AI CEO Building Platform Based On Human Nature Is Confused By Human Nature

Anthropic’s CEO expressed concern that his employees are more preoccupied with money than the mission of the company.

Every so often when I log in to the bleak online realm that is LinkedIn, I catch a glimpse of job postings that Anthropic is recruiting for. One posting listed a storyteller position with a salary of $400,000 per year, another was an events manager role with a salary range of $200,000 to $400,000.

It’s wild to think about such a salary range for mid-level corporate roles—especially when many people can’t even get an interview with an AI “recruiter” because of how many qualified people are applying for one job, and how skewed the labor market currently is in favor of employers these days.

This is not to say that tech workers should be underpaid, or not paid handsomely. (Anthropic is valued at $965 billion.) But when I read that Dario Amodei, the CEO of Anthropic, was concerned that people were joining his company for the salary and not the mission, it did make me wonder: how is a company that is compiling billions of data points on human nature so confused about human nature?

The tech labor market is not exactly ideal right now. Layoffs abound, because everyone from software engineers to marketing managers are either being “replaced by AI” or AI is being used as the excuse to wipe away those jobs. Even the people conducting the fires are said to be informed by AI’s analyses and suggestions.

In other words, to be gainfully and securely employed by a tech company today is, in many ways, to work for an AI company. Especially for employees who haven’t yet experienced a blockbuster IPO—there’s no denying that Anthropic’s upcoming IPO probably is the golden carrot of generational wealth taunting them to join over anything else.

To Amodei’s point, though, I’ve seen several people across various social feeds announce that they are joining Anthropic for “mission-driven” or for prestige reasons. One woman who sold her online harassment mitigation startup announced recently that she was joining Anthropic. Presumably she doesn’t have to work for money, and is interested in being at the epicenter of the “least evil” AI employer.

It also may be that recruits like her want to be close to one of the most consequential technological revolutions in human history. They may feel excited to influence and build from the inside—to have an “impact” on products that hundreds of millions of people may be using, carrying an ethos that may be adjacent to effective altruism. Perhaps they also see it as a mark of their own superintelligence to be recruited to build superintelligence. (Speaking of human nature, ego is a thing, too!)

Anthropic is also a public benefit corporation—meaning that its governance structure is built to prioritize long-term societal gains over short-term investor gains. So it follows that Amodei wants to hang his hat on attracting the people who believe they have a moral compass. I have to believe that he is drinking his own Kool-Aid, that he and his intelligent band of less-than-evil employees will be better than the alternative.

However: Anthropic, to be competitive in the AI ecosystem, is still beholden to interests that require them to pour an unbelievable amount of money into AI infrastructure. Anthropic, in order to compete in the same game as OpenAI and others, must still participate in the strategies of the game in which everyone else is operating.

For example, although it was a brilliant PR move, we’d be naive to think that Anthropic—which was providing its technology to the U.S. government’s Iran war efforts—didn’t want to steal some users from ChatGPT when it vocally declined to work with the military.

So yes, sure: recruits to Anthropic could rationalize that this is the less evil option, that they are doing this for humanity.

But as a former tech worker myself, I’ve seen people move from one hyped tech product to the next: ridesharing to scooters, crypto to artificial intelligence. And while AI is far more consequential in its macro and micro implications than some of these circle-jerk products like crypto, the AI train may be perceived by some tech workers as the last money-making opportunity to benefit the insiders before even more job losses occur. For someone like Amodei who is compiling one of the largest databases of neural networks in human history, to miss this insight seems like a massive blind spot.

What else we’re paying attention to this week…

Mayor Mamdani is supporting a union-backed bill that would require Amazon to hire delivery workers rather than relying on contractors.

Is it the end of Google search?

A PR firm composed of AI avatars is sending out pitches to journalists. The firm’s clients said they weren’t aware of the practice.

Mark Zuckerberg is at it again with a 6,500 word manifesto making the case for AI superintelligence—and predicting that we’ll all have “AI agents” predicting all of our wants and needs in the near-future. I enjoyed this commentary piece about it entitled Mark Zuckerberg doesn’t understand how to live.

Speaking of Zuck, Meta has pledged to invest $1 billion in communities where it’s developing and operating data centers; meanwhile, they are investing $600 billion in data centers by 2028.