Alex Karp Warns Humanities Majors About AI's Job Impact

Karp’s Warning on Humanities Majors’ Jobs

I keep coming across two kinds of pieces side by side: ones arguing that AI makes the humanities more important than ever, and ones arguing that a humanities degree is becoming worthless. They reach opposite conclusions, yet both are stated with total confidence. Before I take either claim at face value, I wanted to look at what specific jobs and skills each one is actually talking about.

At the Davos World Economic Forum in January 2026, Palantir CEO Alex Karp, speaking with Larry Fink, said AI would hit the job prospects of humanities majors hard. He used himself as an example—someone who studied philosophy at an elite university—and said that without some other technical skill, that kind of academic training would be hard to market.

What’s worth separating here is the value of studying the humanities versus the skills a specific job actually demands. Karp’s comment is really about the second question: a forecast about hiring and economic value.

In a TBPN interview on March 12, he named people with vocational training or neurodivergence¹ as groups holding an advantage. Karp himself went from Haverford College to Stanford Law School, and later earned a philosophy PhD from Goethe University in Germany. That résumé is exactly what makes his comment stand out—but a forecast built on his own credentials and traits isn’t automatically a rule that applies to everyone.

In earlier interviews, Karp has also argued that an elite-school pedigree and high intelligence alone aren’t enough—you need concrete, specific expertise. It’s worth checking this warning against actual hiring data.

What Hiring Statistics Actually Show

What Karp emphasized wasn’t credentials but actual usable skills on the job.

He pointed to technical staff involved in battery production, and someone who trained at a community college, worked as a police officer, and went on to operate the U.S. Army’s Maven system. His point was that people can pick up new skills and take on important work. But this shouldn’t be taken as a guarantee that people with vocational training face no risk of being replaced by AI.

The difficulty of finding a job is itself visible in the data. According to Q4 2025 figures released by the New York Fed, the unemployment rate among U.S. college graduates aged 22-27 was about 5.7%, with an underemployment rate² of 42.5%. Here, underemployment means the share working in jobs that typically don’t require a bachelor’s degree. It doesn’t measure whether people are dissatisfied with those jobs.

These figures aren’t tallied only for humanities majors. We should also be careful about comparing them directly against the entire workforce, across all ages and experience levels, and concluding that a college degree itself puts someone at a disadvantage.

Karp also said that his job is to find what each person is particularly good at and place them where they can focus on that work. I think this involves the ability to understand people and make judgments suited to a given situation. But this kind of ability isn’t produced by a humanities degree alone. It’s built through education, work, and a variety of experiences.

Palantir also launched a Meritocracy Fellowship in 2025 for high school graduates. Those who complete the paid program get a chance to interview for a full-time position. It’s an attempt to cultivate talent in-house while criticizing conventional college education. The existence of this program shows that education paths outside college are possible, but it doesn’t mean it’s already been proven better than a typical college education.

Business Leaders Who Emphasize Communication and Judgment

There are other voices in the same AI industry who emphasize different capabilities. Not all of these remarks were direct rebuttals to Karp, though.

Anthropic co-founder and president Daniela Amodei said in an ABC News interview in February that the study of humanities will become more important. She explained that when hiring, she looks for communication skills, interpersonal skills, kindness, and curiosity. She also spoke positively about her own choice to major in literature at UC Santa Cruz.

An illustration representing the debate over humanities and AI. This is not an actual scene from a real conversation.

McKinsey’s Bob Sternfels also emphasized creative thinking and judgment on the HBR podcast in January. He said the firm is taking a fresh look at talent with liberal arts backgrounds. This is a statement about the company’s hiring direction. It’s neither proof that AI will remain incapable of creative work going forward, nor proof that everyone with a humanities degree excels at such abilities.

We should also distinguish between a company saying it values communication and adaptability as hiring criteria, and a company saying it has actually increased hiring from particular majors. An executive’s interview alone can’t tell us the actual hiring outcomes.

When discussing capabilities, it’s better to break them down into concrete behaviors. Communication skill would mean explaining things so the other person can understand; judgment would mean checking the limits of the data and stating the reasons behind a choice. Only by unpacking things this way does it become clear what should actually be evaluated in education or hiring.

Karp emphasizes that a degree in a given major doesn’t guarantee employment, while Amodei emphasizes the value of the ability to work well with people. The conclusions differ, but neither statement is grounds for having to choose between the humanities and vocational training as an either-or.

You Also Need to Look at the Hiring Environment, Not Just the Major

I think both claims are exaggerated: that every humanities major is now at a disadvantage, and that studying only the humanities is somehow an advantage in the AI era. What matters is both what you learned and how you apply that learning to actual work.

A February 2026 briefing summarizing Goldman Sachs Research looks at a labor market where both hiring and firing have slowed down. It found that employment in the industries with the highest share of college graduates fell by an average of 9,000 people per month between 2023 and 2025, while employment in the industries with the lowest share of college graduates rose by an average of 12,000 people per month. This is net employment change across the industries covered — not a figure representing the job placement probability of any specific major, or the raw number of new hires.

People trying to land a new job can be hit harder by a hiring slowdown than existing employees are. Even if companies aren’t laying many people off, if they open fewer new positions, it becomes harder to get hired. Goldman Sachs judged that, as of that point, there was limited evidence to attribute the deterioration in college-graduate employment to AI alone — though it separately noted that AI’s impact could grow going forward.

