Skip to content

Blog /Teamwork /

Outsourced Mind

Outsourcing your writing to AI is brilliant -- until it isn't.

May 12th, 2023

by Andy Didorosi

in Teamwork

People are already rushing to let ChatGPT do their work and the main task they’re subbing out is writing. Paul Graham, the guy behind Y Combinator and lots of early internet, says this is going to be a huge problem.

Observation suggests that people are switching to using ChatGPT to write things for them with almost indecent haste. Most people hate to write as much as they hate math. Way more than admit it. Within a year the median piece of writing could be by AI. - Paul Graham

As a professional writer, and therefore completely unbiased on this particular subject, allow me to share my own take: you should do your own writing now and forever because writing and thinking are synonymous. And you need to think.

Outsourcing your writing to ChatGPT or other Large Language Models (LLMs) seems mildly genius at first: it knows everything! This once-excruciating task of coming up with some “bad words,” editing them down into “okay words,” and organizing them finally into “pretty good paragraphs” can now be instantly outsourced to this thing on the web I paid $0 for – terrific! Moreover, my boss will never know! Now I get all this time back to watch “Love is Blind” and feel guilty about the wilting bag of mixed greens in my fridge. Life is great.

This may hold up for some short amount of time before falling over. If your writing is mostly the “gray goo” we call “content” for social media and company blog posts then generating whatever via AI will work for a little bit. Some people just have to hit X number of blog posts or LinkedIn articles a month and - viola! - their role as a “writer” is fulfilled regardless of the actual outcome. Blame the algorithm or something.

Or you could be writing investor updates, internal process documents, or a eulogy for your mother-in-law. The Monkey’s Paw of AI gives you everything you want with no downside whatsoever. Right?

Slowly though, over time, there’s less and less of you in this process. The AI is performing in the background within a little proverbial black box that tells you nothing of how it came up with these pithy conclusions. You’re not party to any of the “thinking” that happens inside of it so you don’t understand the path it took to get there. Just keep nodding as it figures everything out.

The very process of writing as a normal human is mildly painful because there is uncertainty along the way: this very post was written first in a jangled, haphazard way, then smoothed and rearranged as the thoughts became more clear through the very process of writing it. You’re not sure of the point even as it’s coming out of your hands, like the proverb of building the airplane on the way down after jumping off the cliff.

Almost nobody sits down to write with it all figured out; it’s just a process that unfolds as you type. Sorry.

The issue goes so much deeper. What we’re now calling “AI” using Large Language Models isn’t so much “thinking” as it is “sexy summarization of a shitton of stuff already out there.” The power in these models is their ability to consume unbelievable quantities of information and give you back a reasonably polished tiny slice of it. There’s no “new thinking” going on. Perhaps this doesn’t matter to you as the Junior Assistant to the Deputy Content Manager at Zazzle who just has to get a blog post out before going to the club, but with a few million reps of not coming up with anything new ourselves we’re just going to get dumber – both as individuals now and as a species over time.

If the computer is a bicycle for the mind, large language models are the Segway. Sure, it’ll get you there, but it’s better for you to walk or pedal yourself.

Writing is thinking. There’s no two ways about it. Thinking through concepts yourself and having your own point of view based on the underlying information available to you is a critical part of being alive. Welcome to the show.

Footnote: It’s not lost on me that I used Dall-E 2 to generate the top image for this article as seen on the web version. While the writer gets all huffy about using LLMs to outsource writing, the visual artists could/should be cheesed for the same reasons — but that’s another post.

ON THE INTERNETS

How much is a smidgen? https://www.laphamsquarterly.org/roundtable/how-much-smidgen

The Boring Report - Using AI to de-sensationalize the news: https://www.boringreport.org/

The history of fire escapes: https://noidea.dog/#/fires/

One day you’ll find yourself: https://www.onedayyoullfindyourself.com/

Organizing an electronics bench: https://badar.tech/2023/04/30/electronics-lab-bench-setup-guide/

When the chips are down: https://www.scopeofwork.net/when-the-chips-are-down/

More in Teamwork

Subscribe to The Steady Beat

A weekly-ish round-up of hand-picked articles and resources for people who make software products: designers, engineers, product managers, and organizational leaders.

Subscribe now