AI is spackle

Every major tech trend seems to draw two types of reactions. Boosters make excited pronouncements about the technology’s stunning and awesome potential, while catastrophists predict impending doom (either for the technology or for society at large).

Today’s mania for AI follows this pattern exactly. Boosters, mostly from the Silicon Valley companies that own the technology or invest in it, promise that AI will save the world. It’s hard to exaggerate the claims being made. Take, for example, Marc Andreessen’s post from June 6 called Why AI Will Save the World. Education, healthcare, government, and on and on – all of them will become exponentially better, Andreessen says, because AI. (It’s like a magic incantation – AI will solve it! – much like, a year or two ago, we heard that crypto would solve everything. But I digress.)

Outside of Silicon Valley, catastrophists delight in pointing out the flaws of AI and its impending failure. Ted Gioia’s August 23 post, for example, reports that AI demand is already shrinking: people are using it less as they realize how erroneous or harmful the output is. No one wants AI except scammers (“cheaters and criminals are taking full advantage of AI as a tool of deception”).

I’m more aligned with the catastrophists than the boosters, but I’d like to propose a different way of understanding AI, something that I hope gets at the truth better than an extreme of “awesome” or “worst thing ever.”

First, I’ll note that part of the problem of the AI discussion is that it’s not always clear what is meant by the term “AI.” Is it chat bots? Is it Skynet? Is it the singularity? Most techies will agree that AI isn’t quite any of these. There’s an old joke among programmers that once computers can do something reliably, it’s no longer AI. I suppose partly because of the clever branding – “AI” sounds so futuristic – people naturally have inflated expectations about the tool.

I’d define AI as the use of complex algorithms on massive sets of data. With enough processing power, the application of algorithms to large datasets gives rise to champion chess programs, and pattern recognition in tumor scans, and ChatGPT. A more accurate name to this technology might be “massive spreadsheets,” but it just doesn’t have the same ring as “AI.” That magic incantation.

With that said, let me propose a different way of viewing AI that acknowledges its use but steers away from the overhype that permeates the public discussion.

Spackle everywhere

What I’d like to suggest is that AI is spackle.

Home repair often calls for a binding substance to fix a crack, smooth over a hole, or just serve as an all-purpose filler. Gash in the drywall? No need to hand-construct that part of the wall, just add spackle. Crack in the ceiling? Spread on the gray goo.

AI is the same sort of tool. Need some boilerplate language? Get the AI to generate the text. You’re coding and need a quick sort function? Spray some AI on it. An AI text generator can be a quick-and-dirty tool to fill a gap in the text. It’s boring, low-budget, a bit of a patch to paint over, in limited quantities, once in a while. Spackle has been with us for a long time, and so will AI.

The problem is that the public conversation has framed AI as a much bigger, more influential tool than it really is. Boosters and catastrophists both tend to overstate the importance: it’s the best thing ever, or the worst thing ever, and we’re living in a cosmically important moment, as we’re present at the creation of this New Thing: an AI brain that will save us, or destroy us. This rhetoric is so tiring.

AI is spackle. It has a place, filling the occasional blank, barfing out a sentence or two. Even as a search engine alternative it might be helpful as a conversational interface, for low-stakes questions (as one can never be 100% confident in the answers). Bits and pieces here and there, covering some gaps to be swiftly painted over with higher-quality tools.

I want to emphasize that AI is not inherently futuristic, any more than spackle is a spectacular marvel of ceiling care. We’ve had spreadsheets, data, and algorithms for a long time. Sure, the algorithms today are more sophisticated, and (thanks to intrusive online surveillance pioneered by Google and Facebook) there’s a lot more data available to crunch. But the act of using computers for pattern recognition is hardly new.

The danger here is that the boosters win the public conversation and people begin to adopt the boosters’ faith in AI as a salvic force. While it has its uses, AI should be used for – at most – an occasional input helping an actual expert, filling a gap or offering a suggestion. But we shouldn’t build our economy on AI, just as we wouldn’t build a house out of spackle. There’s a better way forward: tone down the hype, understand what AI is actually for, and keep humans in charge.

To support my newsletter, please join the Creative Good Forum. (You can also post and read comments.)

P.S. Question for anyone living in Washington, DC: do you know anything about what’s happening with the Whole Foods that installed QR-code surveillance throughout the store? We're discussing it on the Forum but can’t find anything except one Newsweek article. Any tips appreciated: mark@creativegood.com

Until next time,

-mark

Mark Hurst, founder, Creative Good – see official announcement and join as a member
Email: mark@creativegood.com
Listen to my podcast/radio show: techtonic.fm
Subscribe to my email newsletter
Sign up for my to-do list with privacy built in, Good Todo
On Mastodon: @markhurst@mastodon.social

- – -