The Alex Caruso Principle
Baseball statisticians measure a player’s value via a stat called “Wins Above Replacement” or “WAR.” A player’s WAR equals the number of additional games his team won for having him on the roster rather than a no-name guy from the minor leagues or free agent market. Looking at the 2022 World Series Champion Houston Astros, we see stars like Jose Altuve (5.1 WAR) and Justin Verlander (5.7 WAR) at the top of the WAR charts. Still, you can’t win a World Series with two players. Verlander represents the clear ace of that roster, though he only pitched 175 innings. Teammates include Jose Urguidy, who managed 164 innings, and Luis Garcia, who clocked in at 157. Those two guys were worth 1.2 and 1.3 WAR, respectively. Decent numbers, but no one child will dream of becoming the next Urguidy. Meanwhile, the trio of Phil Maton, Hector Naris, and Jake Odorizzi combined for 191 innings and 0.3 WAR. In other words, these guys are the “Replacement” in “Wins Above Replacement.” Those 191 innings amount to 34 more than the amount thrown by Verlander. For every inning thrown by the future Hall of Famer, 1.2 were thrown by the Diet Dr. Shasta equivalent of a ball player.
Yet, when the confetti fell, Maton, Naris, Odorizzi, Garcia, and Urguidy each earned the same number of World Series rings that Verlander did. Greatness wins titles. Mediocrity does too.
Basketball guard Alex Caruso gained minor stardom by making his way from the G-League (basketball’s minor league) into the rotation of an NBA championship roster. On his rise, Caruso said this:
A big reason guys get stuck in the G-League is because they don't realize the position they're trying out for. It's like going to a job interview thinking you're going to be the CFO of the company and they're looking for someone to clean the bathrooms.
The 2020 Los Angeles Lakers didn’t need the next LeBron James. They had the current one. They just needed someone who could defend the opponents’ wings and guards, and Caruso fit that role. Although the stars get the most attention, anyone who follows professional sports knows that titles are won and lost by backups, benchwarmers, and guys on minimum-salary contracts. This also applies, of course, to my dear Golden Knights. Our Stanley Cup victory included an overtime goal from Brett Howden, a player who advanced stats once ranked as one of the worst players in the entire NHL!
Uh… Isn’t this an Article about AI?
Freddie deBoer recently published an article that I’ll use to represent the AI-critical perspective. AI proponents dream of future products that revolutionize our world. What we have in front of us, he argues, remains mediocre. However, deBoer shouldn’t forget that he comes from an unusual line of work. When it comes to writing, only the top 1% (if not less) can hope to make a living from it. I guarantee this blog beats 75% of the Substacks out there, yet I don’t even dream about making a cent from it. In the rest of the professional world, though? C+ work gets it done.
In my early days as a data analyst, I misunderstood my value proposition. I spent too much time trying to optimize complex solutions via the causal modeling I learned in grad school or the fancy machine-learning stuff I learned in my free time. I thought that I could impress my colleagues and superiors by producing A or A-minus work. I couldn’t. I looked clueless and naïve. I soon figured out that companies don’t look for analysts who find the perfect solution to the businesses’ most pressing issues. Such analysts either don’t exist or demand much higher salaries. Rather, companies need people who can kinda help with a bunch of different problems. I learned to produce C+ reports and C+ dashboards, hold C+ meetings with C+ agendas, offer C+ solutions via C+ presentations, and maintain C+ relationships with a bunch of different teams. As a result, I receive tons of praise for “getting a ton of stuff done.” Even if they hired some super-genus who could perfect all those items, not much would change. There’s just not that big of a difference in value between the C+ and A+ solutions. I emphasize this concept in job interviews. I received job offers when I stopped trying to sell myself as some data prodigy and started to sell myself as a dude who does things, even if those things aren’t perfect. That’s why the hype of “data scientists” seems to have died down a bit. Yes, you can hire someone to perfect machine learning models. You can also hire someone to commit RandomForest(max_depth=10) into production, and the business won’t lose a cent.
