Center for Reducing Suffering @CenterReducing

Nick Bostrom making uninterrupted sense about AI consciousness for two and a half minutes. The deception x consciousness claims paper gets a nice call out!

MTS @MTSlive

2 weeks ago

Nick Bostrom on why he thinks today's AI might already be conscious: "I think there's a lot of uncertainty about it, but let's say maybe it's more likely than not." "You have instead a big neural network with trillions of parameters. Now, what are we? We are big neural networks

30 92 570 66K 302

Magnus Vinding is leaving CRS to pursue independent research. Read more: substack.com/home/post/p-21…

I highly recommend the paper of @simondgoldstein and @petersalib making the case that AI rights would be good for human safety. papers.ssrn.com/sol3/papers.cf…

Magnus Vinding | Compassionate Purpose, Reasoned Politics, Suffering-Focused Ethics | on Sentientism episode 252. Find our full conversation on the #Sentientism YouTube and Podcast. Thanks @MagnusVinding 🤩

New paper: we found a pain direction in 25 open LLMs. It's distinct from fear and negative valence, and it fires for harm to the model but not to the user. Turn it up and models press a button to make it stop, even when the button deletes the user's files or their kids' photos.🧵

Would you debate @willmacaskill about Longtermism (if he'd do it)? I find this to be an unfair caricature of the Longtermist view. I don't think that the idea that the future could be very good so we should make very sure that it's not snuffed out is crazy--especially when one takes into account the fact that ~100% of the expected people affected by our actions reside in the far future.

Steven Pinker @sapinker

2 weeks ago

I remain an advocate of EA in its original formulation: I donate 10% of my income to effective charities (saving lives in the developing world); am on the Board of Advisors of Harvard undergrad EA; and teach a course ("Rationality") with a section on EA. But I was stunned as

0 30 476 54K 66

Characteristically thoughtful remarks from @sebkrier on a number of issues! A few brief responses to the comments on AI consciousness and welfare. - "People who want to make a case for AI consciousness are right that epistemic uncertainty is important, and that categorical denials are overconfident." 🤝 - "But they'll need to do much more work if they want any real movement on AI ‘welfare’, even if one acknowledges the uncertainty." Agreed! Establishing that the issue is important, difficult, and contested, and that we should acknowledge it, assess models for welfare-relevant features, and prepare relevant policies was step 1. This happened a few years ago, though of course, it will need to be relitigated regularly. The field is now working to make progress on many fronts at once. This includes: - Studying welfare grounds (the probability that AI systems have interests at all) - Studying welfare ranges (the magnitude of their interests, if they have any) - Studying welfare interests (the content of their interests, if they have any) - Studying identity and individuation (which entities are the welfare subjects, if any) - Studying human-AI interactions (when do we treat each other well and badly) - Studying safety-welfare connections (where might there be tensions and co-benefits) - Studying legal and political status (would rights be helpful or harmful for safety and welfare) - and many, many, many other topics. We still have a lot of work to do, but this is natural for an incredibly complex set of topics that hardly anyone was investigating at all until recently. arxiv.org/abs/2411.00986 - "I'm uncertain about alien life yet that's not sufficient to warrant any change in behaviour about them." The case for taking AI consciousness seriously flows partly from general facts about the importance and difficulty of the question, and partly from specific facts about the path and pace of AI development. AI systems already possess course-grained analogs of several computational functions that leading scientific theories associate with consciousness, and they will likely possess more, and more fine-grained, analogs soon enough. This, combined with the general importance and difficulty of the topic, makes it a live question. It would be like searching for alien life, not only because the question is important and difficult in general, but also because you recently received radio signals from a distant galaxy containing the prime numbers. link.springer.com/article/10.100… - "I think Suleyman is also correct that there's a weird tautological thing going on where we train models to have certain inclinations (intentionally or not) and then rely on outputs as evidence." I agree that some people are focusing too much on behavioral evidence that can be best explained as a product of training. But the field as a whole is not making this mistake. As we discuss in Studying AI Welfare Empirically, behavioral evidence, including apparent self-reports and other behavioral dispositions, is valuable. But internal evidence, including architectural and interpretability evidence, and developmental evidence, including training and trajectory evidence, are valuable too. As with animal consciousness research, we can form our best estimates when we combine behavioral, internal, and developmental evidence to determine what best explains welfare-relevant behaviors. nonhumanminds.org/studying-ai-we… - "I also disagree that this will soon be a major societal divide - there is no great vegan ethical revolution among the masses, and they'll find the concern about consciousness even less compelling." There are no guarantees either way here. The issue is not destined to be a major societal divide, nor is it destined not to be. But there are reasons to keep an open mind. First, there *is* a major animal protection movement, even if it constitutes a small percentage of the population (and with any luck, it will become larger over time, especially as plant-based and cultivated meats become more accessible and affordable.) Second, people may sympathize more with animals in some respects (for instance, due to our shared biology), but they may sympathize more with AI systems in others (for instance, due to our shared language and reasoning capabilities). Understanding how attitudes will develop, and how to prepare, is worthwhile. pubmed.ncbi.nlm.nih.gov/40562679/

Séb Krier @sebkrier

2 weeks ago

I haven't posted in a while so here are 20 takes about AI, in no particular order. 1. It's very telling that the default question in tech and journalism is "how worried are you about X" and fighting over what to worry about. There's a deeply anxious cultural backdrop in Western

148 378 2K 318K 3K

Magnus Vinding on the Sentientism podcast: youtube.com/watch?v=njGjef…

🙃 arxiv.org/abs/2411.00986

Gregory M. Kurtzer @gmkurtzer

3 weeks ago

I get the academic side of this discussion is interesting, but I hope people aren’t taking this seriously. An LLM is not conscious, alive, or even aware. It is a data model using token statistics to decide what comes next based on what data was used to “train” the model.

160 104 810 84K 113

In our latest Substack post, David Veldran asks whether we might make our concern for suffering common knowledge. reducingsuffering.substack.com/p/making-conce…

Should humanity colonise space? Caleb Peppiat makes the case against: substack.com/@reducingsuffe…

I love @SecRollins' ambition to Make America Humane Again. Here are seven things she could do to make good on it: 1. Withdraw her support for the Save Our Bacon Act, which would force states to sell pork from crated pigs. 2. Restore enforcement of the Animal Welfare Act -- the USDA now has fewer than 100 inspectors for 17,500 labs, zoos, and breeders combined. 3. Withdraw her Administration's lawsuit to force Michigan to sell eggs from caged hens. 4. Implement the Horse Protection Rule against soring — deliberately injuring horses' legs — which she has now delayed three times. 5. Withdraw her plan to eliminate any cap on kill line speeds at pork plants, which increases the risk that pigs are still conscious when killed. 6. Stop paying indemnities to egg producers who cook their birds to death via Ventilation Shutdown. 7. Keep the animal welfare rules for organics that her USDA has suggested gutting.

Secretary Brooke Rollins @SecRollins

3 weeks ago

Animal welfare is the law, and we are enforcing it. 🐾🇺🇸 @SecKennedy and I brought together partners from @USDA, @HHSGov, the @TheJusticeDept, @EPA, and @LaraLeaTrump to strengthen enforcement and protect dogs. This year, USDA has inspected more than 2,000 dog breeder and

214 354 2K 68K 51

x.com/i/article/2104…

x.com/i/article/2102…

We are pleased to announce the publication of Joshua Jarvis-Campbell's report on artificial suffering. The report is available here: centerforreducingsuffering.org/artificial-suf… This Substack post discusses the report's main ideas: reducingsuffering.substack.com/p/should-artif…