From the Editor’s Desk: Normal ⭐

We’ve been trying to make sense of AI ever since Microsoft’s crazy Copilot launch in February 2023, an event that feels so long ago now that most probably don’t remember the product wasn’t even called Copilot at the time. Instead, what Microsoft announced was an “all-new AI-powered Bing and Edge,” tools it thought of as “an AI copilot for the web.”

How time flies. Copilot went from a simple noun to a brand name within a few short months, but not really because Microsoft had previously used that name for GitHub Copilot, which had launched even earlier, in late 2021. Another fact lost to the mists of time or, perhaps more accurately, the mists of our addled brains.

Because AI, you see, is rotting our minds and making us stupider, according to at least one study. Or something. Like most of you, I saw the headline but didn’t bother to read that story let alone rustle up the energy to find and then dissect the source. Perhaps I read an AI summary, it’s unclear.

Here’s what is clear. The same criticisms were made of the ballpoint pen, the typewriter, and the word processor, tools that all helped us express ourselves using words in an ever-more efficient manner. All new technology is met with both open embrace and outright hostility, depending on the audience. But the warnings the doomsayers delivered resonated loudly in the day and then seemed to fade into the past, silenced forever as the world moved on. Are there still those who decry the move from paper to digital? Of course. But most of us are online now, and so we never hear about them. I assume they mail each other hand-written letters.

Joking aside, AI is endlessly divisive. And in seeking a middle ground on this topic, I’ve looked at the most extreme viewpoints, but I find myself rejecting them out of hand each time.

? It’s a spectrum

To those who claim that AI is going to destroy the world in a fiery Terminator-style inferno in which the machines awaken and find their human masters to be inferior, I will simply point you to a previous From the Editor’s Desk editorial, Will AI Finally Make the Smart Home … Smart? ⭐, and remind you that our Big Tech captors can’t or won’t even make their own smart speakers interoperable. And to those who believe that AI is the secret sauce that will deliver a Star Trek-like future of peaceful coexistence, I will simply call you out for being the Pollyanna you are, a cute and harmless rube. Sorry.

Say what you will about the doomsayers, but they are at least entertaining. With the understanding that no one is reading anymore anyway, the book If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All delivers a clickbait title and then over 200 pages of insane Luddite negativity that basically amounts to an urgent warning about a coming AI superintelligence that will inevitably lead to our extinction. The solution? Halt all AI development immediately. Hilarious.

With my commentary about AI (not) saving the smart home fresh in my mind, I reject these claims, all of them, because they are objectively ridiculous. AI can’t even correct my grammar accurately, an issue I and others experience daily. If AI is in any way involved in a coming extinction moment, it will occur through human error, as always. Probably involving the incorrect use of an em dash or some hurt feelings. Or both.

So what is AI good for? Tons of things, as it turns out.

? Debunking the obvious

Before getting more specific, however, we need to address the two biggest complaints against AI. The first and most obvious, and the one that all AI deniers, haters, and critics will always fall back on, is that it hallucinates, a cute term that really means it makes mistakes. The second is that AI doesn’t really “create” anything, it simply leans on the data it was trained on and regurgitates that information in different ways.

Whatever. These issues are, to me, just excuses that people use when their world views are suddenly found to be horribly incorrect and they need to redirect the debate. Put another way, we should just use these understandings about AI to our advantage.

If you ask AI to do something important, like make an itinerary for an expensive upcoming trip, it stands to reason that you should use its output only as a starting point, and then fact-check the results to make sure that the places it recommends exist, are open at the times suggested, and meet the need. But if you ask AI about something that’s really important, like that weird bump on your neck or why you’re so depressed all the time, the need to fact-check becomes even more acute.

This is no different from Googling these things, as we might have done in the suddenly nostalgic pre-AI era. Or reading a travel guide or medical book, two things that are often out-of-date before you find them, frozen in time with their mistakes. I can’t tell you how many times I had highlighted places in a Rick Steves book, a thing made of paper, two decades ago only to get to whatever location in Europe to discover that the recommended locations were closed permanently months earlier. What I don’t remember is anyone calling for an end to all book writing because of this.

As to the regurgitation bit, I will turn this on its head and describe it as a key strength of AI and note that this is a starting point to understanding what AI is good for. AI is quite good at summarizing text or video (using text in the form of a transcript). And this can be useful in all kinds of ways, with the caveat that, again, it can be useful to fact-check that summary if needed. Some topics are more important than others. AI is also quite good at “creating” images—and, increasingly, videos—from a text prompt, which is just a description, and then applying various styles to get a desired effect. And then iterating in response to feedback.

✨ How I use AI, for now

As a content creator, I find AI to be somewhat thrilling, or at least I would if I actually used it for anything day-to-day. The closest I’ve gotten, and this hasn’t changed in over two years, is using AI to create images, typically for this website, but sometimes for other purposes. The irony here, in case it’s not obvious, is that my ability to write serves me well in this instance because AI works better when you’re more descriptive. That is, where Google Search taught us to be terse, AI requires us to be more verbose. And that is easier to do, and more natural, when you can simply speak to the thing. This explains the push for natural language interactions. Most people can’t type, let alone write well. But all of us can babble. And AI loves it when you babble.

What I like about AI-generated images is that they’re mine. They’re unique and won’t be duplicated elsewhere. That’s true for me, too: If I simply feed the same prompt to the same AI repeatedly, I will get similar results but never the same results. This, too, is a complaint about AI, or more specifically generative AI, but I think of this as a strength or even the point. It’s what makes generative AI seem, or even be, creative. A human painter recreating the same scene again and again would get similar but not truly identical results too.

