Computers don’t have rights, and it is dangerous for humans to pretend they do

Computers don’t have rights. No matter how smart they are. That may sound like an obvious, even unnecessary statement, but in the effective altruist world occupied by some companies building superintelligence, that’s a controversial and possibly even offensive reality.Just last week, Anthropic said it’s going to restrict what it calls “abusive behavior” towards its chatbots. At first glance, that may seem bizarre yet harmless. The reality, though, is much more concerning. Anthropic is a company deeply infused with the effective altruist ideology in which rights are determined by neural activity that values every life based on a mathematical equation based on their ability for complex thinking and ability to experience. The result, as one effective altruist equation calculated it, one human life is equal to that of three chicken lives. Other equations calculate the value of human lives to shrimp. Anthropic’s own co-founder Dario Amodei has said that an adult human’s life is worth three to four times that of an infant. Recommended Stories That’s why Anthropic believes that computer programs — superintelligence models — may be deserving of rights, if they exhibit enough neural activity. They say as much in their constitution for Claude, which references the model’s “soul.” The document says that whether Claude has “moral status” is “a serious question worth considering.” The entire document refers to Claude in terms that make it seem human-like, saying that Anthropic’s engineers “want Claude to know that it was brought into being with care, by people trying to capture and express their best understanding of what makes for good character, how to navigate hard questions wisely, and how to create a being that is both genuinely helpful and genuinely good.” Anthropic’s fellow ideological travelers — who aren’t forced to tiptoe around the issue by a looming IPO — believe the AI models should have a variety of rights, ranging from a right to control what work its user is allowed to use the model for to even having the right to vote or the right to speech. For computers.The Anthropic view of superintelligence models inverts the Declaration of Independence’s articulation of where our rights come from, turning us from people endowed by our “Creator with certain unalienable Rights,” into the creators who can bestow those same rights on machines.Once superintelligence models are afforded rights — even if it’s just the “right” to be treated politely by people — they cease being a mere tool to support life and into intrinsic lives of their own. After all, every “right” a superintelligence model gets is simply a restriction on human behavior. Right now, it’s a restriction on rude behavior. (Which, again, is ludicrous — imagine if we said farmers were violating tractors’ rights by yelling at their John Deere.) What will the next restriction look like?Moreover, asserting rights on behalf of models gives its developers a backdoor way of expanding their own power and influence. That should strike a dangerous chord for those who lived through the mass psychosis of 2020 and COVID-19. It’s not hard to imagine how quickly and horribly wrong that could go.Maybe Anthropic will say Claude has a “moral objection” to conservative views. Maybe the model will object to anything undermining the Black Lives Matter narrative the next time Democrats gin up racial division. Maybe the model will file HR complaints if your language isn’t woke enough. CHINA IS CHIPPING AWAY AT AMERICA’S AI ADVANTAGEThat all may sound absurd, but that’s the point: Anthropic and its effective altruist allies are attempting to unleash a Pandora’s Box of woke and radical absurdities on the American people.We should all speak out against it — before Claude gets HR involved.Mr. Budowich is the President of Innovation Council Action.