Which brings me to AI. Yes, ChatGPT writes poorly optimized and awkwardly documented code with errors and deprecated coding packages. You know who else does? Me. But now I write it in 15 minutes rather than 4 hours. I spoke with a marketing specialist who also finds ChatGPT helpful. She uses it to automate the simple parts of her job, which increases the amount of time she can spend on the more important and valuable documents. I also see people from a variety of backgrounds and roles use ChatGPT to write the first draft. Sure, I wouldn’t publish a ChatGPT-written article or send a ChatGPT-written email. Still, Anyone all writes knows that sometimes you need fucking anything on the page to get you started. ChatGPT is great for that.
deBoer also highlights the mediocrity of AI art. He contrasts it with old-school cartoons, such as the drawing on the right side of this image. I find the example a bit odd. That John Candy cartoon is fine, I guess, though I don’t think the Vatican will tear down The School of Athens to make room for it anytime soon. I find the cartoon art mediocre and forgettable. However, I concur with deBoer’s view that AI art is mediocre and forgettable. You know what there’s a lot of demand for, though? Mediocre and forgettable art. Consider board games like Terraforming Mars or Race to the Galaxy that feature half-assed “outer space” pictures just to put something on the cards. In a recent game of Terraforming Mars, I earned a card that featured (we believe) the Hoover Dam with a purple filter over it. The future of space colonization is, apparently, a half-hour drive from my house. If these games were published in 2028, the makers probably would have typed “space stuff” into an AI and called it a day. Many Substack blogs use AI art because something has to go in that thumbnail. Think of all the boring, C+ art you encounter in your daily lives. AI can automate that.
Others criticize the translations produced by the AIs. Yes, they’re flawed and the Goddess of Perfect Translation disapproves. As long as she stays in Mount Olympus, though, we can enjoy the benefits afforded to travel, education, and collaboration by these C+ translation systems.
Again, the world runs on C+ work. If you want to get promoted, don’t “nail” a single project, make meaningful improvements on a bunch of different ones. My linguistics articles may read like bargain bin John McWhorter columns. Very little similar content exists, making me one of the best linguistics writers for a general audience by default. While I’d hesitate to refer to anything I’ve done in the dating world as “success,” I’ve gotten better at talking to women by attending okay venues where I talk to women I’m mildly interested in using conversational strategies that sometimes work. One of my friend groups takes an infamously long time when it comes to agreeing on where and what to eat. When it’s my turn to order, I quickly pick something that mostly works for everyone. It’s more important to put some food in front of our faces than to get all signatories to concur on the finer details of the Treaty of Pizza Hut.
The Personal Aspect
Why did I write this article? I’m not thrilled by AI, and this piece doesn’t amount to a passionate defense of the technology. Rather, I’ve grown tired of non-stop negativity and pessimism. For example, it’s been about a year since I wrote about Bullshit Jobs, even though the topic began as one of the blog’s main focuses. I stand by the commentary, but I just don’t want to bring any more cynicism into the world. This article, then, amounts to a criticism of my past self. I’ve spent too much of my life focused on limitations, flaws, and imperfections. I’ve grown tired of that perspective, and I’ve grown downright exhausted by how popular that perspective has become among certain parts of our culture. Yes, lots of things have lots of flaws. No, you’re not that smart for pointing them out. In most areas of life, a pile of C-plusses beats a mountain of critique. The one who usually gets the promotion, the girl, or the World Series ring isn’t the one who finds perfection, but the one who most reliably produces the C-plusses. Next time you want to criticize, produce some C+ work instead. Trust me, it’s worth its weight in bronze.
What a relief to read something sensible about AI! When news about Chat GPT first broke, I was surprised by all the panic over what struck me as an obviously useful tool (but nothing more than that). Your article also reminds me of an ongoing debate my husband and I have about maximizers (he is one) and satisficers (I am one). My husband likes to note that if it weren’t for maximizers seeking something better, we would still be living in caves. But, as you point out, if it weren’t for us satisficers being fine with C+, we would never be able to agree on a pizza.
Incidentally, I agree that you are up there with John McWhorter as one of my two favorite writers on linguistics!
This is fantastic. I'm taking this advice to heart immediately.
Also: I'm overall bad at following this advice as of now, but I'm also the food-decider in the friend group (and my family, who reliably get extremely moody when they're hungry but can never decide where to eat). I once saved a vacation by swerving off the road into the parking lot of a shitty RV park diner, because it was the only food we had seen in the Utah wilderness for three straight hours and my family STILL wondered if we should hold out for "the next thing."