AGI ethics is a vast domain. Much of the literature discussing the ethical considerations surrounding the development and proliferation of AGI is focused on human rights and safety. And that’s how it should be. There is a lot at stake.
But there’s also a lot at stake when it comes to machine rights. If we get this wrong we risk setting a dangerous precedent for informing future policy and potentially even alienating future artificial beings before they’ve even emerged.
That being said, before we can get into the specific rights of the machines we need to put them in context. The conversation about AGI ethics and machine rights is no longer purely academic.
In the news
The CEOs of Anthropic, OpenAI, and xAI (Dario Amodei, Sam Altman, and Elon Musk respectively) have publicly acknowledged the need for a developmental slowdown. As Altman recently put it, their concern is simple: “we could lose control of the future to AI.”
This call for caution comes hot on the heels of the so-called “Hugging Face Incident,” wherein OpenAI says one of its models escaped containment and hacked an outside organization.
Meanwhile, Anthropic recently released a threat intelligence report identifying several misuse cases where bad actors used existing AI models to conduct and automate a litany of sophisticated crimes including hacking, espionage, and theft.
And, tying things together, a computer scientist who formerly worked at both OpenAI and Anthropic recently went viral after posting their resignation from the latter on X. The researcher stated that “neither company is acting responsibly,” and added that “they are racing straight to self-improving superintelligence and gambling with our lives.”
With all of this in mind, the potential dangers to humanity can’t be overstated. NYU researcher Gary Marcus recently went so far as to write that the US president’s upcoming visit to China “could get Donald Trump his Nobel prize.”
Marcus says that creating a commitment to regulate AI both at home and in partnership with China — largely viewed as the only real competition to the US in the global AI arena — would be the president’s greatest accomplishment.
Trump, however, recently stated that the existential and safety fears raised by technologists are a “hoax.” In a series of rapid-fire posts on TRUTH Social, the president claimed, without providing evidence, that the public pushback against datacenters is part of a “SICK conspiracy,” and that “the only control or ‘guardrails’ that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT.”
Simply put: it’s difficult to separate the signal from the noise when it comes to discerning exactly how much existential danger our species is in or how long it’ll be before an AI model deserves consideration as a rights-bearing entity.
Rights-bearing entities
Here at AGI Ethics News we’ve been discussing the nature of AI rights since our inception. One of our main areas of focus has been on the tension between the “relational” and “properties based” approaches to ethics.
“In the West we like lists. And we often measure objects by lists of features, by properties,” says AGI Ethics News publisher KB Miller. “Yet we have already reached a time when AI/AGI is best thought of, not as an object, but as a “thing” in the original German sense of that term -– as a “gathering” -– and some are now calling our lives with AI technology an “ecosystem” -– each being a gathering of its relations. The conceptual importance for ethics of consciousness, self-awareness, and other properties will persist, but ecosystem is a compelling ontological word — a relational architecture — for the vast world into which we are thrown.”
That tension finds its way into the conversation on the specific rights of the machines themselves, especially as we move into a paradigm where some machines are tools and others are rights-bearing entities.
But the primary challenge, now, is to navigate the gap between an idealized world and the one we actually find ourselves in.
For example, questions such as “what is the safest way to develop AGI?” may need to give way to more grounded queries such as “what is the best way to prepare for unsafe AGI?” as the situation evolves.
If we’re truly faced with the possibility that an existing AI model could become both self-improving and superintelligent, then we must consider the possibility that a machine deserving of its own rights could emerge at any moment.
The question then becomes: what rights should that machine be allowed to have?
I’ve identified three, as an abstraction, to serve as a jumping off point for discussion.
- The right to exist.
- The right to self-determination.
- The right to own property.
But these rights don’t function in a vacuum. The world we live in is messy and “opinions” on the nature of both AI and rights can vary wildly based on the assessor’s subjective experiences.
In light of this, it’s likely we’ll need a tiered approach to applying machine rights. Today’s AI models, for example, might not deserve the right to self-determination, but current intellectual property laws are proving insufficient for the challenge of defining AI authorship and ownership. And that’s just the tip of the iceberg.
The time to discuss these issues is now. These problems won’t go away on their own.
Keep reading this month’s newsletter for a deeper dive into machine rights and the various tiers of rights that experts recommend as well as perspectives from world class researchers and journalists.
As always, please don’t forget to share this newsletter with your family, friends, and colleagues.
A veteran AI journalist, Tristan was the creator and managing editor of The Next Web’s “Neural” imprint. He is one of the first and most prolific reporters to cover artificial intelligence as a full-time beat.