HACKER Q&A
📣 behnamoh

Should I cancel my GPT-4 subscription and get Copilot instead?


GPT-4 seems to suffer from some limitations after the last update. It doesn’t remember what it did just a couple interactions ago, and keeps writing buggy code that I have to fix. It used to be much more useful but over the weeks, it’s gotten almost useless for my work. Aside from coding, its responses to general questions have also gotten more shallow. I have to push it to generate something I need because it keeps producing bullet points about what needs to be done but then never does them.

I suspect that the API might be better (if what the oai employee said is true), but I don’t see services that use the API for a cost. Poe is one option but they cap the number of messages to 300/mo which is basically nothing for my usage.

I just canceled my GPT-4 subscription plan for next month and I’m increasingly thinking about switching to alternatives. For coding, I don’t mind closed source solutions as long as they work like the good old GPT-4. I’m also open to open source models but I hear they’re not even as good as gpt3.5.

It’s sad that we’re stuck in this monopoly of powerful LLMs.


  👤 GartzenDeHaes Accepted Answer ✓
I just canceled Copilot and my experience was:

PROS

- Good at boilerplate

- Good at unit tests

- Fairly good at pasting from Stack Overflow

CONS

- Most of the time it does nothing

- Large code blocks pop in and out while I'm typing, which is distracting

- Sometimes it correctly guesses what I want, but the code is completely wrong with non-existent variable and method names

Overall, I'm more productive now that I'm not hesitating and waiting to see if Copilot does anything.


👤 smoldesu
> I’m also open to open source models but I hear they’re not even as good as gpt3.5.

WizardVicunaLM claims ~97% performance relative to GPT3.5: https://github.com/melodysdreamj/WizardVicunaLM

It's not particularly great at generating code, but it's uncensored and writes fantastic prose. I've been using it for the last week and I'm really satisfied with where it stands.

> It’s sad that we’re stuck in this monopoly of powerful LLMs.

Won't anyone just sponsor a few months of dedicated GPU training, finetuning and quantizing so they can be held legally accountable for it's output?

I wouldn't hold my breath.



👤 yuppie_scum
Get an API key and a client for GPT, it’s super cheap.

Copilot is at the very least a great time saver and well worth $10 considering that’s a rounding error of my salary.