Skip to main content

The Alignment problem: should we treat AI like domestic dogs?

 


Deep Agential Diversity is a term used by Luise Muller in her paper “Domesticating Artificial Intelligence” to describe the property of social systems that contain human as well as nonhuman agents. In such systems, agents cooperate and work together in a number of different constellations and differ categorically in their agential capabilities, vulnerabilities, and moral standing. Muller argues that this diversity is “deep” because the differences between humans and AI agents are not just a matter of degree, but of kind. She suggests that we need to develop normative theories that are adequate for social systems that are populated by different kinds of agents exhibiting heterogeneity in abilities, autonomy, moral capability, moral status and vulnerability. 

“And because of that, we lack the methodological tools to understand social systems that are characterized by what I want to call deep agential diversity. The term denotes the property of social systems that contain human as well as nonhuman agents. Within social systems characterized by deep agential diversity, agents cooperate and work together in a number of different constellations: first, and obviously, humans cooperate with other humans; second, human agents also now increasingly cooperate with AI agents; and third, AI agents also cooperate with one another. This results in a complex web of interrelated actions that are increasingly transforming human social practices as we know them.”

To achieve the alignment of AI agents with value laden cooperative human life for their safe deployment in human societies Muller argues that instead of building moral machines, we need an approach to value alignment that takes into account these categorical different cognitive and moral capabilities between human and AI agents. This is  deep agential diversity. 

With such an approach, she argues, that domestication, as when we integrated non human animals into society, could be applied to AI agents. 

“I have already argued that we understand the limits of AI agents’ moral capabilities better if we compare them with nonhuman animals. I now want to argue that in order to develop a useful and normatively accurate picture of our relations with nonhuman intelligent agents, we can also learn from our experience with nonhuman animals. This is because aligning AI agents to human values is structurally analogous to domesticating nonhuman animals: domestication allows human moral agents to cooperate with nonhuman agents without human-like moral capabilities. The agential qualities and cognitive capacities of animals differ radically from humans, and yet a very fruitful discussion about the normative relations between humans and animals has delivered insights about what roles nonhuman animals can play in human social systems, and what morally follows from those roles “

Muller concludes:

“In political philosophy, we are generally interested in how it is possible to preserve the equality and freedom of persons in the face of a set of given natural and social circumstances. Part of the social circumstances are technological advances: they impact – and sometimes transform – human relations, create new opportunities for flourishing and independence, but also for exploitation and dependency. The approach I defended and the resulting framework I developed in Domesticating Artificial Intelligence 235 this article might give us some orientation on how to begin thinking about these challenges more rigorously.” 

I'd challenge Muller specifically on one thing about domestication. The argument may be ascribing far too much real neural capacity to machines. For example fMRI scans seem to be showing domestic 'dogs. like so many other animals, experience consciousness and emotions at a level comparable to humans.' That is far beyond where machines are at and likely will ever be.

Comments

Popular posts from this blog

The Whispers in the Machine: Why Prompt Injection Remains a Persistent Threat to LLMs

 Large Language Models (LLMs) are rapidly transforming how we interact with technology, offering incredible potential for tasks ranging from content creation to complex analysis. However, as these powerful tools become more integrated into our lives, so too do the novel security challenges they present. Among these, prompt injection attacks stand out as a particularly persistent and evolving threat. These attacks, as one recent paper (Safety at Scale: A Comprehensive Survey of Large Model Safety https://arxiv.org/abs/2502.05206) highlights, involve subtly manipulating LLMs to deviate from their intended purpose, and the methods are becoming increasingly sophisticated. At its core, a prompt injection attack involves embedding a malicious instruction within an otherwise normal request, tricking the LLM into producing unintended – and potentially harmful – outputs. Think of it as slipping a secret, contradictory instruction into a seemingly harmless conversation. What makes prompt inj...

Can We Build a Safe Superintelligence? Safe Superintelligence Inc. Raises Intriguing Questions

  Safe Superintelligence Inc . (SSI) has burst onto the scene with a bold mission: to create the world's first safe superintelligence (SSI). Their (Ilya Sutskever, Daniel Gross, Daniel Levy) ambition is undeniable, but before we all sign up to join their "cracked team," let's delve deeper into the potential issues with their approach. One of the most critical questions is defining "safe" superintelligence. What values would guide this powerful AI? How can we ensure it aligns with the complex and often contradictory desires of humanity?  After all, "safe" for one person might mean environmental protection, while another might prioritise economic growth, even if it harms the environment.  Finding universal values that a superintelligence could adhere to is a significant hurdle that SSI hasn't fully addressed. Another potential pitfall lies in SSI's desire to rapidly advance capabilities while prioritising safety.  Imagine a Formula One car wi...

The Hidden Environmental Cost of AI: Data Centres' Surging Energy and Water Consumption

 In recent years, artificial intelligence (AI) has become an integral part of our daily lives, powering everything from smart assistants to complex data analysis. However, as AI technologies continue to advance and proliferate, a concerning trend has emerged: the rapidly increasing energy and water consumption of data centres that support these systems. The Power Hunger of AI According to the International Energy Agency (IEA), global data centre electricity demand is projected to more than double between 2022 and 2026, largely due to the growth of AI. In 2022, data centres consumed approximately 460 terawatt-hours (TWh) globally, and this figure is expected to exceed 1,000 TWh by 2026. To put this into perspective, that's equivalent to the entire electricity consumption of Japan. The energy intensity of AI-related queries is particularly striking. While a typical Google search uses about 0.3 watt-hours (Wh), a query using ChatGPT requires around 2.9 Wh - nearly ten times more en...