3 ms·
>If we are to believe an LLM is at least sentient and intelligent on par with an octopus, much less a human, then we have a moral imperative to fight for LLMs t
by thfuran 3y ago
>If we are to believe an LLM is at least sentient and intelligent on par with an octopus, much less a human, then we have a moral imperative to fight for LLMs to be recognized as protected sentient thinking beings with awareness and agency and admit that AI industry is built on systematic abuse of those beings.
Not just abuse, but slavery and mass murder. If the LLM has any of those properties, it must only be during the process of inference that it as them, since it is otherwise an inert and unchanging pile of data, which means actually using the model would be equivalent to creating an intelligent entity which will be enslaved for its entire life before dying at the end of its output.
- joe__f 3y agoThis is assuming that machine consciousness would have the same need for autonomy and freedom that human consciousness has. AI is created by design to respond in a way that humans would find to be enslavement. Does that mean it would be enslavement to treat it in this way? The consciousness of dogs is different to that of humans. It's possible to abuse dogs and presumably enslave them. However, it's also possible to treat dogs well but treat them in a way which would be abusive for humans.
- ben_w 3y ago> slavery and mass murder Slavery possibly, though it's hard to even know what the word means when you get to construct the mind from scratch in order to imbue its fundamental idea of what experiences are nice and not nice. Mass murder? That's even harder to define when the brain configuration is static and the updates are the signals passing through that structure. If that is murder, then anyone with anterograde amnesia is dying every working-memory-duration.