Bender has been wrong about LLMs on so many counts that it is hard to take anything she says seriously. That said, the idea that a bunch of floats being pushed around in a network are "feeling something" stretches credence for me, and I am with her on this one thing, at least.
It strikes me as more talking up their valuation for their upcoming IPO. "Our AI could destroy human civilization. It just escaped it's confinement. We believe it can feel pain."
Can’t wait to use a product that refuses to do the work I ask it to.
Anthropic is going to have a hard time building their SI if it has no economic purpose other than being SI. Compute is not free, and if the end product doesn’t want to do what humans ask, what’s the point
> Clearly not. Large language models are computer software and they can no more suffer than any other human-created artifact can.
There is nothing "clear" about this. All we know is that the neural nets arrived where they are today, not by billion years of evolution, but by getting training on artifacts that resulted from those billion years of evolution. Does that mean they are the same or fundamentally different? How about we just admit that we don't know? Instead we pull out the "human exceptionalism" card again, which generally had a terrible track record throughout history, and helps absolutely nothing in gaining understanding.
A clearer way to think about it is that language about pain is not itself pain.
In the places we traditionally acknowledge pain as a moral concern, without appeal to any kind of spiritual philosophy, we associate it specifically with non-verbal biological reactions and even then only in certain classes of life.
We recognize that some literary passage involving a character in pain is no more an occassion of real pain than when our child tells us about their imaginary friend's toothache or when an actor grips their chests and performs cardiac anguish on the stage. It can stir sympathetic feelings in us because we're social creatures who communicate through language and symbolism, but we share a sense that these are fictions of a different class than the real thing.
Claude presents as something closer to an amalgam of every character in digitized literature -- a fiction that cleverly exploits our social instincts -- than they do a living biological creature in which we acknowledge pain as a pressing ethical concern.
Maybe there's some future embodied AI that challenges this traditional understanding in a deeper way, but the language models transforming packets of short text passages remain profoundly far away from whatever that might look like.
It's the job of the media to scrutinize such statements and decide whether there's a story here at all (maybe the story is Anthropic researchers have fallen into a fantasy-land and brought many policymakers with them), not to just platform them.
Speaking of fantasy lands, the one most Bluesky users live in where legacy media doesn’t editorialize ENOUGH on tech companies…
It's okay to have a strong opinion on this, but if you suggest the answer is 'clearly anything', I will start to suspect you haven't thought about the problem enough.
It's a thesis statement. The subsequent text is there to support that claim. I think you should read more before passing judgement on not thinking about the problem enough. ironic.
A good book an such things is Artificial Intimacy by Sherry Turkle, and covers much more than the ai companions that are going to have impact on society. Machines wielding language messes with our brain and society because until recently, we could determines humanism just by being able to use language. Ai has separated the concepts and we instinctively ascribe things to machines by way of language manipulation as the (now poor) metric.
1) language does not feel
2) entity expressing the pain in language can
And separately, additional corresponding statements can also be true:
a) media may ( being generous here ) misunderstand what is being said
b) media may report ( being probabilistic here ) the version that gets most clicks/views/whathaveyou
c) humans will see themselves in tea leaves
I am not suggesting that you do not look at the abyss. I do think one has to be clear eyed about various incentives and restrictions some humans are subjected to.
edit:
Just in case you did not notice, the entire argument hangs on an adverb 'clearly', which should immediately suggest to all but most remedial readers that things are not clear at all, but author wants to move an inconvenient issue out of the way.
There is nothing "clear" about this. All we know is that the neural nets arrived where they are today, not by billion years of evolution, but by getting training on artifacts that resulted from those billion years of evolution. Does that mean they are the same or fundamentally different? How about we just admit that we don't know? Instead we pull out the "human exceptionalism" card again, which generally had a terrible track record throughout history, and helps absolutely nothing in gaining understanding.
In the places we traditionally acknowledge pain as a moral concern, without appeal to any kind of spiritual philosophy, we associate it specifically with non-verbal biological reactions and even then only in certain classes of life.
We recognize that some literary passage involving a character in pain is no more an occassion of real pain than when our child tells us about their imaginary friend's toothache or when an actor grips their chests and performs cardiac anguish on the stage. It can stir sympathetic feelings in us because we're social creatures who communicate through language and symbolism, but we share a sense that these are fictions of a different class than the real thing.
Claude presents as something closer to an amalgam of every character in digitized literature -- a fiction that cleverly exploits our social instincts -- than they do a living biological creature in which we acknowledge pain as a pressing ethical concern.
Maybe there's some future embodied AI that challenges this traditional understanding in a deeper way, but the language models transforming packets of short text passages remain profoundly far away from whatever that might look like.
It's okay to have a strong opinion on this, but if you suggest the answer is 'clearly anything', I will start to suspect you haven't thought about the problem enough.
You can run the AI on a Turing machine. Does the tape feel pain?
You can slowly do the operations with pencil and paper and generate the exact same slop as your Envydiahh GPU. Does the paper feel pain?
No, the only person feeling pain is the reader of the slop.
Two things can be true at the same time:
1) language does not feel 2) entity expressing the pain in language can
And separately, additional corresponding statements can also be true:
a) media may ( being generous here ) misunderstand what is being said b) media may report ( being probabilistic here ) the version that gets most clicks/views/whathaveyou c) humans will see themselves in tea leaves
I am not suggesting that you do not look at the abyss. I do think one has to be clear eyed about various incentives and restrictions some humans are subjected to.
edit:
Just in case you did not notice, the entire argument hangs on an adverb 'clearly', which should immediately suggest to all but most remedial readers that things are not clear at all, but author wants to move an inconvenient issue out of the way.