I just don't want the world to be destroyed, and there are already enough ways the world could be destroyed without AI being added to that list.

Do you think those war crimes wouldn't happen without AI? Just because war crimes are committed with the help of AI doesn't mean they wouldn't happen otherwise. At least I don't see how war crimes happened less often in the wars some decades ago.

And then you have to forbid computers because they are used for war crimes. And without computers people will use pen and paper to plan their war crimes so let's forbid that as well...

AI will be used in wars but those wars wouldn't be nicer without it.

Then we have robots and drones. Robots will be a huge part in future wars and drones are already but those don't even need AI because they could also be controlled by humans far away. So let's forbid humanity to work on anything robotic or anything which can fly. 

Ok, fair enough. But wouldn’t it be easier to send a drone to go kill say a family in a vehicle than physically shooting them down? If you did the latter you would have to get them in your sights and pull the trigger, much more visceral of an experience. I think that ai will make war crimes much easier and therefore common. 

War crimes happen because those people want them to happen or don't care. Those who decide this stuff will tell their AI drone or some bomber pilot to do the job and both will do it without asking twice. Or did any pilot bombing Gaza ever stop because they thought "nahh"? They will destroy every single house in a country if they get told to do it. 

Yeah, I agree. To name just one example: AI is currently finding bugs and more importantly security issue in open source software. Following this as a software engineer can be a bit scary as so many issues are raised in a short time. But we have to remember: raising these issues means the projects now fix them. This ultimately improves the safety of the software. I guess in a few years or maybe a few months already the pace of bug reports will slow down, as the bugs AI can find will be found and fixed. Which makes us all a bit more secure.

And that is fair to criticize. AI after all is just a tool. I don't want to use the tool to kill people and worry about military use. At the same time the same tool can be used to improve software, and it already does. Not by vibe coding, which leads to horrible code. But by code review, security bug finding and deliberate improvements. And many projects already use AI in this responsible way. So use it for good things, not for military and surveillance.

You're mocking a very real possibility of AI destroying us.

What AI needs are restrictions, safety measures, and a way to function without data centres.

ai still another 2-3 years to reach full potential

copilot is doing good in some aspects, but in most it is true that it's lacking compared to chat gpt and gemini :/

Artificial general intelligence is probably way further than that still. Until then, any AI that tries to be too general-purpose will probably keep being naive/stupid and make mistakes accordingly and hence be too unreliable to rely on without humans supervising them. Personally I wouldn't expect even AGI to completely solve those issues, but it should significantly improve things (although AGI is probably where concerns about AI safety get more serious).