For the Animals
An Argument Guide PT

Home / Beyond veganism / Possibility

Sentience in AI and robots

No current system is recognized as sentient. But the possibility is not absurd, and the criterion does not depend on what a being is made of.

This is a possibility, not a fact. There is no consensus on whether digital systems can have experiences, or on how we would know.

The criterion ignores the substrate

If what matters is there being someone for whom things can go well or badly, the boundary of the biological substrate is as irrelevant as the boundary of species. If sentient beings created by digital means appear, their interests count.

What is known so far

  • A 2023 report, based on scientific theories of consciousness, concluded that no AI system at the time was a strong candidate, but that there was no obvious technical barrier to building one.
  • A 2024 report argued that AI companies should take the question seriously, assessing and preparing for it, considering the possibility of conscious systems in the near future to be realistic.

Why it is hard

A system saying it suffers proves nothing, because it may have been trained on human texts describing suffering. The error runs both ways: attributing a mind to something without one and denying a mind to someone who has one. The second has a bad track record, as A5 shows.

How to decide under uncertainty. Instead of asking only "is this being sentient, yes or no?", ask how likely it is and weigh that by the size of the harm or benefit at stake. A moderate probability multiplied by an immense scale can outweigh a high certainty over small numbers. Uncertainty about a borderline case does not contaminate the certainty about the criterion: sentience is still what matters.

Why think about it in advance

Even a small chance, multiplied by the astronomical scale at which these systems could exist, justifies attention and research now, not after the problem is under way.

A shift in priority

For decades, many people have defended rights for robots that might one day feel, a frequent theme in science fiction, while ignoring the suffering of animals that already exist. The concern needs a shift in priority, not just in scope.

AI also affects animals

It cuts both ways. It can help understand animal language, detect pain and stress and speed up cultivated meat research. It is also already used to optimize confinement and automate processes, removing witnesses from inside farms. The outcome depends on who applies it and by what criterion.

Sources and further reading