Rishabh emailed me the other day, asking me to answer the 7 questions of his new blog challenge, and who am I to say no to such a request? So here we go.

How was your first experience with AI models?

I assume by AI models we mean the current crop of LLMs, and not AI models in general, because I’m old enough to remember when “Machine Learning” was a thing. What even is AI anyway at this point, since everything is lumped together into one useless definition? Anyway, I believe my first experience was trying out chagpt back when it first came out. I don’t think I spent more than 10 or 15 minutes using it at the time. It was impressive tech, but was also completely useless for me at the time, and that’s why I didn’t bother spending more time using it.

Do you use AI or are you completely against using it?

This is an interesting question. Do I use AI? Well, I guess the answer is yes since it’s almost impossible to avoid using it if you use the web at this point. Pretty much all tools and services are integrating some sort of AI-powered functionalities, and it’s become harder and harder not to use them. If, instead, the question is do I use one of the various LLMs directly to do stuff, then the answer is still yes, but the amount of usage is so low that some people might consider that to be the same as not using them at all.

I don’t directly pay for any of the models, but my work email has been powered by Google for more than a decade, and so I do get access to Gemini Pro. Workspace has usage data for everything, and I just looked it up: In the last 90 days, the only AI-related feature I used was the Gemini App (that’s not surprising considering I turned off everything else), and I have apparently used it 62 times.

I’m now looking at the history of those chats, and pretty much all of them are single-question queries related to something web dev I was doing. Things like how to do a specific thing inside Kirby, or how to achieve something using a particular JS library. This is stuff one should be able to find inside documentation websites, but the search there is often awful and so after a google search, I try my luck with AI.

And as I wrote somewhere else, I never copy-paste. I ask very narrow questions so that I can be pointed towards the correct answer. And once I have that, I do the coding and I re-implement everything myself.

Am I against using AI? As a generative tool, yes. I refuse to ask AI to do something for me or to generate content from scratch. As a tech in general? I think it has some potentially useful applications in narrow contexts. As always, the answer is not cut-and-dry, and it can be yes or no depending on the framing and the scope.

Do you have any preference among different models, for example Claude vs ChatGPT? If yes, how do you choose?

Nope.

What aspect of AI models do you like and what do you not like?

The only aspect I appreciate is the ability to ask questions in natural language. Because sometimes you have a problem or an answer you’re looking for that can’t be described in a more structured way.

As for what I don’t like, how anthropomorphised these stupid tools are is definitely high on my list. I don’t want my computer to talk back or to make jokes or to say «I’m sorry». If I input a question, I want an answer back, and that’s it. I don’t want follow up questions, I don’t want some pointless preamble. I get why this happens, but I fucking hate it. This is software. I don't want my software to have a personality. I want it to perform a task and get out of my way.

I also don’t like the lying, the gaslighting, and all the other crap, and I also don’t like what the AI industry is doing as a whole, but that’s a separate issue.

How do you feel about AI generated images? Does it annoy you if someone use them in a blog post?

Again, another question that has different answers depending on the scope. The idea of being able to generate images, in general, is neutral to me. It all comes down to what you use it for. There are some potential use cases that are totally fine, others are completely insane. As a whole, I think the ability to generate slop is bad, but that’s because humanity can’t be trusted to do anything the right way.

As for their use in blog posts, I think stock images were useless, and I don’t see images generated with AI to be any different. Unless you have generated an image as part of the content to explain or visualise something. That’s fine, label it as a generated image and move on. That’s no different than including a render, or a sketch on paper, from a content perspective.

Internet is flooded with AI slop now, full of generated text, images, audio and videos. How do you filter it from authentic human creation? Do you have a strategy?

My consumption of online content these days is so limited that I don’t have this issue. I read very few blogs, and I know they are not AI generated because I emailed the people behind them more than once, and I know what their stance is. I watch almost no YouTube, and I only read a few news sites. My strategy is to simply stay away from the digital world as much as possible, and I’m at the point where I’m considering dropping my digital consumption down to zero and quit the internet as a place for content.

Are you hopeful for a better future with A.I. or a dystopian one?

I have zero hope. And that is because I have zero hope in anything that’s in the hands of mega corporations. The incentives are totally skewed, and they’d do everything they can in order to keep the line go up. I don’t see people with strong morals in positions of power and so unless we decide to go full French Revolution, I see no reason for things to improve.