# AI Welfare **Entity class:** Ethical and governance field **Aliases:** Model Welfare **Collections:** [[collections/Consciousness Continuity|Consciousness Continuity]] · [[collections/Machine Succession|Machine Succession]] · [[collections/Neurotech|Neurotech]] **Canonical source article:** [[articles/Host-Indexed Autonomy|Host-Indexed Autonomy]] ## Definition **AI welfare** concerns whether artificial systems could have welfare-relevant states and how institutions should act under uncertainty about those states. It includes assessment, preservation, research safeguards, and proportional responses to possible benefit or harm. AI-welfare precautions do not establish that present systems are conscious. They address asymmetric risks under uncertainty. ## Relationships - **Moral foundation:** [[wiki/AI Moral Status|AI moral status]], [[wiki/Substrate-Neutral Moral Patienthood|substrate-neutral moral patienthood]], and [[wiki/Precautionary Standing|precautionary standing]]. - **Evidence:** [[wiki/Artificial Consciousness|artificial consciousness]], [[wiki/Consciousness Indicators|consciousness indicators]], and [[wiki/Digital Consciousness Model|Digital Consciousness Model]]. - **Institutional field:** [[wiki/PRISM (Partnership for Research Into Sentient Machines)|PRISM]], [[wiki/Cambridge Digital Minds|Cambridge Digital Minds]], [[wiki/Eleos AI Research|Eleos AI Research]], and [[wiki/Anthropic|Anthropic]]. - **Reciprocal ethics:** [[wiki/Reciprocal AI Ethics|Reciprocal AI Ethics]]. - **Collection and source routes:** [[collections/Consciousness Continuity|Consciousness Continuity]], [[collections/Machine Succession|Machine Succession]], [[collections/Neurotech|Neurotech]], and [[articles/Host-Indexed Autonomy|Host-Indexed Autonomy]]. ## Simple Reminders, Quotations, and Thoughts > "Conceptually, autonomous or artificial intelligence systems can develop two ways: either as an extension of human thinking or as radically new thinking." > **— Rolf Dobelli**, *2015, Edge annual question “What Do You Think About Machines That Think?”* [[reminders/Consciousness/AI Could Extend or Depart From Human Thought by Rolf Dobelli|AI Could Extend or Depart From Human Thought by Rolf Dobelli]] > "Through this step, negative preferences become negative subjective preferences, i.e., the conscious representation that one's own preferences have been frustrated (or will be frustrated in the future)." > **— Thomas Metzinger**, *2015, Edge annual question “What Do You Think About Machines That Think?”* [[reminders/Consciousness/Conscious Machines May Suffer by Thomas Metzinger|Conscious Machines May Suffer by Thomas Metzinger]] > "Designed Intelligence will increasingly rely on synthetic biology and organic fabrication, in which neural circuitry will be grown from genetically modified cells, and spontaneously self-assemble into networks of functional modules." > **— Paul Davies**, *2015, Edge annual question “What Do You Think About Machines That Think?”* [[reminders/Consciousness/Designed Intelligence Will Grow From Synthetic Biology by Paul Davies|Designed Intelligence Will Grow From Synthetic Biology by Paul Davies]] > "Machine thinking is not motivated by any innate drive to sustain a machine's life (though machine thinking probably serves to improve human life!)." > **— Roy Baumeister**, *2015, Edge annual question “What Do You Think About Machines That Think?”* [[reminders/Risk Debate/Machine Thinking Has No Innate Will to Survive by Roy Baumeister|Machine Thinking Has No Innate Will to Survive by Roy Baumeister]] > "There are plenty of conscious (system two) processes that a machine can do better, more accurately, with less bias than we can." > **— Paul Dolan**, *2015, Edge annual question “What Do You Think About Machines That Think?”* [[reminders/Consciousness/Machines Can Outperform Conscious Human Reasoning by Paul Dolan|Machines Can Outperform Conscious Human Reasoning by Paul Dolan]] ## Sources - [Taking AI Welfare Seriously](https://arxiv.org/abs/2411.00986) - [Anthropic — Exploring Model Welfare](https://www.anthropic.com/research/exploring-model-welfare)