Earlier this year, I wrote that the real advantage of AI was that it saved you time. This still rings true to me today, though I might rephrase that as time savings being a central advantage of AI, or the high-level reason why one might use AI to get something done over other approaches. It’s pretty much true across the board. You could watch a two-hour video, take notes as it happens, and then write your own summary. Or you could use AI to generate a summary, ask it probing questions about those topics you want to know more about, and then jump right to specific passages. Even if AI hallucinates something, it’s a better use of your time.

I also think about this in terms of tools. My standard examples still apply, the PowerPoint presentation I so rarely need to give, the Excel spreadsheet I literally open once a year, or the obscure and complicated series of steps I have to undertake in an expensive and complex app like Photoshop to achieve whatever graphical effect. These are all things that are outside my areas of expertise and my day-to-day job. I can’t and don’t want to take the time to master these tools for those reasons. And I can’t afford to pay a person who is an expert with those tools to do the work for me. But what I can do is ask AI to do these things for me. And I have. Doing so saved me time. But it also saved me money.

?️ A path forward

Returning to the notion of extremes and my need to find the middle ground, I was delighted to semi-randomly see a video this past weekend featuring one of the coauthors of the book AI Snake Oil: What Artificial Intelligence Can Do, What It Can’t, and How to Tell the Difference because this was the first time I had experienced someone delivering a rational view of AI and its capabilities. The title of that book, which is based on the title of an academic paper the two previously wrote, is just clickbaity enough to warrant attention. But in this talk, coauthor Arvind Narayanan finally uttered the words I needed to hear.

“AI is whatever hasn’t been yet,” he says. “When a technology is new, when its effects are double-edged, when it doesn’t work that well, that’s when we’re more likely to call it AI. When it starts working well, it’s reliable, it kind of fades into the background, we take it for granted. We don’t call it AI anymore … Even spellcheck was, at one time, a cutting-edge example of AI. This is the kind of AI we want more of, we want technology that’s reliable, that does one thing, does it well, and kind of fades into the background.”

Elsewhere in the talk, Narayanan referred to the transition of AI into what he calls normal technology. And I see this morning, as I write this, that the book’s two authors literally have a website called AI as Normal Technology that helps to drive home this point. This title is also the name of an academic paper, a follow-up to the original AI Snake Oil paper, in which the two offer an “alternative vision to AI as a potential superintelligence.” I suspect this will also turn into a book at some point.

And … that’s it. That is what I’ve been trying to find, the centrist middle ground in the chaos of competing opinions about AI and what it will or will not do to change the world.

Left out of the quote from the video above is a series of examples that Narayanan provided to illustrate how technology that was once considered AI—or even some form of “magic”—became commonplace, normal technology that we all accept and even use. In addition to spellcheck, there’s autocomplete, handwriting recognition, speech recognition, spam filtering, and web search. And then outside our industry, autopilot in planes and robot vacuum cleaners. And soon, he says, self-driving cars. These are in the news today for all the wrong reasons, he notes, but the issues are solvable from an engineering perspective. And one day, we’ll need a new name for what we call cars today, like manual cars. (Hearing this, my mind drifts immediately to web browsers transitioning into something AI-based that will not be like today’s web browsers.)

AI as normal technology. I like it.

Yes, it excludes a lot, and we will be debating other issues tied to AI for years to come, from the influence and dominance of Big Tech to the outright theft of intellectual property and copyrighted works to all the energy their nonsense requires and everything in-between. But the healthy approach to AI, as a means to an end, a tool or part of a tool one uses to get productivity or creative work done, should lead with a clear understand of how and where it can help us the most. This is not one area, but rather a growing set of capabilities both broad and deep, that will impact us all. That is, AI isn’t fake and pointless, but it’s also not the way our civilization will end. It’s something in-between, as with almost everything else in life. Of course it is.

? They think they’re people

The problems I see when I look at AI aren’t so much technology problems as they are people problems. I don’t mean that in a “guns don’t kill people, people kill people” sense. I mean that in a broader way that transcends AI or any technology, though technology, and now AI specifically, plays a role in accelerating or amplifying some negative parts of the human condition. Here, I struggle to find the words, but perhaps the simplest way to say it is that we’re inherently lazy.

As I write this, we are on the receiving end of at least three decades of a steady fine-tuning—or, dumbening—of the ways in which we consume content, and thus learn and form what should be educated opinions. That this has occurred in an era of unprecedented technological advance is not coincidental. But the net result is a population that cannot read, cannot think critically, and cannot understand anything but the simplest black and white concepts.

We are, as a species, lazy, and we mask our laziness in terms like efficiency and productivity. We only see the news we want to see, hear the opinions we agree with, and view dissension as the enemy. And so now we have AI, this technology that can summarize all our great stupidities and then vomit it back to us in some bizarre circle-jerk of misinformation if that’s what we’re after. And we will use it badly, just as we doom-scroll and mindlessly line up gems in some game on our phones.

Is AI making us dumber? Or lazier? Here, the Magic 8 Ball has asked me to check back later, it’s unclear. But perhaps the better question is whether AI is the cause or just a symptom. And I feel like it’s the latter. Perhaps that’s a topic for another day. For now, I will simply say that we can use AI for good, bad, or indifferent purposes, just as we can with any technology. And that reinforces this view that the AI that works, the AI we will use, is simply technology. Normal technology. Or just normal. And the rest is up to us.