6 ms·

Wouldn’t the end game here be that the AI just writes straight to machine code? It’s not like some devs are safe, those who maintain coding languages would be o

Wouldn’t the end game here be that the AI just writes straight to machine code? It’s not like some devs are safe, those who maintain coding languages would be out of a job too. I think AI would find it awfully silly to have to code up a language, maintain it, then turn around and use that language to code with. Bare metal AI seems like the end game to me.

efitz 2y ago
We’ve lots of programming languages; most businesses need only an OS (Linux- free), an application language (python, java, C*), a database language (SQL), and a front end be language (js + html). No need to go to bare metal in most cases as generative AI is already good at these. The mostly unsolved problems right now are around build, deployment, and operations.
im3w1l 2y ago
No I don't think so. Both human and AI benefits from building on top of existing abstractions: langugages, frameworks and libraries.
righthand 2y ago
Code is written to be auditable by humans, if you just generate machine code how will you know the level of correctness? Would you save the prompts then I have to prompt review before merge? That would end up being much wordier than a code file.
- tasuki 2y ago
  Humans write the tests. AI writes the implementation. Just write good tests: property based. Or actual formal specification...
  - episteme 2y ago
    It's much quicker to write code than a formal spec for that code.
    - rootnod3 2y ago
      Yes. Because at the end, the code should be the spec or the final source of truth for it.
      - tasuki 2y ago
        I disagree. The spec should be the spec. The code is just an implementation detail.
        righthand 2y ago
        The style of the code is an implementation detail, once implemented it enters the realm of spec.
    - batshit_beaver 2y ago
      This right here is why I have serious doubts about the ultimate usefulness of LLMs in coding. Code is the spec, and it's far faster to write the code yourself (if you know what you want to do and how to write code) than try to pass all this context to an LLM in the form of lossy and imprecise human language.
    - tasuki 2y ago
      Yes, especially so if you do not much care whether it's correct.
ilaksh 2y ago
I don't think that makes sense in general, but does remind me of an idea I had where it does make sense: 8-bit program synthesis. Your dataset is all of the 6502 binaries, corresponding manuals, and descriptions of gameplay generated by multimodal AI looking at videos, screenshots, etc. Maybe even screenshots and memory captures of games could be input along with sequences of key or controller inputs. But if you feed in a large chunk of the Z80 and 6502 programs and as much metadata as you can and for a training signal make sure that the thing runs without freezing on an emulator and maybe shows the correct loading screen or something.. I think there are enough 8-bit programs and games for this to work, at least to some degree. Maybe. Start with shorter simpler programs and work up. Generating BASIC might be easier in a way, I don't know. But the idea is you would describe your game along with maybe a screenshot or cover art or something and it would generate the machine code. You could also do something similar but on a frame-by-frame input-by-input basis like Oasis. Anyone in this thread want to fund that? I would love to work on it.
utopicwork 2y ago
[dead]

efitz 2y ago

We’ve lots of programming languages; most businesses need only an OS (Linux- free), an application language (python, java, C*), a database language (SQL), and a front end be language (js + html). No need to go to bare metal in most cases as generative AI is already good at these. The mostly unsolved problems right now are around build, deployment, and operations.

im3w1l 2y ago

No I don't think so. Both human and AI benefits from building on top of existing abstractions: langugages, frameworks and libraries.

righthand 2y ago

Code is written to be auditable by humans, if you just generate machine code how will you know the level of correctness? Would you save the prompts then I have to prompt review before merge? That would end up being much wordier than a code file.

tasuki 2y ago
Humans write the tests. AI writes the implementation. Just write good tests: property based. Or actual formal specification...
- episteme 2y ago
  It's much quicker to write code than a formal spec for that code.
  - rootnod3 2y ago
    Yes. Because at the end, the code should be the spec or the final source of truth for it.
    - tasuki 2y ago
      I disagree. The spec should be the spec. The code is just an implementation detail.
      - righthand 2y ago
        The style of the code is an implementation detail, once implemented it enters the realm of spec.
  - batshit_beaver 2y ago
    This right here is why I have serious doubts about the ultimate usefulness of LLMs in coding. Code is the spec, and it's far faster to write the code yourself (if you know what you want to do and how to write code) than try to pass all this context to an LLM in the form of lossy and imprecise human language.
  - tasuki 2y ago
    Yes, especially so if you do not much care whether it's correct.

ilaksh 2y ago

I don't think that makes sense in general, but does remind me of an idea I had where it does make sense: 8-bit program synthesis. Your dataset is all of the 6502 binaries, corresponding manuals, and descriptions of gameplay generated by multimodal AI looking at videos, screenshots, etc. Maybe even screenshots and memory captures of games could be input along with sequences of key or controller inputs. But if you feed in a large chunk of the Z80 and 6502 programs and as much metadata as you can and for a training signal make sure that the thing runs without freezing on an emulator and maybe shows the correct loading screen or something.. I think there are enough 8-bit programs and games for this to work, at least to some degree. Maybe. Start with shorter simpler programs and work up. Generating BASIC might be easier in a way, I don't know. But the idea is you would describe your game along with maybe a screenshot or cover art or something and it would generate the machine code. You could also do something similar but on a frame-by-frame input-by-input basis like Oasis. Anyone in this thread want to fund that? I would love to work on it.

utopicwork 2y ago

[dead]