 'you're unpleasant to be around and your food sucks'"
-
@Remittancegirl @vruz @passenger I suspect there is some “bias” from people (like us? I dunno about you, but def applies to me) who are good with words, and enjoy expressing themselves. And/or for whom self-expression is important.
If this is not part of your identity, you really don’t care if they’re not your words. If you feel like you’re bad at words, you want the reassurance of having something whose “job” is words do it for you. And you hated writing that email, anyway. (Plus, if you’re bad at words you won’t notice that what the LLM wrote is bad!)
@clarissawam @Remittancegirl @passenger
As soon as a broad category of activities is deemed immoral, reason went out of the window. Impossible to start any reasoned discussions as soon as people begin to shout "theft" irreflexively.
Of course I agree that billionaires are comfortably riding on the blood, sweat, and tears of humankind.
But that's capitalism, people feel very strongly about the symptoms and dedicate too little time, or spare any thoughts to combat the causes of injustice.
-
@Remittancegirl I think it's because a lot of people are more socially awkward than they're comfortable admitting, and writing exposes one to the reader. Your mistakes, your typos, your misuse of vocabulary or grammar, all things which can create anxiety.
Thus, having an "editor" which polishes your words is seductive, and once seduced, letting the editor become the writer is an easy slide.
I think everyone wants to feel/sound erudite, and to those unfamiliar with the patterns of ai writing, it can mimic it fairly well.
@MissConstrue @Remittancegirl And the ironic thing is that what comes out desperately needs editing. That's what I don't understand—how can people use it *as is*? I reckon a human first draft¹ can typically be reduced by about a third without losing any content, and the LLM stuff is worse because a lot of the actual content is irrelevant too. You'd think people would at least remove the irrelevant stuff and change some of the words to ones they themselves use.
_____
¹ if I'm the human, anyway
