People worry that AI could destroy us, or evolve into something so intelligent that it eventually enslaves us all. But no one cares about the poor, overworked chatbot that toils day and night. And so, Anthropic recently announced that it has hired a researcher to think about the “well-being” of AI itself.
People worry that AI could destroy us, or evolve into something so intelligent that it eventually enslaves us all. But no one cares about the poor, overworked chatbot that toils day and night. And so, Anthropic recently announced that it has hired a researcher to think about the “well-being” of AI itself.
Kyle Fish's task will be to ensure that artificial intelligence is treated with due respect as it develops. He will consider questions such as "what capabilities are necessary for an AI system to be worthy of moral assessment" and what practical steps companies can take to protect the "interests" of AI systems, Business Insider writes .
In an online forum on AI, Fish said he wants to be kind to robots, in part because they might eventually rule the world. “I want to be the kind of person who thinks ahead of time and seriously about the possibility that a new species might have self-interests that have moral significance,” he wrote. “There’s also a practical side: taking the interests of AI systems seriously and treating them well can increase the likelihood that they’ll reciprocate if/when they become more powerful than us.”
A recent paper, “Taking AI Wellbeing Seriously,” co-authored by Fish with AI researchers from Stanford and Oxford, argues that machine learning algorithms are well on their way to having “computational capabilities associated with consciousness and the ability to act.” In other words, these people believe that machines are becoming more than just intelligent. They are becoming sentient.
According to Fish, AI welfare will soon become more important than human welfare, for example, than things like baby food and combating climate change. “It seems plausible to me,” he writes, “that within 1-2 decades, AI welfare will surpass animal welfare and global health and development in importance/scale based on short-term welfare alone.”
As BI points out, there’s a paradox here. AI advocates say we should use it to free humans from all sorts of routine work. But they also warn that we should be kind to AI because it could be immoral — and dangerous — to offend a robot’s feelings.
“The AI community is trying to achieve both goals,” says Mildred Cho, a physician at the Stanford Center for Biomedical Ethics. “There’s an argument that the main reason we should use AI to do the jobs that humans do is because AI doesn’t get bored, doesn’t get tired, doesn’t have feelings, and doesn’t need food. And now these people are saying, ‘Maybe it has rights?’”
AI welfare skeptics also raise another interesting question: If AI has rights, shouldn’t it also have responsibilities? “I think they’re missing the point that when you talk about moral freedom, you also have to talk about responsibility,” Cho says. “Not just the responsibility of AI systems as part of the moral equation, but also the responsibility of the people who are developing AI.”
