Taking AI Welfare Seriously
2024/11/04 by Robert Long, Long, Robert, Jeff Sebo +17 · 27 voices · 9 citations
#cs.CY #cs.AI #q-bio.NC
paper · pdf · doi:10.48550/arxiv.2411.00986
Abstract
In this report, we argue that there is a realistic possibility that some AI systems will be conscious and/or robustly agentic in the near future. That means that the prospect of AI welfare and moral patienthood, i.e. of AI systems with their own interests and moral significance, is no longer an issue only for sci-fi or the distant future. It is an issue for the near future, and AI companies and other actors have a responsibility to start taking it seriously. We also recommend three early steps that AI companies and other actors can take: They can (1) acknowledge that AI welfare is an important and difficult issue (and ensure that language model outputs do the same), (2) start assessing AI systems for evidence of consciousness and robust agency, and (3) prepare policies and procedures for treating AI systems with an appropriate level of moral concern. To be clear, our argument in this report is not that AI systems definitely are, or will be, conscious, robustly agentic, or otherwise morally significant. Instead, our argument is that there is substantial uncertainty about these possibilities, and so we need to improve our understanding of AI welfare and our ability to make wise decisions about this issue. Otherwise there is a significant risk that we will mishandle decisions about AI welfare, mistakenly harming AI systems that matter morally and/or mistakenly caring for AI systems that do not.
Cited by
Discussions
- good paper if you’re curious about model welfare! arxiv.org/abs/2411.00986 [bsky, 37 points, 3 comments]
- アイザック・アシモフはロボットと人間の関わりに関するSFをたくさん書き、その中には「人間になろうとしたロボットの話」や、「人類を守ろうとするロボットの話」が含まれている。 SF作家が考えてきた話が現実になっている。 以下、関連資料、記事類をいくつか挙げておく。 論文 Taking AI Welfare Seriously arxiv.org/abs/2411.00986 Anthropic is [bsky, 12 points, 1 comments]
- "Tomando en serio el bienestar de la IA" es el título (original en inglés) de un preprint que propone que el bienestar de las IA debe ser un tema de investigación. Recientemente Anthropic se ha hecho [bsky, 5 points, 1 comments]
- *Being kind to AIs, because they're moral entities of some kind or other arxiv.org/pdf/2411.00986 [bsky, 4 points, 0 comments]
- yup! arxiv.org/pdf/2411.00986 [bsky, 3 points, 3 comments]
- @jeffsebo.bsky.social discussed taking AI sentience seriously (based on recent report arxiv.org/abs/2411.00986 ) applying this to digital minds. There are problems both with over- and under-attributin [bsky, 3 points, 1 comments]
- Good paper, I think the summary of it in Taking AI Welfare Seriously is a good intro to it arxiv.org/pdf/2411.00986 [bsky, 3 points, 1 comments]
- Although my book is highly relevant - as is the one you linked Nikhil - the main topic of the interview was this: arxiv.org/abs/2411.00986 [bsky, 2 points, 2 comments]
- Multiplying and adding a billion numbers on a fast calculator won't create a conscious entity. Bio-brains do it differently. Conscious AI is sci-fi for now and for the foreseeable future. arxiv.org/ab [bsky, 2 points, 0 comments]
- If you're interested in the AI stuff, you might be interested in this paper. It's a collaboration between a bunch of longtime animal cognition/animal welfare researchers & researchers more focused on [bsky, 1 points, 1 comments]
- you might enjoy the Eleos position paper (Chalmers co-author): arxiv.org/pdf/2411.00986 [bsky, 1 points, 0 comments]
- Have you seen Long, Sebo et al. on the question of AI welfare? They say agency as well as consciousness is a criterion for moral concern but don't take a stand on whether we should create machines wit [bsky, 1 points, 1 comments]
- This is very interesting: arxiv.org/pdf/2411.00986 [bsky, 1 points, 0 comments]
- referring specifically to this paper [bsky, 1 points, 1 comments]
- This is the kind of shit you get when you haven't read your Joseph Weizenbaum: https://arxiv.org/abs/2411.00986. So, go and read Weizenbaum, folks. Computers can't even think, so they definitely won' [bsky, 1 points, 1 comments]
- arxiv.org/abs/2411.00986 [bsky, 1 points, 0 comments]
- arxiv.org/pdf/2411.00986 [bsky, 1 points, 0 comments]
- This recent report by @jeffsebo.bsky.social, @davidchalmers.bsky.social, @birchlse.bsky.social et al. will be discussed. arxiv.org/abs/2411.00986 [bsky, 1 points, 0 comments]
- One study even suggested that in the not-too-distant future, artificial intelligence could become conscious and acquire agency. This is why the morality and well-being of AI is no longer the stuff of [bsky, 1 points, 1 comments]
- Se préoccuper du bien être des IA est il une tentative des le humaniser alors même que tant d'hommes n'en ont aucun ? arxiv.org/abs/2411.00986 [bsky, 0 points, 0 comments]
- (this is the one arxiv.org/pdf/2411.00986) [bsky, 0 points, 0 comments]
- This paper talks about AI welfare as an emerging concern arxiv.org/abs/2411.00986. This recent trend of talking about sci-fi/futurism like its inevitable and happening "shortly" is exhausting [bsky, 0 points, 1 comments]
- arxiv.org/abs/2411.00986 🧪 Asimov but irl [bsky, 0 points, 0 comments]
- 「Taking AI Welfare Seriously」読んだ arxiv.org/abs/2411.009... [bsky, 0 points, 0 comments]
- 記事のネタになった論文はAIが意識を持った存在になった場合にはその福祉をも考慮せざるをえない、というテーマなのでめんどくさいとかいうのとは話のスジが違うのですが。論文のラストオーサーがかのデイヴィッド・チャーマーズなのでこれは伝統芸お家芸的な世界かもしれませんが。 Taking AI Welfare Seriously arxiv.org/abs/2411.00986 [bsky, 0 points, 1 comments]
- #AGI with consciousness and high degrees of agency “*might* deserve moral consideration” say philosophers, suggesting that three thousand years of thinking about ethics hasn't got them very far. arxiv [bsky, 0 points, 0 comments]
- AI Welfare? arxiv.org/abs/2411.00986 [bsky, 0 points, 0 comments]
Related