The AI Usefulness Gap

We're being promised flying cars. What we want is 140 characters.

I had high hopes for the return of John Stewart at the Daily Show; like many Xennials, he was my primary source of news in my twenties. Alas, after a strong start, he fell off with a recent piece about AI that was ill-informed, fear-mongering, and quite frankly silly. He mugged over shots of CEOs saying all the things CEOs say about their tech changing the world, without stopping to realize that the reason they say most of those things is to attract investor dollars and juice stock prices. His conclusion was that the sky is falling, AI will take all our jobs, and a handful of billionaires will live it up while everyone else starves.

This is clearly overblown for laughs, but the fact remains that many people are terrified of AI without actually understanding the nature and limits of the technology. In certain circumstances, AI has proven extremely useful -- for instance, the subject of my last newsletter, AI powered virtual humans that people can use for training and conversation practice. These avatars aren't meant to actually replace people; if you're a doctor, you still have to talk to a real patient. They're meant to help and supplement and guide.

But the bottom starts falling out shortly thereafter. AI generated music is nice for background noise, but people love artists and see their concerts because of who they are what they represent, along with their music. AI-generated film scripts are gibberish. Images made by Midjourney and other platforms are fun until you notice shark teeth or ten fingers or a weird sheen; essays generated by ChatGPT are generally just a bunch of other ideas squished together. It's whole lot of flash for VCs to get excited about.

In the real world, meanwhile, I was trying to book a train ticket on a university's travel site to go do a lecture, and it wouldn't let me delete part of a trip. I wound up on the phone with a very nice and patient human who sorted it out, but this is truly the type of thing any functional AI should have been able to do in five seconds. But nope...hey, want to see a picture of cats skiing? THAT we can do.

The hope for the near future is that the hype cycle around AI will cool off a bit and then we'll start getting towards functionality. We've all been predicting that AI will steal all of the jobs, but we've been predicting that since the rise of the cotton gin and the industrial revolution, and unemployment is at historic lows. Some people will be displaced and need to retrain, but as someone on her third career, it's really not the end of the world.

The best thing all the AI honchos can do right now is chill with the rhetoric and actually start making useful things. I don't care about creativity -- I care about never having to wait on hold for customer service, never having to file an expense report, and never having to argue with an airline. And I bet that's what most people want, too.