Get Your AI off My Lawn!
I don't care if I sound like a grumpy old person. I probably am a grumpy old person.
As I logged onto a Zoom meeting I’d scheduled, an uninvited visitor tried to crash. It even used my name. I booted Fireflies.ai Notetaker Rosana out of the meeting before it could start recording, which was apparently its goal.
As I texted with my sister, the Messages app suggested some super-clever responses to her I’d never have come up with myself, like “Haha, that’s a win!”
As I checked a work email inbox, I noticed that Gmail wanted to summarize my emails for me.
Seemingly overnight, like some kind of volunteer butler I never hired, AI was everywhere I turned. I was in a futile game of AI whack-a-mole. When I wrote about ChatGPT two years ago, this kind of AI for the masses was still new. Since then, AI seems to have taken over everything. Its strange promise is that we’ll never again have to do anything on our own.
That was driven home to me the last time I updated my Mac OS. A message popped up on my screen letting me know I could enhance my writing with “Apple’s built-in personal intelligence,” billed as “AI for the rest of us,” which “draws on your personal context to give you intelligence that’s most helpful and relevant for you.” Wondering what that means? Yeah, it’s pretty vague mumbo-jumbo, but apparently it’s about helping you get stuff done effortlessly.
I mean, I hope I have my own personal intelligence that I use for writing. And anything worth doing takes some effort.
Let me be clear: I’m no Luddite. Though ever aware of any new technology’s dangers and unintended consequences, often welcome it. I embrace home automation like porch lights that turn on at sunset and off at sunrise, irrigation we can control with an app, and a smart thermostat that theoretically saves energy. I appreciate how GPS has kept me from getting lost on many occasions. I don’t know how I lived without my Paprika recipe app, I spend too much time online, and I love the cloud much more than its energy-guzzling ways warrant.
But I think AI is already going too far — and it’s just getting started.
I can understand the desire for AI to liberate us from tedious stuff. That was the original vision: that AI would do tasks that we don’t want to do, can’t do, or take lots of time for us mere mortals, freeing us up for more interesting and rewarding endeavors. Even that idea relied on some miraculous shift to a Universal Basic Income that would keep us all going when AI took away our jobs, something that doesn’t seem likely to happen. Still, it was a vision with some merit.
I can also see AI’s amazing potential as a tool to aid in humans’ work. It can help (though shouldn’t substitute for) doctors in diagnosing illnesses. It can be used to discover new drugs and tailor them to patients. It can enhance GPS navigation and emergency response. It can balance the energy in a building or on our nation’s electricity grids. I’m sure there are numerous, maybe countless, other solid uses of AI.
But for some, AI now means not “having to” do the things that make us human. Things that bring us joy, like writing and making art — you know, creating stuff. Things like reading books.
In case you have trouble understanding this Tweet, no need to rely on or practice your basic reading comprehension: Grok, the AI of the platform formerly known as Twitter, is here to help.
Good thing I had that added insight! I wouldn’t want to have to think for myself.
Okay, so I don’t want AI to read books for me. Or write for me, God forbid. Or even rewrite my texts or emails in a more friendly or more professional tone.
What about images? Despite the image at the top of this post, I’m very uncomfortable using AI for image creation because it feels like stealing actual artists’ work. However, in my work I now and then need a generic image — like the image of a church with solar panels I tried generating recently. I asked ChatGPT for an image of a large brick church with solar panels far from the edge of the roof and no shade on the panels. I got this:
Despite ChatGPT’s confidence in its creation and cheerful offer to modify it, you can see shade on some of the panels — plus, the panels at the top are too close to the edge of the roof (I won’t even get into the odd shapes of some of the panels). I asked for a refinement and got this:
ChatGPT was lying to me. Again there was shade on some panels. Plus, I didn’t want the smaller roof sections to have panels on them, so I asked to remove them. The next image I got:
More lies! Not only do two small roofs on the church have solar panels, but one is in full shade. Not to mention the weird panel layout. And why did all those other roofs with panels pop up? I asked to remove them, only to be met with even more blatant lies:
After a few more rounds, I gave up. Even the most basic AI prompt wasn’t working for me. Of course, I was using the free version of ChatGPT; my father always did say you get what you pay for. But I had no better luck with the paid Canva Pro I use with the client I was creating this image for.
When I tried to create a simple image of a scale with apples on one side and oranges on the other, one side heavier than the other, neither AI was up to the task — and both lied to me about what they’d produced.
If AI — at least, the AI accessible to me — can’t even create a simple image without lying, how can we trust it to provide accurate information about anything?
I’m sure there are better AIs than the ones I tried, and I’m sure these cheaper, less robust text- and image-generating AIs will improve over time. But I’m not sure that will be a positive development. In addition to the ethical issue of them stealing artists’ and writers’ work, we’re increasingly being faced with the ethical issue of images, and even videos, potentially being fake. In our current environment of rampant disinformation, that’s the last thing we need.
Like all technological advances, AI has its pros and cons. Like all genies that have been let out of the bottle, it can’t be squeezed back in. Whether we like it or not, AI is here to stay. That makes it all the more critical that we address these issues now.
We have yet to see what AI will do to our world. For now, I still wish it wouldn’t pop up everywhere I look. I still wish it would get off my lawn. Is that too much to ask?
The first human to get an idea about how to use fire also probably got a little over excited. That's human innovation for you: it works on this, so how about that? And before you know it not only are we cooking with it (which is amazing), but also accidentally burning down the house (major bummer)!
The AI image generator you're using has no spatial sense. I see the same thing with AI code generation. It seems like the kind of thing that could be trained in, maybe, I dunno how.