At an anti-AI protest in London, a flyer from activist group Pause AI read: “Step 1: Grow a digital super mind. Step 2: ? Step 3: ?” It was a deliberate riff on South Park’s underpants gnomes—Phase 1: Collect underpants, Phase 2: ?, Phase 3: Profit—and it perfectly captures the state of AI today. Companies have built the technology (Step 1) and promised transformation (Step 3). How they get there remains a giant question mark.
Two Camps, Same Blind Spot
Pause AI argues Step 2 must involve regulation—though what kind and who enforces it is debated. AI boosters, convinced Step 3 is salvation, glaze over the middle. They see “economically transformative technology” (OpenAI chief scientist Jakub Pachocki’s phrase) racing toward sunny uplands, but the path is hazy, everyone’s taking a different route, and it’s unclear if anyone will actually arrive.
Conflicting Studies, Same Problem
Consider two recent studies. Anthropic predicted which jobs LLMs will most affect—managers, architects, media professionals should prepare; groundskeepers, construction, hospitality less so. But these are guesses based on what LLMs seem good at, not how they perform in actual workplaces.
Meanwhile, Mercor (an AI hiring startup) tested agents powered by top models from OpenAI, Anthropic, and Google DeepMind on 480 real workplace tasks from banking, consulting, and law. Every agent failed to complete most duties.
Why the Disagreement?
- Who’s claiming what (and why): Anthropic has skin in the game. Most “big thing coming” claims rest on coding tool speed—but not all tasks are hackable with code. LLMs struggle with strategic judgment.
- Real-world deployment isn’t a cleanroom: Tools must work in environments contaminated with people and legacy workflows. Sometimes adding AI makes things worse. Refashioning workflows around AI takes time and guts.
The Information Vacuum
That hole where Step 2 should be creates an information vacuum filled by the latest wild claim—evidence be damned. We’re so unmoored from real understanding of what’s coming and how it’ll deploy that a single social media post can shake markets.
We need fewer guesses and more evidence—requiring transparency from model makers, coordination between researchers and businesses, and new evaluation methods that reveal what actually happens when AI rolls out in the real world.
The tech industry (and the world economy) rests on the promise that AI will be transformative. But that’s not yet a sure bet. Next time you hear bold claims about the future, remember: most businesses are still figuring out what to do with their underpants.
Source: MIT Technology Review (originally in The Algorithm newsletter)
