Technology

I work/study in AI, and it is completely over-hyped. For one thing, the C-suite can't wrap it's head around the fact that AI != LLM; they all seem to think all AI is just LLMs. On top of that, they are way too eager to throw humans out of the loop.

That said, I think LLM applications, even in their current form, are super useful in development and business practices. I myself use it to increase my productivity in coding. But, I use it as an augmentation rather than a replacement. One of my friends put it best the other day, "LLMs are like a junior dev to your senior dev. You need to be hyper-specific, and you need to check it's output." In other words, it's great for off-loading some work, but it isn't going to completely replace humans.

With that said, I'm a bit annoyed that other AI fields are being over-shadowed by LLMs. There's a ton of other interesting work being done in those fields that is super useful and important. All of them, though, are not going to replace humans but rather augment and make humans more productive. I've found that an AI-Human team is most effective.

I promise you they will not rehire those they fire to replace with LLMs. I cannot post which company I work for but it one of the largest employers in my country with tens of thousands of employees. We have laid off entire branches of our organization in favour of enterprise GPT-4 licenses.

Several sectors of HR have been replaced, our program analytics and accounting branches have all gone through 3 levels of cuts (we now have a chatbot that handles data viz and HR mediations/parts of various HR processes. Our graphics and design teams are now Stable Diffusion.

These changes were implemented in August of last year. Thousands of employees have been replaced by a handful of different generative AI licenses and we had to take on an additional cost of hiring an offshore consulting firm to feed these models our business data so they would be more customized for our org.

I’m looking for my exit because despite the increased costs and automation, things aren’t working well at all. The HR bot sometimes tells you rules that don’t exist (had to look deep into my contract because HR bot told me I wouldn’t be eligible for severance pay which was false). All of our workloads have quadrupled with losing a good chunk of our staff and trying to fix things that are now broken. I went from 40h/week to 60+ and I’m definitely on the lower spectrum of this.

(Imagine a "you are here" arrow pointing at the peak.)

I use gpt-4 daily for my work, I love it. I get more done and save time.

The thing i don't trust the simulation for one bit. If you analyze it properly you will always find errors.

Its way easier to fix those errors or to adapt the solution into my needs. But the most essential component of the entire workflow is that i a human with real life expertise green light the good parts from its response. A bit like i became a teacher and chatgpt is the student. Its works extend mine but I remain at all times responsible.

Maybe in a few years it could be ready, lets just hope it doesn’t go the same way as nfts, which could have had a brilliant use to track the authenticity of journalism in a fake news world, but no, monkey craves golden bananas even if it cant digest them. And now the aftertaste is so bad people get sick at the mention.