Anonymous #2 28w

i mean the term wasn’t cut from whole cloth, it first appeared in a star wars game in 2005. it’s actually interesting because it was used by the clones to mock battle droids, which really feels a bit like pot and kettle

3

AI isn’t bad as a result of design at a high level, it’s the technical reality of the thing. It’s a computationally intensive, fundamentally limited technology that’s being sold as a panacea by tech companies. LLMs essentially string words together based on previously observed patterns of words, and while they sometimes correctly encodes meaningful information, they fundamentally do not think or reason.

1

Comment Deleted

0

Anonymous #6 28w

yeah because my argument here is definitely that LLMs deserve to be treated in a humane manner. you’ve cracked the code.

0

Anonymous OP 28w

People can call that bullshit whatever idc

1

it’s all matrices under the hood. a reasoning loop fundamentally can’t be a feature of an LLM, you can approximate it, but the system works by abstracting every concept into a vector. you fundamentally cannot reason with vectors.

1

Comment Deleted

0

Anonymous OP 28w

It’s a perfectly justified response to want to belittle a machine that keeps being advertised and sold as a replacement for humans actually

1

sorta like how the hardest part of building a non-circular wheel is hiding the circle; the hardest part of building lifelike AI is hiding the vectors. they're just vectors no matter how well you hide them

1

Anonymous #6 28w

I literally agree that AI is a net negative for society and that the way it’s being sold is fundamentally misrepresentative of its capabilities, and that the speculation-fueled economic bubble is going to result in extremely damaging market readjustments on a global scale, not to mention the misallocation of key resources this has caused

6

Anonymous OP 28w

but creating a slur for robots doesn’t do a single thing to further your cause, and again, i think it’s kind of weird to dislike something and go “ah i shall craft an n word for this” like you can just dissent without replicating patterns of behavior that have been extremely harmful to society as a whole. do i think you’re hurting anyone by using the term clanker? fuck no. do i think it’s a childish response that’s ineffective? absolutely.

4

Comment Deleted

0

i said lifelike, not humanlike. you're wrongly assuming that i mean like human life - probably because the straw man is easier to address than the fundamental reality that it's vectors all the way up and down

0

Comment Deleted

0

Comment Deleted

0

this is why i can't stand debate bro culture because i correctly use a term like "straw man" to compel you to actually read what i said closely and not just continue shadow boxing and you get all hostile and act like i'm trying to debate you. honestly fuck you. have a day

0

Humans are more important than AI, not because AI isn’t human, but because humans can feel pain and experience suffering. Even if AI can do either of those things, which I am deeply skeptical of w/r/t LLMs, that pain/suffering can be alleviated by just unplugging the god damn data center that’s increasing electricity costs and keeping people awake all night with unregulated noise.

1

Comment Deleted

0

Anonymous OP 28w

It is this deep actually

11

Comment Deleted

0

Comment Deleted

0

I think the origin of AI really means that this isn’t a comparable conversation to discourse on whether other biological species can feel pain, which i assume is what you mean? LLMs are constructs that take in text input, turn it into vectors, do a bunch of matrix algebra, and spit out more vectors. Sure there’s a philosophical question about what consciousness is, but fundamentally a computer is an elaborate system of logic gates, and that does not constitute consciousness imo.

3

the subjective experience of these models and networks is certainly interesting, but if I’m being honest I think your anthropocentrism point kind of misses most people’s conversation here. The overwhelming majority of criticism of AI has to do with its real world impacts, inaccuracy, and the way it is being marketed and developed. When the average person says AI, they mean chatgpt, grok, nano-banana, and whatever else, and they dislike those tools because their development directly harms humans.

6

Anonymous OP 28w

I don’t mean to suggest that criticism of AI rooted in its ‘inferiority’ or lack of humanity doesn’t exist, but simply speaking, the people who can’t afford their power bill because a data center went up and they can’t sleep at night haven’t thought about the consciousness of AI in any meaningful way, they just want to stop being hurt.

