Searches by Vauhini Vara

I read Searches by Vauhini Vara. It’s one of my favorite books I’ve read in the past couple years. The book is both factually dense and beautiful, vulnerable. A book about the internet and feelings—that’s my kind of book.

On dissonance, contradiction

What I liked so much about this book is that she pokes at the cognitive dissonance and internal conflicts we can’t avoid in modern life. We have contradicting goals and beliefs and opinions. Like, you want to shop local for groceries but need to save money, and you want to buy books at the cute neighborhood bookstore but Amazon is so darn convenient and affordable, and you don’t want to eat animals but everyone is screaming at you to up your protein, and you walk to work to spare the greenhouse emissions but are flying to six weddings in other states this year. Et cetera.

We couldn’t possibly ever logically defend all of our choices simultaneously. Our heads would explode. She interleaves her conflicting behavior so well. She’d covered a story about an Apple factory with such terrible working conditions that more than a dozen people had died from suicide. She writes:

Jobs said that Apple had been “extraordinarily diligent and extraordinarily transparent” about its suppliers’ working conditions and that Foxconn was “a pretty nice factory,” though he acknowledged the “difficult situation” and said that Apple was trying to understand and address it. I learned all this on my MacBook. I continued to write, on my MacBook, my novel about technological capitalism. That year, after finishing graduate school, I bought my first iPhone. (98-99)

I haven’t seen enough of this sort of writing that acknowledges and tries to work through our dissonance. I’d love to read more like it. With respect to Twitter, for example, because I have a whole lot of dissonance there.

On her chatbot conversation

The ongoing chatbot transcript—a conversation between Vara and ChatGPT or somesuch about the very book that you are reading—is such a good idea. It’s a clever parallel track for us to come to our own conclusions about AI’s role and value. I say our own conclusions because she doesn’t directly tell us how we should feel.

The long-winded chatbot responses to her short questions—and this might be part of the point—are extraordinarily bland compared to Vara’s own writing. They are watered down, boring. I found my mind wandering off when the language model went on and on (but not while reading her side of the conversation). It’s kind of crazy to stick thousands of words into your book that are directly from the mouth of the machine. I am not against her doing it in the way that she’s doing it—no deceit—but I did get bored and skim through those parts.

That struck a nerve because it has been generally bothering me of late: LLMs are so fucking long-winded. When two humans talk, it would be very weird to reply with a thousand words when your conversational counterpart has spoken ten. It’s a social faux pas to dominate conversation. Why then, do the language models blab on?

Another thing I now crave after reading her chapter “Ghosts” where a language model completes drafts of her writing: I’d like to play more with language models that do not have any invisible system prompt or harness whatsoever. I want to see the next token prediction machines in action so that I can start a sentence and see how it’ll end it. I gather that’s what the interface was like when she (Vara) was playing with a beta version of GPT-3, before ChatGPT came out. I think that’d help me, and people in general, understand these models better, more intuitively.

On format

The other reason I loved this book so much is Vara’s creativity with format. I mentioned the interleaved chatbot chapters, but she did a few other things that woahed me. (Aside: I’d like to nominate “woahed” for your consideration as an alternative verb to “wowed” when appropriate.)

One chapter is a stream of her Google search history over 20 years (apparently you can go to some “my ad center” in Google to see a rough draft of what Google knows about you.). Pages and pages of her mundane, vulnerable, and funny searches. Another chapter is a similar stream of her Twitter topics of interests, as labeled by the algorithm.

There is also this bilingual Spanish-English chapter that was pretty startling to see in a printed book. It had two columns, one written in Spanish by Vara herself, who’d recently learned the language while living in Madrid, and the second in English, direct from Google translate with errors intact (both her’s and Google’s). It was a well placed analogue to a separate chapter about AI writing. With both AI writing and Google Translate, our own skills atrophy (or never build at all) in exchange for convenience. But again, Vara isn’t too on the nose about this. She acknowledges her own contradiction: re having AI write for her, she says the idea “disgusted me. It also attracted me.” My read is that she’s telling her personal tale and letting us come to our own conclusions throughout, because this stuff is not straightforward and we all must choose and live with whatever tradeoffs we make.

I’ve been thinking about this quite a lot in the last few years, my own skill atrophy from convenient technology. The other weekend I insisted on turning off my phone for 24 hours on a weekend camping trip, printing out maps ahead of time to drive and hike around Red River Gorge. It was a blast, if impractical. I also learned to drive stick shift recently (after dragging my feet for a while), which I feel good about, I’ll say.

The English-Spanish chapter was astonishingly good, and it’s self-contained enough to standalone as an essay. I’ll admit to reading the English column after the first page even though I could understand 90% of the Spanish because that would have taken me at least twice as long, which itself is a good illustration of how attractive the convenient option can be.

On a fork in the road

I liked reading about the choice Vara had to make between two attractive paths—the prestigious creative writing program at Iowa and the Apple beat at the Wall Street Journal. It’s not often you read (or I read, at least) someone writing about choosing between two good paths, probably for fear that it’ll sound braggy. I don’t think she sounds braggy.