It depends on what you mean by “wrong”. Bugs can exist in code where the shell doesn’t hit the target it was supposed to. There can be a hardware malfunction. There can be user error where the user picks the wrong target. There’s multiple types of “wrong”.
I think you’re making my argument for me.
I don’t even understand what your argument is. You seem to act like code can just be churned out of a computer like a printer prints a report.
When did I say that?
Saying they can write code is simple enough. Saying they can write complex code is simple enough. You seem to act like a computer is just going to decide to hack Wall Street for no reason. How do you imagine it’s going to do that besides “code”? A bug? A bug is going to cause it to write code to hack Wall Street? Somehow it will write simpler programs without this issue coming up neither during testing (which would likely take years) nor its initial release (which would also last years), and not a single issue will arise, and yet somehow it will create a program that targets the systems Wall Street uses on accident? So somehow this computer is able to not only understand code and its own architecture well enough that it can program in machine code on its own processors, but it apparently knows the architecture of Wall Street’s systems well enough to exploit them on accident. All the while this incredibly complex code goes completely outside the notice of the developers who would likely be checking this machine’s code since it’s for some other extremely complicated task, and during their likely years of reviewing and testing, they failed to notice such a gaping error. Or they were stupid enough to let it run freely yet were smart enough to teach it to code. And we haven’t even discussed the code written by other developers used by Wall Street.
How do you come up with that? When did I say that?
How are they going to write code we don’t want when it’s been presumably not coded to do so?
But if I write the code, I may program it to do things you don’t want it to do. So you can’t presume that at all.
Ah, so you want it to be a Stuxnet that writes itself, a worm that took the combined resources and brain banks of both Israel and the United States. And this program is just going to whip that up?
When did I say that? Making up an argument and then attributing it to me is not proper debate.
You say someone is presumably going to write a program which will write a program targeting Wall Street. Stuxnet was a computer worm designed to target Iran’s nuclear systems, and that’s what it took to make it.
I certainly don’t. No one and nothing exists in a vacuum. Every entity needs some starting point, human or A.I. What would a feral child come up with? Nothing at all.
Complete non sequitur. Programs have already been written targeting Wall Street. Such programs will presumably become more powerful and capable as time goes on. Computers may be employed to write the code – and the complexity makes unintended consequences more likely – even ignoring the evil intent of the original programmer.
Those programs weren’t written by another program. I would expect code that could write code to hack Wall Street would at the bare minimum require what Stuxnet had. And even then I doubt they could pull it off with fifty years.
And so will security.
There it is again, your answer is the computer may just “code” it up. Somehow this computer will go from hacking Wall Street, which would require a narrow enough focus as it is, to somehow growing beyond its specifications to something worse. And these hackers are too stupid to even review what their program is doing since it would then likely not be doing what they told it to do.
Yes, but mostly because we don’t like being disturbed.
Unless you’re bringing free pizza, in which case, leave it on the console and please leave!Valid point, but just for the sake of argument let’s assume that you’ve created an AI that could pass the Turing test. Of course that wouldn’t prove that it was conscious. But let’s assume that you really want to know if your AI is conscious. After all, there might be some things that are ethical to do to a mere machine, that aren’t ethical at all once that machine is conscious. So what could you do to determine if your “machine” is conscious?
