Author Luke Burgis puts forward the question, is typing dead?

Luke had tried a new AI voice dictation app. From a text generation perspective, both language learning models and voice dictation seem more efficient. I did the same a few years ago, which I covered for Forge here. An excerpt:

In dictating this piece, I spoke for a little over 16 minutes and ended up with 2,935 words. The final piece is around 750 words. With the dictation process, there’s always a lot of cutting. That’s okay — somewhere within that massive chunk of thoughts will be some of your best ones.

…

After a dictation, I’ll export the text out of my software program and into a Google Doc. For my first edit, I simply make the writing readable, putting my thoughts into full sentences and creating different sections. I also add research and fill in any blanks. In my second edit, I firm up the structure, tighten up the writing, and strengthen the ideas. The third time around, I triple space it, print it out, and mark it up by hand.

I was familiar with this process because I’ve been in the professional ghostwriting business for over a decade. It’s definitely worth noting the efforts in cutting and editing the text.

My first job at Xtreme Labs involved interviewing team members and putting their thoughts into the written word. I elaborated on this practice with my work at my editorial studio Wonder Shuttle, where my team and I developed a collaborative writing process that we worked on with clients like Shopify, Wealthsimple, and Flipp.

I was mildly surprised to read Luke write:

I spent a few days with the economist Tyler Cowen last week, who told me that he may never truly “write” another book again (and he has written many). This was a wake-up call to me. I think Tyler is at the forefront of the curve on this stuff, but I also thought he discounted a very real aspect of writing: the written word as we write it is not the same thing as the words that we speak. 

Luke hits it on the nose with a hammer. I’d take it a step further; we don’t think to write, we write to think. It’s similar with talking, though the process happens a lot faster, which also means we don’t have as much time to think. It’s like R.P. Blackmur told Robert Caro, “You’re never going to achieve what you want to achieve if you don’t stop thinking with your fingers.”

It goes the same way with talking. The reason I need to spend so much time editing my transcription—which involves a lot of typing, through rewriting—is because it’s often a complete mess. It waxes, it wanes, and it branches off into tangents. Perhaps I’m just more comfortable with a pen and pad, or in front of a keyboard; I’d wager that a lot of people do better, more serious, thinking with the writing in front of them and not just through dictation.

One possibility that does excite me is an equal time typing, though in different parts of the book writing process. If the writing and text generation comes across more easily, then that means an author can spend more time finding facts, discovering patterns, and architecting interesting stories. Imagine an author writing up three different versions of a chapter, in the same amount of time they spent writing one; they have three different combinations to choose from. Then, it’s simply a matter of picking the best one.

Either way, I’m curious about the possibilities. I have yet to read Reid Hoffman’s new book, which was written with GPT-4; I would imagine Tyler Cowen writing something like this, and spending more time on research, interviewing, and synthesizing—all aspects that will probably involve typing.