Cengage’s 2025 graduate survey points to similar difficulties. Among respondents who graduated in 2025, 30% said they’d landed a full-time job related to their major. But that doesn’t mean the remaining 70% are all unemployed — it includes people working outside their major or in other job formats. Nor should this kind of survey be treated as equivalent to a national unemployment rate.

What these sources actually tell us is that new job-seekers and recent graduates are facing a tough hiring environment. Which majors are affected, by how much, and whether AI is the cause — those questions require breaking the data down further by major, occupation, and industry.

There’s a longer-term trend here too. A 2025 Cleveland Fed study found that the monthly job-finding rate for college graduates aged 22 to 27 has been declining since around 2000, converging toward the rate for young high school graduates. Decades of change can’t be explained by the recent spread of generative AI alone. At the same time, the study notes that college graduates still hold an advantage in wages and job stability once they are employed. The difficulty of landing a first job and the overall value of a degree are two separate things.

How many jobs AI will eliminate or create going forward remains, for now, a matter of forecasting. Optimism or pessimism from business leaders, and expectations from HR managers, are worth referencing — but they aren’t the same as actual employment changes. That’s why I try to look first at a forecast’s time horizon, assumptions, and underlying evidence, rather than reaching for a firm conclusion.

The draft looks accurate and complete against the Korean source; no fixes needed.

Oswarld’s Lens

When I listen to what Karp and Amodei say, I also pay attention to what kind of talent each company wants and what business each is running.

Karp’s criticism of universities fits neatly with Palantir’s own in-house talent development program. Amodei’s talk of communication and understanding people connects to how Anthropic describes its hiring. That said, the existence of this context doesn’t let us conclude that the two men said these things purely for PR purposes. We should understand the background of a claim, but judge whether it’s right or wrong on separate grounds.

In my own experience building GTM (go-to-market) strategy, splitting knowledge from execution was never useful. I needed both the ability to understand what problem a customer was facing and the ability to handle the tools that could solve it. For me, the more important question is where and how humanistic thinking and technical skill can actually be applied in a given job.

The case of the former police officer Karp mentioned can also be read as evidence that a degree title alone can’t tell you much about someone’s aptitude. What matters is looking at both a person’s existing experience and their capacity to pick up new skills together. There’s no reason that capacity has to come from any one particular major.

If I were evaluating someone who uses AI tools, I wouldn’t just look at how fast they produced a result. I’d also want to see whether they could spot missing information or false premises, and explain what impact the output would have on a client. The ability to use a tool well and the ability to critically read its output can be trained together.

In a January interview, Sternfels explained that McKinsey is now working with 40,000 people alongside 20,000 AI agents. He expects every employee to be supported by at least one agent within the next 18 months. That agent count doesn’t mean the same number of jobs have been replaced. In an environment like this, I think what matters is checking what the AI has produced and adjusting it to fit the actual work at hand.

I think what’s needed is an attitude that lets you study one field deeply while still picking up other tools and perspectives. That’s not a call to abandon what you’re already good at. It’s a call to keep looking for ways to apply that skill to new problems.

Closing

I wanted to write this piece because I kept noticing how easily people accept phrases like “the humanities’ moment” or “technology is making traditional education obsolete.” People love to say critical thinking matters, but then they don’t actually ask what evidence backs up those very claims.

I’d suggest reading this kind of argument in three separate layers:

• The observed change: Check which age groups, majors, or occupations actually saw shifts in hiring or wages.
• The explanation for the cause: Look for evidence on whether AI, the broader economy, or hiring policy is what’s actually driving it.
• The advice for individuals: Ask whether that explanation even applies to your own situation.

When you’re weighing whether to study the humanities or pick up a technical skill, it helps to start from the problem you actually want to solve, and work backward to what kind of learning that requires. A famous CEO’s one-liner can’t make that decision for you.

References & Further Reading

Primary sources

• Fortune, Alex Karp’s remarks on humanities and vocational training, 4.11.2026: covers his January Davos comments and a March interview.
• Fortune, Daniela Amodei’s remarks on humanities majors and hiring, 2.7.2026: reporting on an ABC interview.
• HBR IdeaCast, Where McKinsey—and Consulting—Go From Here, 1.6.2026: full interview with Bob Sternfels.
• Goldman Sachs Research, Why Is the College Graduate Unemployment Rate Higher Than Usual?, 2.18.2026: analysis of slowing hiring and firing, shifts in net employment by industry, and AI’s role.

Background

• Federal Reserve Bank of New York, The Labor Market for Recent College Graduates: US labor market data for college graduates aged 22-27. The figures cited in this piece are from the Q4 2025 release.
• Federal Reserve Bank of Cleveland, Are Young College Graduates Losing Their Edge in the Job Market?, 11.2025: explains long-term shifts in job-search success rates alongside differences in post-hire job stability and wages.
• Cengage Group, announcement of the 2025 Graduate Employability Report, 9.9.2025: graduate survey data and employment outcomes by field of study.
• Palantir, Meritocracy Fellowship: overview of the company’s own talent-development program.

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

Footnotes

1. Neurodivergence: the view that people’s neurological traits and ways of thinking vary widely. As Karp uses the term “neurodivergent,” it refers to people whose neurological development differs from the typical pattern — including those on the autism spectrum, with ADHD, or with dyslexia. This trait alone cannot predict a person’s job performance or career prospects. ↩

2. Underemployment: in the New York Fed data cited in this piece, this refers to the share of college graduates working in jobs that don’t typically require a bachelor’s degree. It should be distinguished from other underemployment statistics that measure insufficient working hours. ↩