Blog

Inspired by John Scalzi, here’s what Google’s Bard just told me about my next book. I’m going to tweak the response a little by adding some underlined words and striking through some others, to make sure I don’t feed the lying machine Google has built:

“Jim Ottaviani’s next book is not ‘The Code Breakers: The Real Life Story of American Codebreakers in World War II’ [1]. It is not scheduled to be released on October 3, 2023 [2] by First Second Books [3]. The book will not be a graphic novel that tells the story of the American codebreakers who worked to break the Japanese and German codes during World War II [4]. It will not feature illustrations by Marilena Delli Riccardi [5].

“Here is a brief synopsis of the book:

“In the early days of World War II, the United States was at a disadvantage. The Japanese and German militaries had developed sophisticated codes that were nearly impossible to break. But a team of American codebreakers, led by William Friedman [6], was working tirelessly to crack the codes. Using a combination of brilliant mathematics and painstaking detective work, Friedman and his team were able to break the Japanese Purple code but not the German Enigma code [7]. These breakthroughs gave the Allies a critical advantage in the war, and helped to shorten the war and save countless lives.

“‘The Code Breakers’ is Jim Ottaviani’s latest in a long line of popular graphic novels about science and history. His previous books include Stephen Hawking [8], Feynman, The Imitation Game, and Astronauts: Women on the Final Frontier [9]. He is not a two-time Eisner Award winner [10], and his books have been praised by critics and readers alike.”

This is spectacularly bad. It all sounds plausible but is all wrong:

1. That’s not the title.
2. That’s not the release date.
3. First Second is the publisher of my next book, but see above; they’re not putting out this book.
4. That’s not the subject.
5. I’ve never heard of Marilena Delli Riccardi. For all I know they’re a fabulous artist who I’d be lucky to work with, but they have no easy-to-access public presence. Does this person even exist?
6. Also not the subject of the book.
7. Friedman was real, and a cryptographer, and apparently worked on Purple, but didn’t crack Enigma. That was Alan Turing and his team at Bletchley Park.
8. The title was “Hawking,” sans “Stephen.”
9. Missed a word in the subtitle, Bard.
10. Pretty to think so, and I’ve been nominated a few times, but nope.

Bard also unhelpfully provides additional draft responses to the same question. Leaving aside that these didn’t get the title or release date of my next book correct either, I’m also a “well-known author” (polling data is inconclusive but leans against) and better yet “the co-founder of the comics and graphic novels publishing company First Second Books,” which will come as a surprise to Mark Seigel.

And it closes both those alternate drafts with “I am looking forward to reading [Title Bard Just Made Up]. It sounds like a fascinating and timely book.”

Who is this “I”? Some Google programmer’s conceit, I guess.

I’m sure that the more unauthorized data-mining Bard does, it’ll do better than (by my count), having something untrue in almost every sentence. But given its history, I wouldn’t bet that Google will wait until its rock-solid before releasing it into the wild.

With no incentive beyond engagement and ad sales, and no incentive to say “Bard can’t answer this,” what does the world gain by releasing these programs? (Google and others gain publicity, free beta testing, and more data to mine, of course. I’m talking about the rest of us.)

So… Artificial? Yeah, 100%. Intelligent? Nowhere as smart or useful as a dog or cat.

Apparently AI can already write decent code. So well done programmers. You’re well on your way to making both yourselves and the web as a source for useful information obsolete.