Your figure for insects is off by multiple orders of magnitude. There are far more of them, making this more pressing.
The list of sentience candidates misses a huge relevant group that is still systematically neglected—spiders. Portia spiders are especially intriguing candidates.
The plans to disenhance animals specifically do not seek to suppress reflexive responses to pain, but conscious experience of it, so the intended outcome would actually be animals that show signs of pain they do not experience. (Which has its own severe risks, especially in light of the fact that we are not in fact sure that this result could be achieved, and farmers may well numb as a result to animal pain expressions, a la Descartes the screeching is that of a malfunctioning clock.)
A development I fear in using AI in factory farming to monitor welfare, based on how I have seen this actually researched and implemented, is over-reliance on AI to signal welfare problems, while having even less human oversight for all the animals, which is troubling in light of the fact that the systems trained to recognise suffering animals have very dubious training data. E.g. it is assumed that your average factory farm animal that doesn’t look egregiously sad to the current vet is probably doing fine, then volunteers are briefly trained to label data accordingly, have their own speciesist bias affect the results, and then this all gets a fancy AI label and suddenly seems objective when it really is not.
Insect farming is indeed turning into a huge thing, and they currently have no protective legislation, even while we are finding increasing indicators of a capacity for suffering. Insect farming is potentially very cheap and space, water, food turnover, climate etc. friendly, and so may well scale up significantly while there are zero welfare steps taken, and then be economically entrenched and hard to reverse, especially as the public tends not to be sympathetic to such alien creatures and many aspects of it are also quite “green”.
Animal communication is also not trivially solvable by just throwing AI at it.
And something relevant people keep forgetting; while decoding wild animal communication is very challenging, teaching non-human animals human language systems (verbal for parrots, sign for primates, or generally working with symbol tables or buttons) has proven surprisingly successful (albeit also ethically disastrous); we’ve gotten far more communication going that way. (They understand us better than we understand them.) A lot of the consequences and issues that follow have already played out that way.
Multiple corrections/additions.
Your figure for insects is off by multiple orders of magnitude. There are far more of them, making this more pressing.
The list of sentience candidates misses a huge relevant group that is still systematically neglected—spiders. Portia spiders are especially intriguing candidates.
The plans to disenhance animals specifically do not seek to suppress reflexive responses to pain, but conscious experience of it, so the intended outcome would actually be animals that show signs of pain they do not experience. (Which has its own severe risks, especially in light of the fact that we are not in fact sure that this result could be achieved, and farmers may well numb as a result to animal pain expressions, a la Descartes the screeching is that of a malfunctioning clock.)
A development I fear in using AI in factory farming to monitor welfare, based on how I have seen this actually researched and implemented, is over-reliance on AI to signal welfare problems, while having even less human oversight for all the animals, which is troubling in light of the fact that the systems trained to recognise suffering animals have very dubious training data. E.g. it is assumed that your average factory farm animal that doesn’t look egregiously sad to the current vet is probably doing fine, then volunteers are briefly trained to label data accordingly, have their own speciesist bias affect the results, and then this all gets a fancy AI label and suddenly seems objective when it really is not.
Insect farming is indeed turning into a huge thing, and they currently have no protective legislation, even while we are finding increasing indicators of a capacity for suffering. Insect farming is potentially very cheap and space, water, food turnover, climate etc. friendly, and so may well scale up significantly while there are zero welfare steps taken, and then be economically entrenched and hard to reverse, especially as the public tends not to be sympathetic to such alien creatures and many aspects of it are also quite “green”.
Animal communication is also not trivially solvable by just throwing AI at it.
And something relevant people keep forgetting; while decoding wild animal communication is very challenging, teaching non-human animals human language systems (verbal for parrots, sign for primates, or generally working with symbol tables or buttons) has proven surprisingly successful (albeit also ethically disastrous); we’ve gotten far more communication going that way. (They understand us better than we understand them.) A lot of the consequences and issues that follow have already played out that way.
- But overall, stimulating read, thank you!