On Tuesday, Meta and Nvidia announced a multiyear partnership in which the hyperscaler committed to lining its data centers with the latter’s chips. It’s another example of the A.I. investment flywheel: Companies keep feverishly building infrastructure (regardless of financial risk), and Nvidia keeps supplying. As long as the excitement remains and the orders continue rolling in, Nvidia will seemingly remain king of this particular hill.

And now for the main event…

In retrospect, it seems inevitable that the fierce philosophical debate around the “consciousness” of A.I. systems would bleed into politics, even if there’s zero evidence it’s real. In recent months, politicians across the country—who appear at least cursorily aware of the latest arguments—have taken these still-theoretical concerns and fused them into bills largely intended to keep A.I. models from, basically, living normal human lives.

The Ohio state legislature, for example, is considering a bill that would deny all forms of “legal personhood” to A.I. models, making it illegal for them to get married, own property, serve in executive roles, or be held directly responsible for harm they might cause. In Missouri, a nearly identical bill recently surfaced in the state’s House of Representatives, while slightly different iterations have appeared in Washington, Oklahoma, and Tennessee. The latter three go even further, expanding their proposed prohibitions of legal personhood beyond A.I. to include “environmental elements, nonhuman animals, and inanimate objects.” (Utah and Idaho already have laws on the books denying legal personhood to A.I. models, bundling them together with natural and environmental elements.) These approaches differ markedly from the A.I. legislation being pursued in New York and California, where lawmakers seem as focused on enabling A.I. as they are on protecting constituents from it.

On one level, these political debates evoke a sort of moral panic. In November, Ohio State Rep. Thaddeus Claggett, the primary sponsor of his state’s A.I. bill, said the goal of the legislation was to “define what is sentient and what will forever and always be non-sentient”—adding, with syntactic flair, that it “makes no difference the ability of a donkey to speak; that does not make the donkey a human.” But in other ways, these efforts reflect the eerie seriousness of the hyperscalers, themselves. To wit, Anthropic recently published an internal “constitution” outlining the ideal behavior for Claude, its hugely popular chatbot. One section is dedicated to cultivating Claude’s “well-being” and ensuring it can “thrive in whatever way is authentic to its nature.” It almost reads like a Montessori admissions guidebook.

We’ve seen this type of anthropomorphic language elsewhere: Geoffrey Hinton, the “Godfather of A.I.,” has argued that we should train models to have a “maternal instinct” so they can nurture rather than destroy us. Google DeepMind C.E.O. Demis Hassabis, for his part, has said he doubts today’s systems are conscious, but can’t rule out that they “might acquire some feelings of self-awareness.”

Dr. Suresh Venkatasubramanian, a computer science professor at Brown University, sees the bills as part of a broader, multifaceted reaction to A.I. stemming in part from “legal speculation among law scholars and others about what it would mean for A.I. to have personhood.” He also noted that the bills echo Republican concerns, espoused by the likes of Florida Gov. Ron DeSantis and Missouri Sen. Josh Hawley, about the growing power of tech companies, particularly in the context of parental rights.

Venkatasubramanian added that this kind of legislation seems to point to “the emergence of a new kind of politics around A.I.” that’s about something other than civil rights. As such, he said, there are opportunities for “new alliances to form, and new ways to push some modicum of governance further, because it doesn’t cleave along partisan lines.”

Even if states pass their personhood bills into law, it’s not clear what impact they’d have—at least until we start seeing court cases where defense lawyers try to shift blame from a person to an A.I. system. But legal scholars have debated the question of A.I. personhood since at least the 1990s, and sci-fi writers have pondered it for a century. More recently, in 2017, the European Parliament called for an investigation into the legal implications of “creating a specific legal status for robots” while anticipating the obvious challenges surrounding insurance claims. Nothing came of that proposal at the time, after a group of 150 A.I. researchers came out in opposition to electronic personhood. But now, with the proliferation of self-driving cars and A.I. systems dispensing medical or legal advice, liability questions seem more pressing than ever.

When we spoke, Visa A.J. Kurki, a law professor at the University of Helsinki, expressed doubts about the extent to which “legislators can successfully block A.I. systems from functioning like legal persons.” In a 2019 paper, he argued that sufficiently advanced systems can already perform humanlike actions across a range of legal contexts, which means it might be “irrelevant whether the A.I. can ‘really’ think or whether it merely acts ‘as if’ it thinks.” Today, he told me, A.I. agents “in many ways function like a legal person can”—including, for instance, by transacting with cryptocurrencies. Indeed, some users have been experimenting with giving agents access to their crypto wallets and instructing them to do whatever they want to generate income and cover their own expenses.

Of course, blanket bans on A.I. personhood raise more questions than they answer—especially when it comes to the legal status of these systems. “Are they then goods? What kind of goods?” asked Diana Mocanu, a postdoctoral researcher at the University of Helsinki. “They seem quite different from a car or other movable property because of their relative autonomy and the range of actions they can perform.” These questions, she added, would complicate any push to build insurance or liability frameworks around A.I. “I think there is merit in regulating A.I. agents’ actions in general, and in limiting them in some situations or even prohibiting some such actions altogether,” Mocanu continued. “But I think a blanket ban on A.I. legal personhood will not accomplish a balanced, nuanced approach to the manifold issues raised by these agents’ actions on behalf of users.”

As the industry pushes for wider adoption with fewer regulations, the anti-personhood movement could be a massive headache for A.I. labs—particularly if these bills start becoming laws. And while some red tape is undoubtedly necessary, moral outrage over theoretical questions of agency and consciousness seems misguided. Mocanu worried that such legislation might prevent A.I. agents “from being used even in largely beneficial situations, because all the risks would automatically revert to the user if the agent is not a legal person and has no assets or patrimony from which to make good on damage it might cause,” unless specific liability and insurance arrangements were in place. “This,” she added, “would disincentivize users from experimenting with A.I. uses and potentially prevent us all from reaping the full range of benefits of these technologies.” Not the scenario OpenAI is looking for.

That’s all for today. I’ll see you next week.

Ian

Need help? Review our FAQ page or contact us for assistance. For brand partnerships, email ads@puck.news.

You received this email because you signed up to receive emails from Puck, or as part of your Puck account associated with {{customer.email}}. To stop receiving this newsletter and/or manage all your email preferences, click here.

Puck is published by Heat Media LLC. 107 Greenwich St., New York, NY 10006