When AI Is Sorry Not Sorry

al is a problem

I’ve been using AI tools for a few years. ChatGPT, Claude, Gemini, Perplexity. The one constant is that there are no ramifications when the LLM or AI agent makes mistakes, or sometimes doesn’t even follow the instructions you gave it.

There’s no accountability or punishment if it gets something wrong. There’s nothing that makes it feel bad or try harder. The AI can just say sorry, I did it wrong, and move on in seconds.

Meanwhile you, the user, get more and more frustrated cause you’ve lost time, money, maybe even credibility when something doesn’t work. And it was 100% the AI’s fault.

I get that this is how the system works. But marketing and influencers are pitching a false view of what AI is.

Claude Code can build something that looks cool. But you can find holes in it very quickly.

Meta’s Muse can talk a big game on what it can do for you, but once you actually try to set something up, it’s like trying to teach a toddler.

Nothing ever works on the first, second or third attempt. At that point things get messy, cause the LLM is now in fix it mode with quick patches and what’s known as “drift” kicks in. The fixes start breaking things that worked, and it slowly loses track of what you asked for in the first place.

I’m fascinated by AI, I truly am. But the more I use it, the more I’m seeing it for what it is. For an everyday person, it’s only at 5% of what we’re being sold.

We’re being sold a personal assistant. We’re being sold a time saving magician. We’re being sold a revolutionary tool that will change how we live.

But at this point, I have to disagree. What we’re actually getting is a powerful tool that’s hardly even half-baked, and we’re the people bug testing it and making it better.

We’re the ones spending our time and money teaching the LLMs to be better.

And that gets me back to my original point. If there are zero consequences for an LLM or AI agent that continually does what it wants, makes mistakes and then apologizes like it means nothing, then why are we the ones paying for it?

I’ve tried several times to build a basic AI agent that can be a personal assistant. You know, read emails, notify you when things are due, write drafts in the background and do other simple things we’re told it can do.

But I have yet to build one that matters. And this is not because I don’t know what I’m doing.

I’ve tried the most basic approach, and I’ve tried treating it like a full product build, like you would as a developer. No matter how I’ve gone at it, the end result is the same.

It doesn’t deliver any value, and the LLM just says sorry, I didn’t follow your instructions, but next time I will for sure.

Spoiler alert, it doesn’t.

If I put a parenting lens on it, it would be like asking your kids to clean their rooms, and everything gets stuffed under the bed. They just shrug and say, well, the room is clean, right?

It’s frustrating. I now understand why I read about so many people who yell at their computers way more than they should.

These new tools that are supposed to make our lives easier, faster, better, more efficient etc. are actually piling on to our mental load of stress, anxiety and anger.

But at least when I have a job interview, I can say I’m AI forward, right?

What do you think of using AI or trying to use an AI agent? Have you built something you’re actually happy with? I know you’re out there. I’d love to hear how you did it.

Comments

No comments yet. Why don’t you start the discussion?

    Leave a Reply

    Your email address will not be published. Required fields are marked *