6

Comment Deleted

0

i made an accurate assessment about the nature of generative AI: that the more "lifelike" it is, the more well hidden the fundamentally lifeless reality is. this is in direct response to your point about stumbling on the building blocks of life; my comment is a rejection of this notion, but you would have to be reading to understand rather than reading to respond to get that

1

Comment Deleted

0

man i should finally get around to playing detroit become human

6

i then correctly identified a straw man and criticized it, at which point you started calling my comments "debate bullshit" which is nonsense and no, rejecting a specific claim within the context of a broader discussion is not dismissing the entire conversation

1

Comment Deleted

0

Comment Deleted

0

you’re both just talking past each other and relying on bad faith interpretations of each others arguments. #5, when normal people say AI, they mean LLMs, which means that in their view, the consciousness question likely isn’t worth asking. #7, you have a point on ‘hiding the vectors’ but you refuse to engage with any of #5’s broader philosophical questions. You’re just talking about entirely different things and yelling at each other for not following the conversation.

3

If i remember correctly it’s kind of mid because the writer really insists on using AI as an allegory for racism, and it just doesn’t fit

5

Comment Deleted

0

Comment Deleted

0

I think the reward structures used to train LLMs make exploring them inherently difficult, because they’re very apt at detecting the slightest hint of confirmation bias, and then reinforcing it to generate positive user experiences. it’s difficult to assess whether or not an LLM has any kind of self determination without perturbing the system, and that’s before you consider the credibility that they gain with users by representing themselves as having subjective experience.

6

Anonymous OP 28w

i mean to illustrate it poorly, look no further than the piss filter that appears in ai generated images which is caused by users having a slight preference for more saturated images, and then the models self cannibalize on generated content and then a system quirk becomes a noticeable artifact

5

Comment Deleted

0

Comment Deleted

0

i’m getting more at LLMs being fundamentally designed/created to be responsive to user input. as far as I’m aware, the only way you get an LLM to disagree with a user is through higher level ‘user input’ in the system prompt or datasets they can query, because they fundamentally cannot grasp objective reality in a meaningful sense. Any investigation of an LLM’s self-determination necessitates querying it, and at that point you color its responses. I’ll admit I’m no expert with ML or LLMs tho.

4

Comment Deleted

0

I’m an acoustics masters student so I know even less than you on ml lmao

5

Comment Deleted

0

i mean i did physics and music for my undergrad so it was a pretty natural continuation, and some of the research i’ve been offered has actually concerned ML tools. there was one cool one about using ML to virtually ‘move microphones’ by using an ML model to estimate the transfer function from a target point to the known point where the mic is.

5

i apologize for my hostility. i've been really short on patience when it comes to online disagreements, and i think we have common ground in wanting "normal" conversation and not just rhetorical debate ping-pong. i was mainly talking about LLMs/GPTs, as these are the main thing people mean when talking about AI, though i think similarly about ANNs. i actually missed your first mention of SNNs (right before i commented) which definitely would've changed my approach - that's my bad

7

Anonymous #7 28w

i think this is an overall good discussion and i apologize for the short fuse i had today

6

Comment Deleted

0

i didn’t end up going for that one because i wasn’t sure I was passionate enough to a PhD, but basically my understanding is you feed a model paired data for measured pressure signals at two locations, with information about their spatial relationship. the application of interest was automotive, so you’d also probably feed in data on the geometry of the cabin.

2

Comment Deleted

0

it’s for active noise control because allegedly you can’t put microphones in people’s ears when they’re driving. in theory it could also be combined with computer vision to feed in realtime data about head positioning to further localize the ‘virtual microphone’ signal. that computer vision implementation used to be fantasy stuff but now that the hardware is already there for awareness/attention monitoring it’s more viable

1

Anonymous canesfan 28w

it’s very demonstrably not but go off i guess

1

Anonymous OP 27w

It’s already increasing productivity which will help keep inflation under control, and there are tons of novel medical applications that can save lives. Like I know there are more sophisticated criticisms but most of the ai hate seems to come from people just getting annoyed by ai slip on social media. https://www.bbc.com/future/article/20260309-ai-is-finding-treatments-for-incurable-diseases

1