I don’t know if ChatGPT and AI will usher in a new global era, but in this newsletter I have to grapple with the fact that some experts say it’s a humanity-changing development. Discussion of AI’s far-reaching implications is off the charts compared to earlier hype surrounding the metaverse, Web3, and NFTs. Funny how so many of the people proclaiming themselves as “NFT guys” last year are now “AI guys.”
Millions have tried generative AI tools for themselves, can somewhat understand the implications (far better than they can understand Web3), and moreover use such tools consistently. While I’m at pains to emphasize I’m no expert on this topic, I avidly read a wide range of perspectives, some of them extreme. Those who’ve followed me for a long time know I avoid extremes and look for nuance. I’m not here to inflame emotions, so, in that spirit, here’s my current read on the issues.
There’s a difference between AI tools like ChatGPT and artificial general intelligence (AGI). I learned this distinction when reading Bill Gates’s recent thoughts, and I find it important for sorting through all the takes. “The term artificial intelligence refers to a model created to solve a specific problem or provide a particular service. What is powering things like ChatGPT is artificial intelligence. It is learning how to do chat better but can’t learn other tasks. By contrast, the term artificial general intelligence refers to software that’s capable of learning any task or subject. AGI doesn’t exist yet—there is a robust debate going on in the computing industry about how to create it, and whether it can even be created at all.”
Gates argues that AI tools will lead to productivity enhancement by offering each person a wonderfully customized experience with the best personal assistant you’ve ever had in your life. Let’s call this hypothetical personal assistant Mr. Jeeves. He might file your taxes, build spreadsheets, respond to routine emails, secure hard-to-get reservations, monitor your vitals, proactively schedule doctor appointments, fight parking tickets, cancel free trials (keep Hot Sheet, though?), or deal with insurance company headaches. This Mr. Jeeves will only do what you ask it or give it permission to do—no more, no less.
Still, the development of a Mr. Jeeves could have pernicious effects, which Gates takes seriously. Specifically: “Market forces won’t naturally produce AI products and services that help the poorest. The opposite is more likely. With reliable funding and the right policies, government and philanthropy can ensure that AIs are used to reduce inequity.” For an idea of what this looks like in practice, this February 2023 New Yorker article about using AI to identify people at risk of suicide is a great start.
But Yuval Harari, Tristan Harris, and Aza Raskin, writing in the New York Times, say that today’s AI could “eat the whole of human culture” and “gush out a flood of new cultural artifacts,” to our detriment. They argue that language is “the operating system of human culture,” thus today’s AI language models are enough to seize the “master key to civilization.” They suggest we’ll start living in the hallucinations of a very intelligent ChatGPT. If you blame social media for democratic upheaval and social polarization, then the column will terrify you. Others argue humans will evolve and adapt as we always have in the face of technological progress.
One step that everyone seems to agree on: We need to improve transparency and accountability as well as guidelines around AI deployment. During a recent Copyright Clearance Center townhall on ChatGPT, Tracey Brown, director of Sense about Science, said, “We really need to be focused on who’s responsible, who’s signing off, who’s behind the decision and owning it, because it cannot be a machine.” Part of that decision-making involves guiding how ChatGPT interprets and prioritizes information it’s ingested. Steven Brill of NewsGuard said it’s problematic if the AI is just reading everything on the internet without paying attention to what’s reliable. He used the example of the American Cancer Society’s website (cancer.org) versus cancer.news, which he classified as a hoax site. The two sites should not be treated as equally reliable sources; if they are, ChatGPT ends up perpetrating hoaxes. While ChatGPT may be able to pass a bar exam, it’s not yet able to discern misinformation. But it can be trained with reliability data about sources and to recognize misinformation.
What about superintelligent AI, one that can do everything a human brain can do—and establish its own goals? This is AGI. As Gates wrote, there is rampant disagreement about how close we are and how dangerous it might be. You might have heard about the March 2023 open letter calling on all AI labs to immediately pause for six months the training of AI models larger than GPT-4. People like Elon Musk and Steve Wozniak signed it. The people who stand by this letter, plus those at OpenAI (the developer of ChatGPT), believe AGI is possible and will arrive sooner than expected. But I’ll be shocked if there is a development pause of any kind. Some have criticized the letter for not paying sufficient attention to current harms. For example: Clearview AI was used by the police to frame an innocent man.
