Hacker News new | ask | show | jobs
by rpdillon 322 days ago
I'm an engineer and my general attitude is that when major new developments enter the space, it's worthwhile to understand how they work, why they were created, and what they could potentially be good for, and what they potentially won't be good for, and why.

I applied this methodology to mobile, cryptocurrencies, Web3, VR, and AI. I ended up being bearish about cryptocurrencies and Web3 and bullish about mobile, VR and AI. VR is having a fair amount of trouble taking off, but I think AI is not facing the same challenges. As such, I'm betting very strongly that it's going to be part of our future, and that means now is a good time to start understanding how it works and what people are saying about it and what might come next. I do this by putting my hands on the technology and running lots of different models locally and also experimenting with cloud models, as well as exploring what tools companies are building with these technologies. Simply reading doesn't quite do it for me.

I don't do this because I'm worried about being left behind, nearly as much as I want to be able to deploy these tools effectively in the future to solve problems. It's also fun!

1 comments

> As such, I'm betting very strongly that it's going to be part of our future, and that means now is a good time to start understanding how it works and what people are saying about it and what might come next.

Getting a feel for what LLM's can and can't do is also proving to be an interesting and fun exercise for me. I use them daily, but haven't committed a line of code written by one. I suspect that will always be the case, but nonetheless not using them would be a career limiting move.

I suspect that's because an LLM is just a specialised AI that's not good for much beyond understanding the written language and regurgitating what they've seen before. (That almost sounds flippant, given how much in awe I am at both their language abilities and how much they remember.) There are lots of other AI's out there. They play chess and go, do protein folding, predict weather and I'm sure a plethora of other things I don't know about. They are all universally better at their respective tasks than humans (and that includes LLM's ability to process language.) I can't shake the feeling that one day, someone will create one that can do more than spit out code similar to what it's seen before.