shortsighted AI praise
Today I’m thinking about how so much of generative AI goodwill and praise focuses on text generation only and relies both on other options becoming worse and believing AI/user interaction stays like it is right now (aside from model improvements).
We use AI as a better search engine. Would we do that (as much) if Google didn’t go to shit and Reddit search is trash? When we use it to chat and roleplay, would we even be interested if social media and chats didn’t become awful or shut down? We use it to generate recipes, but would that be necessary if sites weren’t cluttered with SEO bullshit, 10 modals and ads covering the content? When we use AI to access specialized info, isn’t that just a sign that many specialized forums died, too much science is paywalled and info is locked in FB groups and Discord servers?
Instead of changing the things we could change, we just shrug, defeated, and perceive AI as good because the bar is in hell. It’s like letting your home deteriorate and then marveling at the cleanliness of a new thing you bought.
As I said, it also assumes things stay like they are right now - no ads, no sponsored content, just a clean message feed with quick replies containing the occasional hallucination. This might stay for freely available models, but what about the rest? Their current state is only possible because it’s the hype right now and VC money is flowing. And who knows for how long free models can compete?
One day, the money dries up or profits need to be made, just like what happened when everything became enshittified to squeeze as much money out of everything as possible.
My prediction: On the free tier, you’ll have to wait 3 minutes between prompts, there are ads between the messages and it’ll suggest you sponsored content first in its responses. You’ll get less and less prompts per day or reduced output size. You’ll see a lengthy reply by the LLM that fades out just like digital newspaper paywalls, asking you to upgrade to see the full thing. It might be biased towards companies and products that paid for exactly that (you’ll be lucky if they’ll even mark it as such). You’ll probably be able to pay to remove it, so the AI providers are getting money from both sides for the same thing.
Could be entirely wrong of course, but I believe it for now.
So when I see praise about its ease of use and swiftness and how much better it is than xyz old method, I think about how the other method used to be that, too. It’s just a cycle. I guess the people that’ll profit the most from it are the ones who get access via their company’s enterprise contract.
Reply via email
Published 05 Feb, 2025