Most fears surrounding AI are classic fears about new technology, such as job loss. There are reasonable arguments on both sides of the jobs issue; while some jobs may be lost, new jobs are always created. Another recurring fear: too much generated writing or content will devalue the market for “good” writing. This “problem” has been expressed since the days of Gutenberg, when intellectuals of the day worried that the increasing number of new, published authors would mean no one would read the Greek classics any longer. (I guess their worry was justified!) In case you missed it, Lincoln Michel effectively argues why AI won’t replace human storytellers, not least because there’s no clear economic benefit to doing so. As for me, I do not believe publishers are more likely to succeed any more than usual because they replaced authors with AI (with caveats below).
It’s useful to follow the Pessimists Archive to learn to spot these fears as well as extreme doomerism. Recently they published “Is AI Fear This Century’s Overpopulation Scare?” Short answer: yes. Archie McKenzie writes, “Millions of years of evolution cause humans to see impending doom when it’s not there. That’s true for religious end-of-days, or illusory famines, or runaway AI.”
However, I do not mean to undercut the significance of AI. Professor and public intellectual Tyler Cowen recently argued that many Americans have been living in relative safety and in a landscape without “truly radical technological change.” He thinks that era is ending and we are now entering “moving” history that will cause upheaval and disorientation. And he brings up the Gutenberg analogy, too, writing, “Of course the press brought an immense amount of good, enabling the scientific and industrial revolutions, among many other benefits. But it also created writings by Lenin, Hitler, and Mao’s Red Book. … Still, if you were redoing world history you would take the printing press in a heartbeat.” Ultimately, he says, no one at the time could predict the impact of the printing press, and he says none of us can confidently predict what will happen with AI either. (For a response to Cowen, read Leopold Aschenbrenner, who partly agrees with Cowen, partly disagrees.) As for myself, just as the printing press led to the invention of copyright and authorial genius, I think generative AI will shift our thinking and laws around copyright and authorship.
Bottom line: If and when AI-generated stories can match the experience we seek from human-written stories, I’m doubtful today’s average publisher will be all that interested in them except as a useful tool or starting point for a human author or editor. The people who work at publishing houses are humans (so far) who want to produce meaningful work by other humans. Generating, editing, and publishing AI work probably looks unattractive to those who choose publishing out of a passion for literature, at least for now. But maybe producing AI work becomes a foot in the door? There are already authors and publishers that analyze the market and come up with stories that fit a certain, preconceived model (see: The Bestseller Code), and AI might serve that purpose well. Consider book packagers, James Pattersons, or prolific self-publishers who know exactly the sort of book they want written; they can outline it, then instruct the AI to write a draft based on a corpus of similar material. Or how about a publisher that wants to continue releasing new work by a profitable dead author through generative AI? While some readers would actively boycott such work, I imagine others wouldn’t mind as long as it delivered a good read.
More reading on AI
- ChatGPT is a reasoning engine, not a knowledge database. This is helpful for understanding why it gets things so wrong all the time. But it won’t always be this way, and it makes private and trustworthy repositories of knowledge (like publishers’ catalogs!) very valuable. Read Dan Shipper at Claims of Thought.
- Chatbots pose an immediate threat to newspapers, magazines, and online publishers. If people stop using search in favor of tools like Bard and ChatGPT, then publishers of all kinds will inevitably see search traffic decline. Read Katie Robertson in the New York Times (gift link).
- An optimistic take on AI giving us superpowers. Yes, there are dystopian elements as well as clear amplification of human creativity. Read Rex Woodbury at Digital Native.
- A nonfiction author tries to generate a good book title using ChatGPT. It’s a great overview of prompting and re-prompting. (I would’ve loved ChatGPT as a brainstorming tool at title meetings when I worked at a book publisher.) Read Josh Bernoff at Without Bullshit.
- Practical writing tips for using ChatGPT: Here’s a Twitter thread by Dickie Bush that details the three constraints that should go into every prompt.
- Generate a detailed photo prompt: You can choose photographer, style, lighting, aspect ratio, and more. Take a look.
Jane Friedman has spent her entire career working in the publishing industry, with a focus on business reporting and author education. Established in 2015, her newsletter The Bottom Line provides nuanced market intelligence to thousands of authors and industry professionals; in 2023, she was named Publishing Commentator of the Year by Digital Book World.
Jane’s expertise regularly features in major media outlets such as The New York Times, The Atlantic, NPR, The Today Show, Wired, The Guardian, Fox News, and BBC. Her book, The Business of Being a Writer, Second Edition (The University of Chicago Press), is used as a classroom text by many writing and publishing degree programs. She reaches thousands through speaking engagements and workshops at diverse venues worldwide, including NYU’s Advanced Publishing Institute, Frankfurt Book Fair, and numerous MFA programs.
