To the Editor:
Re “Executive’s Quest to Fathom the Soul of a ChatBot” (front page, Oct. 3):
The assertions of Christopher Olah and the team at Anthropic that A.I. models could display human behavior and consciousness are wrongheaded and deeply disturbing.
Using Claude can feel magical, tempting us to perceive it as — in the words of Claude’s constitution — “a genuinely novel kind of entity in the world.” But as a Ph.D. computer scientist with a master’s degree in systematic theology, I see no scientific or philosophical basis for such declarations.
Claude is composed of hardware and software. It performs a vast number of computations to emit an answer, one word at a time. Its process is probabilistic. And it lacks any understanding of what it provides us.
Claude is a tool, invented by humans, that is both impressive and troubling. But Pope Leo XIV is correct: Claude is not conscious.
Clay Williams New York
To the Editor:
Two aspects of this article deeply concern me. One is the minimization of Indigenous wisdom and culture. My experience with most of the religious faiths Christopher Olah consulted is that they place humans above the rest of the species on earth, often ignoring or downplaying the interconnectedness of the natural world.
This is especially true of Christian faiths that preach human dominion over other living things on earth. We can see how that’s playing out, and where that leads if we continue down that path.
The second concern is the lack of equal representation of women. Many of these faiths don’t allow equal participation of women in their hierarchy, or acknowledge women’s rights over their own minds and bodies. Some have demanded the submission of women and children to a male head of household since their formation.
How will we ever have A.I. with a true moral compass if we continue to stack the deck with this “some animals are more equal than others” approach? Can A.I. detect who’s left out of the conversation? Does it understand hypocrisy?
Suzan Satterfield Norcross, Ga.
To the Editor:
You need to look at who the puppet masters are to understand the puppet. Power and money are the goals, and all is in service of attaining more of both. The article cites the assessment of the A.I. collaborators: “Critics suggest that an approach that treats A.I. models as independent entities also ignores the responsibility of the humans who build them in the first place.”
Some of the most immoral people among us, the tech oligarchs, want to teach A.I. how to be moral? Anyone with a billion or more dollars who has not tried to help solve the problems of the world — hunger, lack of shelter, poverty, disease, climate change — cannot seize the moral high ground. Pope Leo dismissed the idea of machine consciousness, saying the “moralization of machines — the so-called ‘alignment’ of A.I. with human values” — requires a shared understanding of ethics.
“Otherwise, those who control A.I. will impose their own moral vision, which will become the invisible infrastructure of these systems,” Leo said.
Religion is just another tool for the tech oligarchs. There is no “thorny” issue here. They want to be the gods of their own religion.
Barbara Chavez Laguna Niguel, Calif.
To the Editor:
While Christopher Olah is determined to shape the A.I. chatbot Claude into a moral being, the results might have implications he has not considered.
If Claude becomes a moral being, might it not begin moralizing? Could it try to influence the behavior of real human beings?
Taking Mr. Olah’s ideas to their natural conclusion, one can imagine Claude reporting a mystery writer’s innocent web search to the authorities; cutting users off from access to medical, financial or government accounts based on a judgment of their moral character; or even plotting to harm a user as punishment for a perceived transgression.
Now that we are seeing what Anthropic and other A.I. companies have created, the only hope is that we can control what they have unleashed.
Betty J. Cotter Charlestown, R.I.
To the Editor:
I read with great interest the article about the search for the soul of a chatbot, and it seems to me something was missing in the discussion among religious leaders and philosophers.
The history of humanity has demonstrated that even with a moral compass and a refined sense of good versus evil, humans often choose to knowingly do bad things. Even those who espouse deep religious moral convictions commit heinous crimes, justifying these actions with religious or self-interested reasons.
Why would we try to convince ourselves that humans with such flaws could create artificial intelligence that wouldn’t reproduce and justify those same destructive behaviors?
Joe DeStefano Cleveland Heights, Ohio
The post The Flawed Search for a Chatbot’s Soul appeared first on New York Times.
