Eric So, a professor of global economics and behavioral science at MIT Sloan, argues that the convenience of AI tools is creating a dangerous gravitational pull he calls "AI gravity." In his new book, "The Collision: What AI Does to Us," So warns that developers and knowledge workers who use AI to bypass cognitive effort are risking significant cognitive atrophy. The core premise is that while AI offers enormous potential, the instinct to surrender thinking to the machine leads to a degradation of memory, judgment, and the tolerance for friction required to solve complex problems.

The Risk of Cognitive Atrophy

So compares the effects of AI dependence to physical muscle atrophy, noting that the human brain adapts to the work it performs. When professionals routinely outsource reasoning to AI, the brain "doesn't really fire very much" and loses the exercise necessary to maintain sharpness. This phenomenon creates a workforce that is less capable of functioning independently and less comfortable with the normal day-to-day frictions of problem-solving. The efficiency of AI, which can complete tasks that once took hours in moments, is seductive but potentially corrosive to long-term skill retention.

Breaking the Junior Talent Pipeline

For engineering leaders and executives, So identifies a critical strategic error: treating AI adoption solely as a cost-cutting opportunity rather than a talent development challenge. If companies use AI to replace entry-level thinking tasks, they may save money in the short term but destroy the pipeline that produces future senior engineers. So argues that AI should be used to free junior employees from routine work so they can reinvest that time in building judgment and learning from experienced colleagues. Without this deliberate intervention, organizations risk having no one left to run them because no one has developed the necessary senior-level intuition.

Managerial and Educational Shifts

Managers are urged to stop tracking AI usage metrics and start assessing whether employees truly understand the output they generate. So advises leaders to model good AI use by asking teams to explain their reasoning and defend their decisions, rather than just rewarding polished output. In the educational sector, So criticizes moves like New York Mayor Zohran Mamdani’s proposed one-year ban on AI in public schools. He argues that outlawing the technology pushes its use into the shadows, whereas educators should teach students to deploy AI as an amplifier for their own ideas rather than a cognitive replacement.

Key Takeaways

  • AI gravity describes the instinct to surrender cognitive effort to AI, leading to atrophy in memory, judgment, and tolerance for friction.
  • Companies risk destroying their future senior talent pipeline if they use AI to cut costs by eliminating entry-level thinking tasks.
  • Managers should assess whether employees understand AI-generated output rather than just tracking usage metrics or rewarding polished results.
  • Educators and workers should use AI to amplify signature thinking skills, not replace them, to remain marketable and capable.

The Bottom Line

AI is not the enemy, but passive consumption is. Developers must treat AI as a tool to amplify their unique signature thinking skills, not a crutch that erodes their ability to function independently.