Back when I was a freshman at Stanford, one of the greatest challenges we faced was procrastination. Whenever our papers or problem sets were due, there was always a more tempting alternative activity, like a game of volleyball, a late night run to Taco Bell, or simply hanging out with friends.
One of the great masters of procrastination in our dorm (Alondra) was our Resident Assistant Eric Lassen, who founded Procrastaholics Anonymous as a way for all of us to fight our desire to goof off. The slogan on our t-shirts was “Carpe Niem” with the image of a student sleeping on a beanbag with a textbook covering his/her face. (Yes, I know it should be Carpe Noctem, but how many people know Latin well enough to protest!)
I’m not sure how successful Procrastaholics Anonymous ended up being, but I think we may need to revive the organization because of AI’s procrastination problem.
It’s not that LLMs themselves procrastinate (though anyone who has run into token limits knows that there is little functional difference between running out of tokens and procrastination!). The problem is that a) AI makes procrastination incredibly easy and b) AI allows us to fool ourselves into thinking that our procrastination is productivity.
One of the great delights I’ve found in using AI is the remarkable ease with which I can start new projects. Productivity seems like it’s just a quick prompt away.
But I’ve noticed something concerning. I’m starting a lot of projects, but I’m not finishing most of them. AI lets me get enough work done to feel good about my progress, but the number of such “90% done” projects that pile up, waiting for my final review and edits, just keeps increasing.
When I was a freshman in the 1990s, we had no AI. Writing a first draft represented 90% of the work; making revisions and final edits was a welcome final step, and seemed easy in comparison.
Today, however, using AI to write a first draft represents 10% of the work we need to do to complete a project. Making revisions and edits is now the bulk of the work. Yet the way that I (and likely most of you) use AI doesn’t reflect this new reality. We happily leverage AI start project after project, feeling like we’re 90% of the way to our goals, without realizing that we’re only 10% of the way there.
What’s particularly insidious is that AI really is boosting productivity! Creating those first drafts was hard; that’s why we procrastinated so much to begin with. But now that creating those first drafts is easy, the bottleneck has shifted. We procrastinate by starting new projects, leaving our older projects forever undone. This new form of procrastination isn’t AI’s fault. It’s ours.
Once you acknowledge this new form of procrastination, the solution is simple (though easier said than done). Block out time on your calendar to devote solely to the non-AI work of finishing what AI started for you. Maintain the discipline of minimizing work-in-progress and thus maximizing final output, rather than maximizing the pleasurable but ultimately empty feeling of using AI create your first drafts.
P.S. The irony of course is that the very act of writing this post (without using any AI I might add) is itself a form of procrastination. Not only is using AI a form of procrastination, so is creating content about AI. Just like creating “Procrastaholics Anonymous” t-shirts, while a sincere attempt to fight procrastination, was also a form of the very thing it opposed! I can only comfort myself the same way Eric Lassen likely did–by hoping that the amount my content helps others outweighs the amount that I used it to procrastinate and avoid my own work.