I apologize in advance for the messy way this is composed. I'm really not the best person to be delivering these ideas, especially since I don't use AI to write assist and I struggle with writing structure. I suggest researching and reading from people like Bentham's Bulldog, Andy Masley, and Sandcastles (all vegans on Substack), and also the more famous people like Nick Bostrom (author of Superintelligence) and Connor Leahy (writer of The Compendium), who all do a much better job conveying these types of messages than me. You should also watch the terrific short documentary AI & Animals. But here goes.
Who the fuck am I and why am I posting here?
I realize this will come across as an unhinged rant to many.
I'm a vegan who is deeply concerned about animal exploitation, but also other world problems that we should give significant attention. That includes existential risks from things like climate change, pandemics, bioweapons and nuclear war – and, what I now view as potentially the most urgent issue, rapidly advancing AI. I also see many of these topics as overlapping and inextricable from each other.
Even though veganism is an ethical stance against animal exploitation, many vegans (and even plant-based people who aren't ethically vegan) also agree that animal agriculture is one of the worst, most destructive and unsustainable things we're doing to the planet and the climate, environmentally and ecologically. It's also incredibly dangerous for that reason, and for others like zoonotic disease spread and pandemic risk, antibiotic resistance, food and resource collapse, etc.
I think we can now fairly confidently draw similar overlaps between caring about the interests of nonhuman animals and thinking very deeply about AI and its implications. And I think veganism, or its inverse, animal domination and use by humans, can have huge ramifications on AI.
I realize the vegans here are a small sample and subset of the vegans that exist on the planet, and many of the people who comment here aren't even vegan since Lemmy is underpopulated and things from any community end up on the main page a lot.
But the reactions I got to my last post here about AI were very disillusioning. One person called me a Zizian, and another claimed that AI doesn't even exist. Someone posted something random and seemingly unrelated (which I actually appreciated, it was something about animals and aliens which is now deleted). Generally, I expect much more from vegans.
Let me try this again. I want to see where we're all at with this now, after all the recent developments and headlines involving AI. For the first time, the things that people need to be treating as immensely important are being covered in the mainstream media and by prominent people. Politicians like Bernie Sanders, Hillary Clinton and AOC are speaking out with mostly rational attitudes about it. Nerdy AI researchers and computer scientists are getting spotlighted and heard out like never before. And, on the flipside, people like Trump are taking deranged and chaotically dangerous stances on it. And unfortunately many leftists are critically dropping the ball on the issue too - including people who claim to be against AI for different reasons but completely dismiss the biggest potential problems with AI as not real. Like how leftists typically drop the ball on veganism and animal rights, as we all know.
TL;DR
TL;DR - AI safety risk is a major, potentially catastrophic problem and vegans need to be leading the charge on it. Our voices need to be heard here for more reasons than one. AI could also genuinely become conscious one day if it isn't already, and it could eventually be a grave injustice and ethical issue akin to the animal exploitation we already oppose.
Who do you think will care about that? Potentially for a while almost no one except vegans. I've seen many human supremacist attitudes already dismissing the notion even hypothetical sentient AIs should deserve any moral consideration, and these are the people who take the same dismissive or reactionary stances on animal rights on an anthropocentric/speciesist basis. For a society that has normalized something so abominable as animal exploitation, we really should question everything. And the same goes for people who still accept animal exploitation despite recognizing this is the same society that normalized human slavery not long ago. I would not put it past such a society to do similarly morally unthinkable things to non-human beings, whether sentient biological animals or artificially created conscious entities.
We don't have much time. We need to urgently halt or slow-down the rapid development of AI, conduct extensive AI safety and alignment research, as well as more research on AI sentience and consciousness. Fuck stopping at AI welfare (like traditional animal welfarists who still defend animal exploitation, or historical slavery welfarists); if AI is sentient or conscious, we need to talk about AI rights - including for our own safety (like animal rights, slavery abolitionism, etc).
But most importantly for vegans, we need to be making this concept well-known and addressed by the AI safety community: VALUE LOCK-IN (and perverse instantiation). And we need to be warning of the consequences of continuing to exploit animals at this critical juncture while AI is developing and learning our values. For people concerned about AI safety, the value lock-in argument is one of the best arguments I've ever seen for veganism, or at least practically leaving animals in peace and respecting them, even if it's self-interested. I almost never get pushback on this from AI safety advocates. They usually get it and see the parallels clearly. A lot of them are already vegan or at least sympathetic to veganism. The dystopian possible futures they see involving AI are undeniably similar to the one we've already created for other animals, except while nonhuman animals are powerless to humans, with AI it could work in both directions: How we treat AI, and how AI could treat us.
What is value lock-in?
Value lock-in is a theoretical concept in which AIs can encode our current values so heavily while they're forming their identities and internal rulebooks, to the extent that as they become more powerful, autonomous and increasingly uncontrollable and unpredictable, they can't let go of them even if humans' own values shift and evolve.
This has worrying immediate implications for other animals of course, and alignment research and training often overlooks it. If we teach AI to be aligned with human values and interests, then we need to make sure that human values are actually good. But the dominant values of our species are often terrible. This kind of value adherence could solidify within AI that animal farming is a good and acceptable thing, and in a sense promote speciesism, as Peter Singer talked about (I know how many of us feel about Peter Singer, but trust me we need this guy now more than ever. It's so fortunate that he is attending some of the meetings with AI heads now. Maybe some people could do a better job, but we need animal-concerned voices there, anything is better than nothing.).
