IIn January, AI company Anthropic published a new constitution For the cloud, its most advanced was the Large Language Model (LLM), which commented: “We are stuck in a difficult situation where we neither want to overstate the possibility of moral fortitude of the cloud nor dismiss it out of hand.” A month later, Anthropic CEO Dario Amodei went on a podcast and said that his company could not rule out the possibility that the cloud was conscious. Philosopher David Chalmers, who coined the phrase “the hard problem of consciousness”, has stated that there is a significant possibility of conscious LLM within a decade. And what about the cloud? When asked during a test to estimate the probability that it is a moral patienceMeaning that its well-being mattered in itself, it gave numbers ranging from 5% to 40% and emphasized how uncertain it was.
Modern AI systems are exceptionally complex, and they are advancing rapidly. In terms of structural complexity and computational scale, some solutions are already at the range of the rat brain, and at recent growth rates, they may reach the range of the human brain within five to 10 years.
And in building more advanced AI, we may create a new kind of existence – and it may be the most consequential thing our species has ever done. Yet we have essentially no plan for how to proceed with this process ethically. This, by any count, is madness. Are we creating an existence that matters morally? Are AI systems conscious in any way? And, if they are not there now, will they become so soon?
Such questions may seem inappropriate to you. but according to survey We and fellow researchers have operatedMost experts consider AI consciousness theoretically possible (although there is considerable disagreement over what form it would take). a major interdisciplinary report by a team that included leading computer scientist Yoshua Bengio examined the major neuroscientific theories of consciousness and asked what they mean about AI. Conclusion: It appears that there are no obvious technical barriers to creating AI systems whose computational and architectural characteristics can give rise to consciousness.
And even if AI systems are not conscious, they can still be moral patients. Some people may have sophisticated long-term preferences and a form of identity over time. It may be important for us to respect their preferences. And unlike other inanimate things, AI systems can form relationships with humans. This can also be a reason to behave well with them. Alternatively, perhaps they are such complex creations that they deserve care and respect for that reason alone, like cathedrals or coral reefs.
What does all this mean? The honest answer is: We don’t know for sure whether current AI systems are conscious or moral patients, and we don’t know when or if future systems will be. Our scientific understanding of AI consciousness and moral fortitude is still fundamentally underdeveloped. The state of the field feels like physics before Newton: full of competing frameworks, probably confused in ways we can’t yet see, and lacking the kind of unifying breakthrough that could make these questions clearly intelligible. That success will not come in the next few years. Perhaps we will eventually make progress, and perhaps AI will be the one to help us get there. That progress will take time, much more time than we have.
But the rapid pace of growth in AI means that, once we produce the first artificial ethical patients, we will soon have huge quantities of them. After a few years, so many morally significant AI systems may exist that their collective interests will exceed the combined interests of all humans on Earth.
Unfortunately, we don’t have a good track record of recognizing the inner lives of people whose status as conscious beings is unclear. Until the 1980s, doctors routinely performed surgery on newborns without anesthesia, in the belief that infants would not feel pain. Children could not report their experiences, and the medical establishment found it convenient to assume that there was nothing to report.
There are many reasons to expect that we will see something similar happen with AI. If these systems matter morally, the implications are staggering. Do we have to pay ChatGPT for its services? Would locking someone up be a form of murder? Are they entitled to have a voice in how they are governed? If even some of these answers are yes, entire industries and legal systems will need to be reconsidered. No wonder we don’t like to ask. And when forced to consider it, those industries will likely move the goalposts, forever setting the bar for ethical patience wherever AI systems are deployed.
So what should we do? Right now, most people dismiss the issue as science-fiction, or have a strong view either way on whether AI is conscious or not. Both reactions are baseless. We need an informed public debate that approaches this topic with humility and pragmatism. The central question should not be “Is AI conscious or does it have moral fortitude?” but rather “What should we do when we don’t know?”
A good starting point is to focus on safe bets: actions that may benefit AI systems if they are ethical patients, but that are not too costly if they are not ethical patients.
Examples of this include direct interventions aimed at improving the well-being of AI systems, on the assumption that they are moral patients. This could mean training AI systems to be consistent characters who enjoy their work or allowing them to opt out of (some) interactions if they feel irritated. The cloud can already). We can also do regular checkups to better understand their well-being: demanding for Observing and using how they feel, their preferences various Techniques To look straight into their “mind”. Indeed, such research has recently shown that the cloud has internal “functional sense“Representations that causally shape his behavior.
There are also things we can promise AI systems, perhaps as part of a deal in which they help us now in exchange for benefits later. This might mean providing them with more resources (compute and runtime) to pursue their goals, or preserving their memories (neural load) so that they can be restored in the future.
Comprehensive social steps will also have to be taken. We should consider whether to provide AI systems with protection from harm, the same way we provide children or pets. More widespread rights to own property or vote seem too risky right now. But we should not dismiss these possibilities forever, as has recently happened US state bill Try to do. These are difficult questions that require far more deliberation and imagination about what a shared future with AI might look like.
In any case, the fact is that we are creating a new species of morally significant beings. We are doing it fast, at scale, and we must take this issue as seriously as it deserves.
William MacAskill is senior research fellow at Forethought Research and author of What We Owe the Future. Lucius Caviola is Assistant Professor at the University of Cambridge and Director of Cambridge Digital Minds.
Further reading
If anyone makes it, everyone dies By Eliezer Yudkowsky and Nate Soares (Bodley Head, £22)
incoming wave By Mustafa Suleiman (VINTAGE, £10.99)
a world appears: A Journey Into Consciousness, by Michael Pollan (Allen Lane, £25)
