2 ms·
> N period on from now, software engineers who haven't adopted or started exploring software assistants, are frankly not gonna make it. Engineering organization
by ben_w 2y ago
> N period on from now, software engineers who haven't adopted or started exploring software assistants, are frankly not gonna make it. Engineering organizations are right now are split between those who have had that "oh fuck" moment and are leaning into software assistants and those who have not.
Mine was InstructGPT, a few months before the public release of ChatGPT. Everything that followed was an unsurprising extrapolation of what I saw in that moment.
Question is, how far will it continue? That I cannot say. At least a factor of 3 from foreseeable compute improvements… unless a war is started over the manufacture of compute. But also, Koomey's law (energy) is slower than Moore's (unit cost + feature size), and if one presupposes that manufacture of compute is unbounded, then we rapidly hit limits of manufacture of power sources for the compute, and that in turn causes huge economic issues well before there's enough supply to cause mass unemployment… unless there's a lot of huge algorithmic improvements waiting to happen, because that might be enough to skip electricity shortages and go straight to "minimum and cheapest calories to keep human alive are more expensive than the electricity to run the servers that produce economic value equal to the most productive human".
I have no confidence in any forecast past the end of 2028.