Microsoft is using AI to replace all their C/C++ code with Rust by 2030. Is that actually possible? What could go wrong? Microsoft’s strategy relies on a new "North Star" metric: 1 engineer, 1 month, 1 million lines of code. 😱
Rust is good, but this approach seems incredibly difficult. Given the risk of AI hallucinations, aiming for a million lines of code per developer feels like an 'Apollo-program' level of difficulty and an excessive task 😱
@nixCraft Everything is stupid now 🤦♀️
@nixCraft I'm not a coder, but I know enough not to interrupt the enemy when he is making a mistake.
A bit of context:
Galen’s group is responsible for some interesting work in languages and memory safety. They did Singularity and Checked-C. They have a long track record in moving things that were traditionally done in unsafe languages over to safer ones.
They are currently in an environment where there is no headcount or funding available for anything unless it says it is AI. Lying to management is the core of Microsoft culture,
There’s a reasonable chance that this is actually looking for someone to do interesting things and lie to management that ‘AI’ did it. If it works, it will be branded as ‘AI’ and Galen will get a bonus. If it doesn’t work, it’s research but it tried AI in an ambitious setting, so Galen will get a bonus.
@david_chisnall @nixCraft "They are currently in an environment where there is no headcount or funding available for anything unless it says it is AI. Lying to management is the core of Microsoft culture,"
That's not specific to AI and/or Microsoft, it's how the world has always worked.
Central policy makers say that money is only available for things that accord with their policy. Local decision makers have to work out how to pretend that that the things they actually need to do match policy.
For example, ask any councillor how much effort their officers have to expend on writing justifications for how the council's activities supposedly match the criteria for the central government funding pots.
