Kinda wish I knew more about "AI"
-
But I can't be arsed to learn it. I do have an idea though - using an LLM to do code synthesis. Code synthesis is similar to code generation, but unlike code generation code created via code synthesis looks as though it was written by a human being. This is particularly important for things like parser generators, where generated code tends to be rigid. For example - a generated LL(1) parser always matches greedy while a hand rolled recursive descent parser will switch between greedy and lazy consumption as the situation calls for it. The end result is more compact and more efficient. I don't recommend using parser generators because of the code they produce - unless you like source files in the hundreds of thousands of kilobytes or even megabytes. Code synthesis stands a good chance of changing that. Things like Github Copilot can produce code. Code synthesis seems like this functionality.
Check out my IoT graphics library here: https://honeythecodewitch.com/gfx And my IoT UI/User Experience library here: https://honeythecodewitch.com/uix
-
But I can't be arsed to learn it. I do have an idea though - using an LLM to do code synthesis. Code synthesis is similar to code generation, but unlike code generation code created via code synthesis looks as though it was written by a human being. This is particularly important for things like parser generators, where generated code tends to be rigid. For example - a generated LL(1) parser always matches greedy while a hand rolled recursive descent parser will switch between greedy and lazy consumption as the situation calls for it. The end result is more compact and more efficient. I don't recommend using parser generators because of the code they produce - unless you like source files in the hundreds of thousands of kilobytes or even megabytes. Code synthesis stands a good chance of changing that. Things like Github Copilot can produce code. Code synthesis seems like this functionality.
Check out my IoT graphics library here: https://honeythecodewitch.com/gfx And my IoT UI/User Experience library here: https://honeythecodewitch.com/uix
I am curious as well, but even more curious about AI 'improvements' for hardware. Exactly what are those improvements and how can they be used? Can I use the arithmetic logic unit (ALU) and memory unit to increase performance for non-AI algorithms? Similar to you, I am just too busy with regular work to delve into it.
