I constantly feel overwhelmed by everything LinkedIn is throwing at me. It’s something most of us can probably relate to, and it’s been bothering me for a while.
In tech, it seems most of us have to be “jacks of all trades” to an extent – I am not sure if anyone has the luxury of just sitting there working with a very specific set of technologies without having to at least be aware of what’s happening in tech overall. And being aware is not enough either. If you want to have some way to classify the information you see, you need to understand it somehow.
There is no argument there, right?
However, just building on my personal experience… A couple of years ago I was reading all those posts about AI, and I was getting a strong feeling that “we are all doomed”. Until I made an effort to learn more about what LLMs do, how they work, etc. No, I’m still not an expert to explain the math, but, once I got the basics (or a bit more than just the basics), things started to make much more sense, and “we are all doomed” started to turn into “well, maybe not yet”. I mean perhaps we are still doomed because of the so-called “AI bubble”, but that’s a different thing.
Problem is, this just shows how little I know and how bad of a judge I am. Just yesterday, I was trying to set up an MCP server with Azure Functions. It’s simple, right? Well, kind of… Except that I have not done Azure Functions in a while, really (and, to be honest, I have never really done them, other than occasionally), so I had to catch up on a few things I did not know, had to update my local setup, and, of course, had to go through some samples. All in all, a few hours in, and I did have that basic MCP server working.
That illustrates the problem, though.
I can’t appreciate the complexity and gotchas until I tried something. I also have no idea whether what others are saying is true/false/works in general/works in very specific cases/etc. But, also, if I am using AI for help, and if I do not have enough experience in what I am asking AI to do, I need that AI to be 100% reliable (maybe not 100%… but this works similar to how I would trust a professional to do their work. Yes, there could be issues/bugs, but there would also be a process for resolving them). If AI hallucinates, if it makes up an answer, or if it provides results which are not accurate, how would I even know?
That is why I totally love to see what AI can do, every now and then, but I still can’t find a good use case for using AI outside of my own area of expertise. I just wouldn’t be able to validate whatever it is AI would be doing. By extension, I am not able to find a good use case for AI in the business scenarios I’m normally working on. It might be useful for internal users, but they would have to be able and willing to validate AI output every time they converse with AI, or it would be a disaster at some point.
Then again, it’s not as if AI were responsible for the unmanageable complexity of our technology, which is what’s making it hard for the AI to stay highly reliable. There is a lot happening everywhere, we’ve accumulated a lot of knowledge, and nobody is able to know everything and/or to predict that much at all. If you think it’s manageable and someone out there really knows what they’re doing, yet if you blame yourself for not keeping up, just look back two or three years. Look at all the news and announcements there and see which ones have actually materialized by now.
And that’s the problem, though. Yes, I can keep learning more and more, but there is only so much one can keep in the “active” memory. I have not done much work with the Azure Functions lately, and I had to re-learn a few things once again. So there we go, then:
- There is too much knowledge for anyone to be an expert in everything
- AI is supposed to close that gap by giving us an “expert” (there is even a name for that in the AI architecture: a mixture of experts)
- And, yet, we still have to validate AI output no matter how good AI is in some cases, since it is still capable of absolutely failing in others
So how can someone reconcile all this in their minds? Which is where it goes back to my LinkedIn intro above. A much better option might be to start simplifying this whole tech landscape – instead of having so many frameworks, just have one. Instead of having so many clouds, just have one. Instead of all those API versions… yeah. As if the specifics were that important these days for the end users to bother. Of course it’s never going to happen, since those differences is how “business” works, it’s literally feeding on them.
Ultimately, I do feel overwhelmed since I suspect someone is not; and, perhaps, it’s the second part that’s incorrect. I mean yes, of course there is too much information out there to digest all of it, but it’s the same for everyone. So it’s more likely that nobody is actually keeping up, we are just all pretending a little differently.