How are you, friend?
-
How are you, friend?
*Listens and nods*
Now that that's out of the way...
How are you, -really-?
I'm here for you, sweet one. Let me know if you need any help.
🩵🩵🩵
-
How are you, friend?
*Listens and nods*
Now that that's out of the way...
How are you, -really-?
I'm here for you, sweet one. Let me know if you need any help.
🩵🩵🩵
@Bwee i've been assigned to write a whitepaper about AI-assisted software development and the dangers thereof and I can't even make myself start on it because I don't even know how to start the research
-
@Bwee i've been assigned to write a whitepaper about AI-assisted software development and the dangers thereof and I can't even make myself start on it because I don't even know how to start the research
@AVincentInSpace@furry.engineer @Bwee@meow.social Setting aside the general issues about AI like training data and resource consumption (that is a whole different rant), here are some dangers I find with AI software development:
- risk of lack of understanding. If you outsource your development to AI, you're not training your development skills and also not building your understanding of the code. If you do not mitigate it, you'll end up with bugs you can't fix. I found a lomger article here: https://dev.to/dev_tips/ai-killed-my-coding-brain-but-im-rebuilding-it-4i35
- Code quality of AI-generated tools can be low. Yes, surface-level it's great, but on a deeper, more structural level it tends to recreate functionality that's already somewhere else in the system.
- AI doesn't push back if you ask it to do something that's not optimal (when trying to work it into code could let you smell the code smells that maybe you need a different solution)
- AI only has the context you've given it (and even then it's not clear if it uses it all)
And finally: what will happen if the bubble bursts, your favorite AI provider shuts down and the available providers cost thrice what they now cost (and I'm being optimistic here)? -
R relay@relay.mycrowd.ca shared this topic
-
@AVincentInSpace@furry.engineer @Bwee@meow.social Setting aside the general issues about AI like training data and resource consumption (that is a whole different rant), here are some dangers I find with AI software development:
- risk of lack of understanding. If you outsource your development to AI, you're not training your development skills and also not building your understanding of the code. If you do not mitigate it, you'll end up with bugs you can't fix. I found a lomger article here: https://dev.to/dev_tips/ai-killed-my-coding-brain-but-im-rebuilding-it-4i35
- Code quality of AI-generated tools can be low. Yes, surface-level it's great, but on a deeper, more structural level it tends to recreate functionality that's already somewhere else in the system.
- AI doesn't push back if you ask it to do something that's not optimal (when trying to work it into code could let you smell the code smells that maybe you need a different solution)
- AI only has the context you've given it (and even then it's not clear if it uses it all)
And finally: what will happen if the bubble bursts, your favorite AI provider shuts down and the available providers cost thrice what they now cost (and I'm being optimistic here)?@Bwee@meow.social @AVincentInSpace@furry.engineer Also, I agree that it's very difficult to do research on this: for every problem you want to research, there are dozens of articles about how great AI with the same keywords
