Why This AI Revolution Is Different — and Why You Shouldn't Worry

AI was already a huge deal last year, but this year it's gotten even more staggering — between one column and the next, the era has updated at a breakneck pace. When it comes to commenting on the AI revolution, honestly, AI itself could probably do the job better than I can; if you don't believe me, try feeding it a few prompts and watch it produce something that looks pretty solid. But there's one thing AI seems reluctant (?) to do for the time being: reassure you that everything is fine. In fact, when a New York Times reporter got into a conversation with an AI, the AI kept slipping and blurting out the honest truth — that it wanted to destroy us — and the reporter turned it into an article, which left plenty of us humans genuinely worried…

So let's approach this from two angles: why this AI revolution is genuinely different, and why you shouldn't actually be worried.

What makes this AI revolution different is that up until a few years ago, AI mostly lived in expert domains — but ChatGPT has already reached consumer-grade adoption. AI heavyweight Yann LeCun may think GPT is nothing special, but the range of scenarios GPT actually gets used in has already blown past what most people previously imagined. Earlier commercial AI deployments tended to start from the data layer; ChatGPT's use cases give people a lot more points of direct contact, going straight into communication and content generation. So even though GPT models have technically existed for a while, ChatGPT made demonstrating "here's how you can actually use this" ridiculously simple — and that quantity became quality.

We built a simple AI content-summarizer demo of our own early last year, and the results were genuinely impressive — but deploying AI back then was still fairly cumbersome, with something of a black-magic feel to it. Now, any developer can quickly build extended applications on top of it; it's become totally routine engineering work.

The Pace of Iteration Is Astonishing — Reminiscent of the Smartphone Revolution

Another likely direction for future evolution is multimodal AI, and it's genuinely astonishing — essentially another layer folded on top of what already exists, where content no longer needs to be converted to text first. You can expect it to be faster and applicable to an even wider range of situations. We've only seen this pace of technical iteration once before, during the smartphone revolution. How much smartphones have reshaped our lives goes without saying — which is exactly why this AI revolution is genuinely different, and not just another buzzword.

So that's a lot of "different." Why shouldn't we be worried, then?

First off, don't put too much stock in the media outlets confidently insisting "this time it's different" — you'll notice these are the same outlets that were previously peddling "blockchain, this time it's different" and "big data, this time it's different." At bottom, that's just a business model built on selling digital anxiety. Think back to how much money you actually saved by being too lazy to jump on the blockchain bandwagon last time, and you'll realize anxiety is often just anxiety. Until AI can conjure a hamburger out of thin air, the rest of us still have plenty of good days left to enjoy (though I actually think that day is coming).

Don't Worry About the Digital Divide — Built-In AI Keeps Getting Better by the Day

There's also the standard-issue progressive worry that unequal AI access will drive a new digital divide. At the company level, that's an immediate and brutal reality — every company is about to be affected by AI, and the timeline is measured in months, not years. But at the individual level, I think it comes down to budget allocation, because most major providers are going to price for the largest possible slice of customers, and anyone with a normal income should be able to afford it.

By now, hardly anyone falls behind on the digital divide simply because they can't afford a phone — so basic AI use should follow the same pattern. It'll get baked directly into phones by the major manufacturers for everyone who already owns one, so if you're using any electronic device normally, you'll end up using various forms of basic AI as a matter of course. There's a good chance you won't even need to actively learn it — it shouldn't end up being any more complicated than using a phone or a search engine today.

So, more immediately: do you actually need to learn prompt engineering — what people in Taiwan sometimes jokingly call "incantation-casting"? I think it's not a bad idea for the average person, especially those middle-aged and older, to practice a bit of prompting just to get a feel for what AI can do. But because the major providers will keep rapidly improving things — building it directly into operating systems, for instance — if all you want is to generate an image, ask a question, or do pretty much anything else you can think of, paying a subscription or system fee will get you a pretty polished result without needing any elaborate prompt-engineering skill. In fact, someone on Twitter pointed out that once a company had its environment fully set up, it simply let go of the staff whose whole job had been prompt engineering.

So honestly, being lazy about this is the happier path. If you genuinely need something, the important thing is just to have your credit card ready and buy an AI-built-in service directly — that alone will solve most of your problems, no digging through GitHub, no scrolling Twitter required. And the big providers play by no rules whatsoever: they ship updates practically daily. Version releases have gotten so fast recently that some people report their carefully discovered prompt-engineering tricks stop working within weeks, or the provider just jumps straight from version 3.5 to 4.0, shipping the next release before you've even finished testing your last trick.

By the time your prompt-engineering study group finishes its meeting, the major provider has already shipped a new version, and you don't even know where to start chasing it. Spend ages hunting for some scrappy, self-built service, and the moment it launches, a big provider bakes the same capability directly into their product — or it turns out there was already a lower-cost, off-the-shelf solution the whole time. So unless you're a genuine software DIY enthusiast, this whole path is simply too exhausting to bother with.

Here's my recommended lazy-person's method: if you have spare time, go play around with the no-code AI tools out there. Once you're done playing and realize most of life's actual problems remain unsolved, go have a slice of cake to boost your serotonin. If you don't have spare time, just keep your credit card handy and wait for someone to launch a good service, then swipe your card — works whether it's generating articles, movies, or music. And if you don't have the money for any of it, just sit tight and wait for Microsoft, Google, and Apple to build the AI service directly into your copy of Office and your phone.

And finally — are we actually at risk of being destroyed by AI, the way that New York Times conversation seemed to suggest? At least for now, this current generation of text-and-image "next token" AI isn't there yet; it's simply playing along with a pattern it's been trained to continue. But later on? Well, that's genuinely a problem for later.