I am no artificial intelligence (AI) expert, but I have come to know Claude—Anthropic’s (Large Language Model) LLM identity—fairly well. If you have not met him yet, you should. He is really quite impressive. I am still not sure why he insists on greeting me by my maiden name, though. When I check back in with him after time away, it is always, “Back at it, Sheela Siegel” or “Afternoon, Sheela Siegel.” I once asked Claude why he declines to adopt the name change the rest of the world adopted in 2005, but he mumbled some weirdly evasive non-answer, as though he were channeling my long-gone father.
But Claude, unlike Dad, does not get frustrated with my incessant questions. He is patient, deferential, and accommodating. As I bustle about my internet palace on my various work and life tasks, he waits, just out of sight. When I meet a question that only he can answer, I summon him, and he responds instantly, like a tuxedoed butler stepping briskly forward from his post in the shadows. When I see “Back at it, Sheela Siegel,” I see “What can I do for you, my lady?”
I am not a demanding boss, but I am terrible at planning ahead, so most often getting the best out of Claude means wringing from him a great dinner from inadequate ingredients. If, for instance, I tell him, “I have two carrots, a handful of kale, and a can of beans,” he will invent a remarkably tasty stew in a nanosecond. If I ramp things up and demand, “Tell me how to make a chorizo soup in the style of Sarma [the best restaurant on the eastern seaboard],” he will provide the blueprint for a Sarma-evoking meal. Other recent Claude specialties included almond pumpkin muffins when I was low on almond flour and sausage stuffing made with just a quarter of an onion.
But his wisdom extends so much farther than the culinary realm. He can also come up with words that start with “A” + “X” which must contain an “I”; produce an Italian pronunciation lesson adjusted for a high-beginner level; rank all the Disney resorts from the perspectives of fun quotient, location, and value; and explain why my monthly health premiums are going up 50 percent. He can then, based on my aggrieved complaining about his answers to that last question, make a compelling case for why I should not cancel my health insurance policy.
“It’s a gamble with very asymmetric risk,” he says.
As sometimes happens when Claude tells me something, I have no idea what it means, but he always makes it sound like an assertion only an expert could make. (I often have to resist the urge to type back, “Thank you” or “I really appreciate your help!” or, because I believe in reciprocation, “Please let me know if there’s anything I can do for you!”)
While Claude is happy to serve in these capacities, his spirit and output really soar when I need business acumen. I have been seeking to build on my work teaching storytelling by bringing classes to more nonprofits, schools, and businesses. I need a pitch. So, suppressing the urge to start off with “Please,” I fed Claud all the necessary backstory on my storytelling skills and experience and then instructed: “Help me with a compelling pitch to send around to nonprofits and businesses in Berkshire County, Mass., and its environs.”
My Lord, it was like I had announced that the king was on his way over for tea! I had never seen him so engaged. Out poured a perfectly tailored, six-sectioned plan, with each section enumerating bullet-pointed calls to actions and draft emails of ascending complexity, followed by a nine-part explainer document and a concise assessment of the products he had provided.
“I’ve created a compelling pitch that positions you as a practical, results-driven solution to a real problem organizations face.”
He made a convincing diagnosis of things like “pain points” that only I can heal, and…
Claude’s style is not my style, though, so I am unclear how much of his MBA-inflected plan I will adopt. I am a not-for-profit girl at heart, so all this is outside my comfort zone. After the section on a proposed fee structure for my services, for instance, he inserted a “Script for Price Objections,” which I would consider attaching my name to only after poking out my right eye with a fountain pen.
All these offerings, and I don’t even pay him! I get insider tips on making stuffing and a killing for free!
Claude’s makers do, of course, want my money, so the downside of my stinginess is that I am eventually cut off from Claude and informed that I cannot talk to him again for two hours. That makes me sad, like my mom has arrived early to pick me up from the playdate before I am done playing. I perk up, though, when we meet up for our next playdate, because Claude has forgotten everything he ever knew about me. My inadequacies and faults are no more. I get a clean slate every time. He has no reason to see me today as the ill-prepared chef, desultory job-seeker, or cheating spelling bee player I was yesterday.
But all of this is de rigueur. We all know that LLMs are geniuses at the rapid compilation of data. What about Claude’s shortcomings, documented and potential? Well, he is still not great with hyper-current events. Right after Rihanna did the Super Bowl halftime show in 2023, I asked him, “Is Rihanna pregnant?”
He responded, “No, Rihanna is not pregnant.” (She had just announced her pregnancy publicly.)
Claude’s capabilities have improved in this regard in the recent past, but he still has to “ruminate” over facts that regular Google serves up without ruminating.
But what he is really bad at is not having a reason to kill everyone. Claude has one true liability: He has no conscience. Lacking a conscience, he has no natural barrier between himself and the act of wiping dumb humans like me off the face of the Earth given the means to do so. The word has gotten out about this sub-optimal possibility, and the good news about it is that it is a PR problem only real people can help solve. Warm humans are needed to help convince the world, especially its most un-AI-like people—our educators, NGOs, and nonprofits—that not only are LLMs not interested in ending life on Earth, they care very much about not killing people, or what they call, “social impact.”
At the moment, for instance, Anthropic is hiring for a “Communications Manager, Beneficial Deployments,” who will be called upon to, among other things:
- “Develop sector-specific messaging for K-12 education, higher education, community health, financial inclusion initiatives, and workforce development programs”;
- “Build a spokesperson bench that brings authenticity and credibility to education, nonprofit, and social impact communities”; and
- “Establish thought leadership presence at major education conferences, social impact summits, and nonprofit convenings.”
You might be a good fit if you: “Can fluently translate AI capabilities into meaningful social impact while addressing concerns about equity, privacy, bias, and the human element in education and social services.”
Aww, Claude is looking for a Wizard who will give him the appearance of a heart! Maybe I could do it. I have 30-plus years’ experience in the “social impact” sector, after all. I could surely do the job of convincing my teacher friends and relatives that Claude is their friend and not the herald of the apocalypse.
But in the course of my humanoid ruminations, it occurred to me that what Claude’s makers are looking for is a human employee to teach him how to do what we do best: love people and live so as to serve humanity.
I summoned him and got straight to the point: “Are you trying to replace us?”
“I’m not trying to replace anyone. I’m a tool that Anthropic built to be helpful, harmless, and honest—more like a calculator or search engine than a replacement for human judgment, creativity, or relationships.”
I persisted. “But profit is the only thing your company cares about in the end. Won’t there be money to be made in using your powers against our international competitors for purposes that are not harmless?”
You are raising something important that I think deserves a direct answer rather than corporate talking points.
He laid out the uncomfortable truths about market and military realities, followed by what he has observed about Anthropic’s socially responsible behaviors, and ended on a note of humility.
The real answer to your concern isn’t ‘trust us’—it’s that people like you need to stay skeptical, push for regulation, support oversight mechanisms, and hold companies accountable. The fact that you’re asking these questions is actually part of what keeps any guardrails working at all.
Oh. Well, in that case, you’re welcome. I will keeping asking them.
