- https://karrisaarinen.com/posts/our-organizations-are-being-ddosed-from-the-inside/
- https://elliotmorris.net/prototypes-are-not-that-helpful-anymore
Two piece that came up in a feed today.
TLDR: LLMs are enabling so much production (good quality production in engineering settings) that it is overwhelming management’s ability to manage.
Management cannot be automated. Understanding cannot be outsourced. Training will be improved. It has to become faster to produce and more efficient to learn. There are hard limits to this.
The difficulty of determining relevance was noted by AI pioneer Marvin Minsky in the 1970s, he called it the frame problem then. Back then they were working with expert systems. Neural networks solved some of it at the low level, but the higher level stuff cannot be abstracted away. Relevance is both biologically and philosophically determined. AI systems have no purpose or needs of their own so this stuff is going to have to be dealt with. It is political and even religious too. Depending on the issue, determining relevance can take a lot people interacting over what can be extended periods, even centuries. Consensus around relevance an emergent feature of an information processing network that is not unlike the brain: a human society.
The leftists who take their morality as a human universal must not be allowed to blithely build it into AI as the work of “professional” ethicists and philosophers. Engineers cannot be allowed to outsource their personal philosophical work to leftists as they seem prone to do. These things are at the core what it means to be living in a human society. Everyone must have the right to both hold a view on these matters and express it.
Much more conflict ahead.