3 ms·

Inventing programming languages is super-fun, but aren't we almost post programming language? In 30 years, do we really expect to be programming in some machin

Inventing programming languages is super-fun, but aren't we almost post programming language? In 30 years, do we really expect to be programming in some machine-human intermediate language? I hope that we'll be having an open ended conversation with the machines to iterate our intent, with working software at most stages along the way.

dragonwriter 5y ago
> In 30 years, do we really expect to be programming in some machine-human intermediate language? Just as when people asked this 30 years ago, yes, though from our perspective today the tooling may seem near magical.
JabavuAdams 5y ago
Perhaps more modestly - how far are we from just being able to translate from one high level language to another, assuming the ability to refine the translation as opposed to a one-shot job that has to correctly interpret the intent? I.e. will it matter what language anyone uses? Here, take this software renderer I wrote in Python, and translate it to C++. "Conversation" about choices and ambiguities ensues...
- machiaweliczny 5y ago
  I think we soon will have NLP capable of that. Too much money involved to resist.
- nahuel0x 5y ago
  Maybe there is the possibility of a meta high level language capable of describing/encompassing all the others without exceeding human cognition capabilities. Instead of compiling to a common low level target like assembler, transpiling all the existing sources to a very high level one optimized for human understandability.
CodeGlitch 5y ago
> In 30 years, do we really expect to be programming in some machine-human intermediate language? I hope that we'll be having an open ended conversation with the machines to iterate our intent, with working software at most stages along the way. This is actually entirely feasible, although perhaps not open-ended conversation. Instead it could be that you feed the machine training data which describes how to solve a particular task, and the output is a program that solves that task. So we would all end up being ML engineers. Scary I know...

dragonwriter 5y ago

> In 30 years, do we really expect to be programming in some machine-human intermediate language? Just as when people asked this 30 years ago, yes, though from our perspective today the tooling may seem near magical.

JabavuAdams 5y ago

Perhaps more modestly - how far are we from just being able to translate from one high level language to another, assuming the ability to refine the translation as opposed to a one-shot job that has to correctly interpret the intent? I.e. will it matter what language anyone uses? Here, take this software renderer I wrote in Python, and translate it to C++. "Conversation" about choices and ambiguities ensues...

machiaweliczny 5y ago
I think we soon will have NLP capable of that. Too much money involved to resist.
nahuel0x 5y ago
Maybe there is the possibility of a meta high level language capable of describing/encompassing all the others without exceeding human cognition capabilities. Instead of compiling to a common low level target like assembler, transpiling all the existing sources to a very high level one optimized for human understandability.

CodeGlitch 5y ago

> In 30 years, do we really expect to be programming in some machine-human intermediate language? I hope that we'll be having an open ended conversation with the machines to iterate our intent, with working software at most stages along the way. This is actually entirely feasible, although perhaps not open-ended conversation. Instead it could be that you feed the machine training data which describes how to solve a particular task, and the output is a program that solves that task. So we would all end up being ML engineers. Scary I know...