As someone who daily drives ChatGPT for a lot of stuff, I agree. I heard someone put it this way "AI is perfect for stuff that is hard to find but easy to verify".

The other day I took a photo of my liquor cabinet and told it to make a cocktail recipe with ingredients on hand. Or if I encounter an error on my PC I'll just describe the problem. Or for movie recommendations when I have a very specific set of conditions. Or trying to remember a show from my childhood with only a vague set of memories. The list goes on.

Particularly for anything coding. If I'm trying to learn something I always learn best when I can just see an example of the thing in action as documentation is not always great. Or if I'm doing data manipulation and I have the input and output and just need the function to convert one to the other. I recently saved a whole afternoon of effort with that one. Or spec tests I'll just drop my whole code file in and ask it for full coverage.

These are all things that traditional search engines are poor or incapable of. I'd have a hard time going back if they just turned all this off tomorrow.

I think there's a lack of education around how to use AI which is actually a problem. Like you shouldn't be using it to identify if a mushroom is safe to eat. You shouldn't be using really for anything food or health related for that matter. You should ask it for its sources when you are unsure of its answers.

I just wish companies would stop cramming it into every product and making it the selling point for everything because they can't think of anything else to do. It's kind of fun to play with sometimes but it just isn't useful at all for most tasks that companies seem to want us to use it for. I'm hoping it will be mostly forgotten outside of niche uses in a few years the same way we forgot about 3D TVs or NFTs as soon as the next new tech buzzword got invented

It will transform mediocre content into crap, and then it will regurgitate that crap into worse and worse crap. All of that at the expensive of untold Terawatt-hours of energy. AI will not take over the world, but it sure will help destroy it anyway.

Its energy usage and effectiveness vs human laziness that worries me.

The problem is not the tech, as usual, it's the people, who have been led to believe it's AGI, who equate forming syntactically correct sentence with intelligence, and that that is enough to perform most white collar tasks.

There are ethical problems with how many of the models have been created and some of what they're being used for

But I generally agree, it's pretty much the same thing we see with every technological innovation—something big changes and a load of things get disrupted, a group of people then get angry about said innovation, eventually those people dwindle and the innovation gets absorbed into the general public's idea of what modern life consists of.

I can't think of any big innovation over the past few decades that hasn't really followed a similar trajectory

The dangers of AI should be determined by application, not capability. It’s a tool, like any other. You can use a hammer to build a house or cave in a skull.

The problem is that not the tech. The problem, as always, is Capitalism.

That's exactly what AI would tell you

To offer a counternarative, we don't really know where AI will plateau, what will a 1 quadrillion parameter language model look like? And will it make my degree worthless by comparison lol?

It isn't as great as it's proponents say either.

Even AI porn is getting good.

AI is awesome! Don't let the luddites convince you otherwise.

Ai is the way!