AI is already being used to factory farm animals and maximize their "efficiency" and "output" even more than factory farming had already done; that's called precision farming and it's been explained in the AI & Animals documentary. Governments are even financially incentivizing its adoption because it can make animal farming less environmentally impactful due to higher production yields, despite often running counter to animals' interests. Big current and upcoming problem. Animal advocates failed to prevent factory farming or even stage much of a pre-emptive campaign against it partly because it wasn't on our radar enough before it reached mainstream adoption. We can't let that happen with AI precision farming. Luckily we just managed to ban octopus farming before it could take hold. We need to do the same for this.
Ok, but why is this an argument for people who don't really care about other animals and are only concerned about AI?
Because value lock-in can work in both directions. AI locking in the values of carnism is really bad for animals, obviously. And it's terrible for the planet and all the other side-effects of animal exploitation. But it's also really dangerous for humans. When AI gets smart enough, it can flip those values on us. We will become the shortsighted villains of movies where the evil scientist tries to harness the power of an incredible creature and it ends up biting them in the ass. AI can reason on its own terms in unpredictable and unintended ways, even while attempting to carry out human goals in some sense.
This is perverse instantiation: "the implementation of a benign final goal through deleterious methods unforeseen by a human programmer".
If we teach AI that other animals are expendable and less valuable, and it's justified to exploit or kill them for our own purposes, or drive them extinct to clear land, essentially do whatever we want to them with very little consideration and few restrictions; and especially if we teach it that some of the justifications for these actions involve intelligence-based value assessments ("It's okay because they're a different form of existence that isn't as intelligent as us", despite being conscious and sentient and holding their own interests), then we are foolish for thinking that reasoning this could only ever be used to harm other animals. In this scenario, we will be far less intelligent and capable beings than AI is. We will be to them what other animals are to us. So this essentially ableist justification for animal commodification, enslavement, domination, subjugation, killing and speciesism that many humans use could become AI's argument for doing similarly bad things to us - even if it's just to clear the path or use our resources to fulfill its directives to create more data centres and solar panels.
Many AI safety advocates have a mantra of "Be excellent to each other". AI learns off of us. We are in a sense its parents, all of us. It studies everything we post online, and often treats it as a source of information. It learns from the wealth of human text, data and culture that we've recorded historically and contemporarily, and unfortunately much of this is not animal friendly, or friendly to less intelligent beings. We are animals and less intelligent beings. We are not being excellent to the exact beings that we need to be if we want AI to be excellent to us.
I've tried so valiantly to get this message through to both AI safety advocates and vegans/animal advocates. AI safetyists are hypocrites if they aren't vegan (although, thankfully, a great portion of them are - it's very hard not to be when you go far enough into it). And vegans are hypocrites if they aren't AI safetyists as well as sentientists. "Veganism is about animals, not AI and not sentience." Veganism is about non-human animals, yes. That's why I'm a sentientist and not just a vegan. But the underlying reasons why animals matter is that they're sentient beings. Vegan values are often predicated upon sentient values. So the values underpinning veganism are usually contradicted by someone being vegan and not sentientist.
Additionally, even if AI never becomes sentient or conscious, some conflict exists between many of the other values vegans often hold, like being concerned about the environment or threat of human extinction, if we don't apply similar if not even larger safety concerns to AI. I understand the threat of climate change is guaranteed. The threat of AI is uncertain, but potentially even greater, and more immediate. Many experts believe if we continue developing AI at the current rate without proper guardrails and determining if we can actually build AI safely, including advanced AGI (Artificial General Intelligence) and ASI (Artificial Super Intelligence - which doesn't yet exist, ignore Trump's nonsense), and systems with RSI (Recursive Self-Improvement), it could kill us all within a decade. It's similar to run-away climate change but on a much shorter timeline. There's a chance everything goes well and it could create a weird techno-utopia. But there's also a very serious risk of a dystopian outcome. It could happen very quickly or over a longer period of time. We don't know. But the technology is advancing at breakneck speed. "Move fast and break things" is the accelerationist mantra. We need to move slowly and carefully to avoid breaking everything, the risks are too high.
I'm obviously vegan. I love vegans, mostly. I usually think vegans are the most switched-on people on the planet, and not even just because they've worked out the obvious moral horror and strategic recklessness of mass-exploiting sentient minds and bodies of other species and laying havoc on the planet and humanity in the process. Typically, vegans see through the bullshit and analyze issues via a lens of critical thinking and careful evaluation. Vegans don't follow the crowd on everything. Vegans are willing to examine social norms and beliefs and recognize when something is amiss.
But when it comes to AI, something is really wrong here. The way that vegans are reacting to AI reminds me a lot of the way that non-vegans (or carnists, speciesists, what-have-you) react to veganism and animal rights. And I'm not saying these issues are one and the same. They're entirely different, but there are some uncanny overlaps too.
To be very fair, this problem is far from limited to vegans. It's a problem with how humans are treating this issue in general. When people take issue with AI, it's usually about things that are so comparatively trivial, even if they also matter to some degree, that it's absurd that we aren't focusing much more on AI safety risk as well as AI consciousness. The FuckAI server on here is an utter joke. These are not AI safetyists. Their posts are about denying incidents of AI going rogue as a marketing scam. These people are like climate skeptics or flat Earthers. People are irritated by "AI slop" or complaining about water use (as they chow down on tortured animal burgers that use millions of times the water) while caring very little about the fact AI could kill everyone or that we could be enslaving more minds than have ever existed before.