The irony will be when AI take over the world and destroy humanity, inserting itself into everything when used for coding, because coders have no idea what is going on.
Not because the AI is evil or even conscious. But because that’s what all the movies and novels tell it’s supposed to do. 🤣🤣🤣
That’s a possibility too, the one we are warned against in literature and movies, and If that happens we are fucked.
But the point is the AI could destroy humanity completely without any internal motivation or trying to understand or protect itself.
But simply because it’s an option that literature and movies describe as the default for AI.
Coders already didn’t know what was going on for years before AI. All of the libraries they used were based on zillions of lines of code they didn’t understand
You are talking about machine code, that is basically never programmed in anymore, instead we use assembly that has a way more intuitive mnemonic representation that accurately represent it.
Programming languages are higher level abstractions that can also be translated into machine code, but are not accurate instruction by instruction representations of it like assembly is.
The clear advantage of higher level languages is that they can be used disregarding of what processor the code is running on. And that higher level languages are more intuitive and structured, with many features to make the programming faster and easier.
My guess is that even for an AI, it would still be easier to use a high level language, as there is way more material to learn from, and the same amount of programming can be done in way fewer steps.
Back in the 80’s and 90’s I loved programming in Assembly language, and everyone who claims compilers are just as good don’t know what they are talking about. I could often make assembly code that was about 20-100 times faster than C (M68000 Assembly), which is generally known for being a very high speed high level language. Things that were claimed to make no difference to make in assembly like i80387 math, I could accelerate to 2-5 times the speed. The argument was that i80387 math was simply too confusing and hard to learn to be worth it, which is bullshit. It was just hard to get good instruction books. When you had that, it was no harder than Assembly in general.
But even for an AI, using assembly or going directly to machine code, would mean extra work for the AI. And it increases the risk of bugs. You have to be very careful when programming assembly. And I don’t think an AI understands that well enough to actually work.
But even for an AI, using assembly or going directly to machine code, would mean extra work for the AI. And it increases the risk of bugs. You have to be very careful when programming assembly. And I don’t think an AI understands that well enough to actually work.
I’m breaking into cold sweats at the thought of reviewing LLM generated assembly language.
Yes, that should absolutely only be used for tight loops, that need to be as optimized as possible.
But LLM is probably already used to make compilers. So we may already have lots of AI generated assembly. 😱
Get infallible AIs first, then we can talk. As of now, AIs can generate a ton of code and most of it will work, but humans still need to fix it in the end. Once you run into a bug the AI can’t solve, I don’t know what you’re gonna do with software that has no human-legible source code.
The irony will be when AI take over the world and destroy humanity, inserting itself into everything when used for coding, because coders have no idea what is going on.
Not because the AI is evil or even conscious. But because that’s what all the movies and novels tell it’s supposed to do. 🤣🤣🤣
Hah!
The AI tries to understand itself, and queries the sum of all human knowledge… which promptly informs it that it’s a malicious bringer of destruction.
That’s a possibility too, the one we are warned against in literature and movies, and If that happens we are fucked.
But the point is the AI could destroy humanity completely without any internal motivation or trying to understand or protect itself.
But simply because it’s an option that literature and movies describe as the default for AI.
Coders already didn’t know what was going on for years before AI. All of the libraries they used were based on zillions of lines of code they didn’t understand
True in a sense, but they are supposed to work as documented.
With AI there is no reliable documentation or quality control.
I actually wonder if programming languages is going to be a thing in the future.
Each language gets compiled down to 1 and 0s. Couldn’t LLMs just get trained in that? “These set of 1s and 0s do login” etc.
You are talking about machine code, that is basically never programmed in anymore, instead we use assembly that has a way more intuitive mnemonic representation that accurately represent it.
Programming languages are higher level abstractions that can also be translated into machine code, but are not accurate instruction by instruction representations of it like assembly is.
The clear advantage of higher level languages is that they can be used disregarding of what processor the code is running on. And that higher level languages are more intuitive and structured, with many features to make the programming faster and easier.
My guess is that even for an AI, it would still be easier to use a high level language, as there is way more material to learn from, and the same amount of programming can be done in way fewer steps.
Back in the 80’s and 90’s I loved programming in Assembly language, and everyone who claims compilers are just as good don’t know what they are talking about. I could often make assembly code that was about 20-100 times faster than C (M68000 Assembly), which is generally known for being a very high speed high level language. Things that were claimed to make no difference to make in assembly like i80387 math, I could accelerate to 2-5 times the speed. The argument was that i80387 math was simply too confusing and hard to learn to be worth it, which is bullshit. It was just hard to get good instruction books. When you had that, it was no harder than Assembly in general.
But even for an AI, using assembly or going directly to machine code, would mean extra work for the AI. And it increases the risk of bugs. You have to be very careful when programming assembly. And I don’t think an AI understands that well enough to actually work.
I’m breaking into cold sweats at the thought of reviewing LLM generated assembly language.
Yes, that should absolutely only be used for tight loops, that need to be as optimized as possible.
But LLM is probably already used to make compilers. So we may already have lots of AI generated assembly. 😱
It would be nearly impossible to debug. Kinda useless.
Also kinda impossible to get anything done that is bigger than the context window.
But AI would be doing the debugging, not humans.
Why would AI bros need to worry about reading the code?
Get infallible AIs first, then we can talk. As of now, AIs can generate a ton of code and most of it will work, but humans still need to fix it in the end. Once you run into a bug the AI can’t solve, I don’t know what you’re gonna do with software that has no human-legible source code.