You might expect that a venerable high-brow publication would deliver industry-leading analyses penned by the most knowledgeable and experienced people across a range of fields. You might expect pieces published in this venue to be carefully scrutinized by editors, meticulously vetted by subject-matter experts, and thoroughly fact-checked for accuracy. You might also expect that these tasks would be executed with additional rigor in the context of a cover story on a controversial subject.
Well, when it comes to the latest issue of The Economist, you’d be wrong. The September edition of the magazine leads with a piece entitled, “Could AIs become conscious? The mind bending question humanity must face.” Aside from the fact that the title is grammatically incorrect (AIs?), overall it’s just not good. In fact, it might be the worst writing on AI consciousness found in a major periodical. You can read it here.
Over the past several years, there have been a number of prestigious outlets that have published work on this specific question. I have cited some of them in previous Substack essays. They are a real mixed bag, but none so misguided as the present one.
For example, a PhD student at Cambridge wrote an essay in The Telegraph arguing that the idea of AI welfare is “absurd.” The op-ed piece is a biocentric screed dressed in narrow philosophical garb, and it gets a lot of things wrong about AI moral status. It clings to certain Enlightenment tropes and treats them as if they have a monopoly on the truth. Mainly, the author assumes that an entity needs to be alive in order to be conscious or sentient, so because AI isn’t alive it can never be either. Case closed. An academic article appearing in the journal Animal Sentience unsurprisingly lodges a broadside along the same lines—AI isn’t alive, so it doesn’t deserve any kind of protection. These are deeply flawed works, but they do not command the kind of attention that a cover story in a major periodical might.
By contrast, the essay appearing in next month’s Economist takes bad argumentation on the moral or legal status of AI to new heights. It is absolutely riddled with conceptual inconsistencies and philosophical leaps, illustrating just how poorly this entire discourse is understood. To substantiate my claim, I have pulled out some of the most egregious errors in the form of direct quotes (in bold) and briefly explained why they are wrong. See the line-by-line take down below.
“Most people believe humans are different. The reason is that Homo sapiens has evolved to be conscious.”
This is an opinion masquerading as fact. Yes, most people do believe humans are different from non-humans. However, they disagree about why.
“Many animals feel pain and pleasure; some appear to be self-aware.”
Now we have already moved on from consciousness to sentience and self-consciousness. The author is just sliding from one concept to the next without any care.
“Many people feel that to confer personhood on AIs would be an abomination.”
Here the author references the Pope’s latest encyclical, Magnifica Humanitas (which I previously wrote about here), and jumps from his concern with sentience (not consciousness explicitly) to personhood. But there are many kinds of personhood. We can only presume the author is talking about MORAL personhood, because it plays by different rules than does LEGAL personhood.
“Whether you believe in a human soul, or that consciousness requires brain cells, or that a mind can indeed emerge from computations in silicon alone, humanity is destined to confront something entirely new.”
This claim obscures the nature of the debate in the philosophy of mind. It’s not strictly over whether “inner life” merely emerges from anything regardless of substrate. Rather, it’s about whether consciousness is an exclusively biological phenomenon or whether a functional equivalent can take shape in an inorganic entity.
“to treat AIs as having even limited rights would be very dangerous.”
This statement begs the question. It suggests that granting rights to one entity necessarily takes away from the rights of (in this case) humans. Even human rights scholars like Sushma Raman and Bill Schulz reject this zero-sum framing. It’s scaremongering at best and ignorant of rights theory at worst.
“Already Argentina’s president, Javier Milei, has taken a step towards legal personhood for AI by proposing that bots should be allowed to run companies.”
Now we have shifted from talking about moral personhood to legal personhood without any recognition of the different rules that apply to both.
As you can see, the claims in the article are all over the place and they evince a stunning lack of familiarity with the main concepts and arguments found in the literature. A graduate student working on AI legal personhood or AI moral rights would have easily come up with a more coherent, well-informed analysis (well, maybe except for that Cambridge student).
The current state of knowledge on AI consciousness stipulates that we don’t really know what consciousness is, can’t really measure it, and can’t even agree upon whether it is something internal to a being or externally diffused throughout the world. Perhaps even more importantly, we don’t know whether it is necessary for moral status or merely sufficient for its ascription. Or something else entirely. It’s all still up for debate and the probability of reaching consensus in the near future is quite low.
Whether we are discussing AI rights or welfare or consciousness, we need to approach the topic with scholarly precision, and that requires a strong grasp of complex legal and philosophical constructs. But this essay (which comes in at just under 1,000 words) offers none of that. In fact, it totally misunderstands the literature on the moral and legal status of AI. It’s hard to believe this was published as-is, let alone made the cover of The Economist.
If I were the author, I would just be glad I placed this in an outlet where no one would know I wrote it.
Coda
There is another essay in the September Economist that addresses the question of AI consciousness. However, this one, written by Google CTO Blaise Agüera y Arcas, makes the case that it’s not consciousness per se that matters in the moral calculus. Instead, Agüera y Arcas finds that “As intelligent, social entities, we decide (or, to some degree, our evolutionary history has decided for us) which other entities we regard as having inner lives and being worthy of care. We do not care for others because they are conscious. Rather, we believe they are conscious when and because we care about them.”
This observation owes an intellectual debt to feminist care ethics à la Maria Puig de la Bellacasa (who wrote the fantastic book, Matters of Care) and the relational approach to moral status developed by David Gunkel and Mark Coeckelbergh (you can read their seminal article, initially focused on animals before they applied it to AI, here). Agüera y Arcas is working on a book about AI and consciousness, so hopefully he gives proper credit to those whose contributions paved the way for his own.
Most interesting and yes as often we agree.
Aside from the fact that the title is grammatically incorrect is a small detail that is doing a lot of work as an opener, because it signals the piece was not reviewed by anyone who actually thinks about this professionally, and that turns out to predict the substance too. The pattern worth naming across outlets is that pieces on AI moral status keep getting assigned to generalists rather than to the people who have spent a career on the underlying theory, which would never happen on a comparably technical physics or medicine story.