Hacker News new | ask | show | jobs
by tester756 1296 days ago
>For the past year I've been looking at all the progress happening in ML/AI and each day I'm more convinced that there's a lot of game-changing stuff that will come out of it (what we're seeing with Stable Diffusion and GPT3 are some examples of this).

Wow, I must admit that out of all big branches of computer science AI/ML is the least exciting for me. I don't know, but all that unreliability is just putting me off.

I do agree that it's better to have something automated in 90% instead of 0% or 40%, but the impossibility to getting it to 100% is annoying.

It's been like over a decade of huge hype on AI/ML and yet I feel like biggest applied AI/ML that affects my life directly or indirectly is search & ad industry or things like chat bots, but it's mehhh.

I don't believe in autonomous cars based on computer vision
2 comments
karmakurtisaani 1296 days ago
I really want to like AI research, but ultimately find it kind of boring. For me it's a combination of increasing complexity of the models and the requirement of having lots of appropriate data and computation power. Things are more interesting on the mathematical side, as long as you are happy with not getting anything practical out of it.
link
imhoguy 1296 days ago
> getting anything practical out of it

You need to start low, e.g. my friend is non-IT but I have shown him SD so now he tries out AI-art t-shirts business. But I agree, AI is completely still dumb, especially at scale - just asking my Android Auto Google if there is paid parking zone at the destination it navigates to, "Sorry I don't understand". We are still far from ELI5 usability.

link
rakejake 1296 days ago
I specialized in ML in my Masters, and I was surprised by how boring my first job in ML was. Sure, the results are "sexy" and the job itself is apparently much sought after, but building the model was frustrating. Like you say, the interesting part for me was what people would think of as boring - getting the data cleaned up and correctly aligned, the scaling work, latency issues with serving the model etc. I knew then that ML engineering was not for me.
link