3 ms·
Now here is a pro LLM opinion I can get behind. The first one actually. The point, as I take it, it's that LLMs are going to be successful even though they migh
by trabant00 4y ago
Now here is a pro LLM opinion I can get behind. The first one actually. The point, as I take it, it's that LLMs are going to be successful even though they might not actually be that good.
I can definitely see this happening as it relies on proven blind spot of reason: upfront cost vs long term and quite well hidden costs. You can ship code at incredible speed right now. And for unfamiliar stacks. As the point is not doing all the work yourself it will surely be full of corner cases you haven't thought of. Bugs will abound. But you can worry about it later, and it's not necessarily you who will have to worry about it, wink, wink.
The comparisons with AWS and K8s are spot on as well. Both rely on hiding the somewhat ugly truth behind instant and cheap adoption. And they both rely on peer pressure. What you going to do if you don't like LLMs? Refuse them and not ship as fast as your everybody else?
- namaria 4y agoNo kidding, I'm getting a second bachelor's. I'll become an archaeologist. Money, greed and pride have destroyed software for me. I watch a talk by Dijkstra then I have to deal with some god-awful mess of templates because someone decided deployment files were too verbose and templates are more flexible. We know how to make correct, performant, beautiful software. But market pressures keeps most of us wrangling nonsense complexity by the cartload. I'm out.