Sam Altman Suggests AI Will Eliminate Jobs That Aren’t ‘Real Work’

TL;DR: OpenAI CEO Sam Altman claimed at DevDay that many jobs displaced by AI might not constitute “real work,” invoking anthropologist David Graeber’s controversial “bullshit jobs” thesis. However, research contradicts this sweeping claim: European studies found only 5% of workers consider their jobs useless, whilst US surveys suggest closer to 20%, with pointlessness stemming more from poor management than inherent role invalidity.

Sam Altman’s remarks at OpenAI’s DevDay conference generated immediate controversy when he suggested that many jobs eliminated by large language models might never have been “real work” in the first place.

Context and Background

Responding to a thought experiment about how a farmer from 50 years ago might view contemporary work, Altman stated: “The thing about that farmer… [is that] they very likely would look at what you do and I do and say, ‘that’s not real work.’” He continued: “This makes me feel a little less worried… [but] more worried in some ways. If you’re… farming… you’re doing something people really need. This is real work.”

The statement, clipped and shared widely on social media, drew criticism ranging from callous to dystopian. However, Altman isn’t the first to suggest swathes of modern work constitute performance rather than productivity.

A decade ago, the late anthropologist David Graeber wrote On the Phenomenon of Bullshit Jobs, arguing that many workers secretly believe their positions are pointless. The essay went viral and was later published as a bestselling book in 2018, with Graeber’s core claim—that entire sectors are built on box-ticking bureaucracy with no social value—cited by everyone from disgruntled office workers to policy think tanks.

However, subsequent research hasn’t substantiated these sweeping claims. A 2021 study using the European Social Survey found only about 5% of people said their jobs felt useless. A similar US study put that number closer to 20%. In both cases, researchers concluded that feelings of pointlessness stemmed more from poor management and work culture than inherent role invalidity. If your manager micromanages and workflow is broken, even valuable work can feel fake—but that doesn’t prove the role should be automated out of existence.

Looking Forward

Where Altman’s commentary holds potential validity is in what it implies rather than explicitly states. Most jobs aren’t fake, but many have accumulated layers of automatable busywork: compliance checklists nobody reviews, reports nobody reads, emails summarising meetings that could have been Slack threads. This “game-playing” work represents precisely what LLMs currently excel at automating.

When Altman suggests these models will wipe out tasks rather than entire roles, this distinction matters. The controversy arose from framing job displacement as eliminating “fake work” rather than acknowledging legitimate concerns about economic disruption affecting real people performing work that serves organisational purposes, however inefficiently structured.

The data suggests a more nuanced reality than either extreme position—neither widespread fake employment nor complete immunity from automation. The challenge lies in distinguishing between automating genuinely wasteful bureaucracy and displacing valuable work simply because AI can perform it cheaper, with the two categories frequently conflated in discussions about AI’s labour market impact.

Altman’s positioning reflects a broader pattern in tech leadership rhetoric: framing disruption as efficiency gains whilst minimising legitimate concerns about economic dislocation. Whether AI ultimately proves a productivity catalyst liberating workers from administrative drudgery or primarily a cost-cutting mechanism enabling workforce reduction likely depends less on technology capabilities than on how organisations and policymakers structure the transition.

Source Attribution:

Share this article