• lloram239@feddit.de
    link
    fedilink
    English
    arrow-up
    2
    ·
    1 year ago

    The work was about 80% reading the API’s documentation, 18% configuring my API keys and downloading the example project and things like that, and 2% writing code to hook everything up.

    Or in other words 100% of the work could be automated by AI quite easily. The fun with ChatGPT and friends is that it doesn’t stop at programming, it can automate all the other stuff as well. It can read documentation, answer emails and produce shell code just fine. It’s also not like you are “programming in English”, you are having a conversation with an AI programmer. It can dynamically change the code and adopt to new requirements. Don’t like the Python code it put out, ask it to rewrite it in C++. Don’t like it using libbar, ask it to use libfoo instead. That’s the power of AI programming, it gives you a much higher level of abstraction than any classic programming tool. And it’s cross domain, if you need some plausible data for a test case, it can write that too. Want to replace your programmer art with something professional, done. Need an icon for your app, done. If you are just looking for a way to name a thing in your code, AI is the best thesaurus you could hope for. If you like, it can even review and criticize your requirements.

    And yes, ChatGPT, especially the free version, gets everything outside of trivial problems wrong. But the point about AI is not were we are now, but where we will be in 5-10 years. ChatGPT is just the first prototype that kind of works. It’s the AltaVista of AI, wait until we get to the Google of AI.

    Given the restrictions under which ChatGPT has to work (no access to documentation, everything done from memory, no ability to actually test code or use a compiler) it’s insane how good it already is.