Engineering via search seems less novel than you indicate. The novelty lies more in the automation of search.
Selective breeding of plants and animals for food has been one of the most important classes of engineering projects over the past 10,000 years. It has involved so little design that I suspect most of the engineers were unaware that they were doing engineering until the past century or so.
The smallpox vaccine was developed before any clear concept of a virus, so it seems to have involved more search than design.
Drug development still seems to involve more search than design. That causes major problems for figuring out whether we should trust a drug. For a number of diseases, it leaves doctors with a need to perform a search over multiple drugs for each patient.
Engineering of new institutions is often too hard to do via design. I would summarize Eliezer’s description of CFAR’s initial steps as “we tried a design-heavy approach, it failed, we switched to a search-heavy approach”. The book Seeing Like a State documents more expensive versions of this pattern.
See also Henrich’s The Secret of Our Success for arguments that for several million years, humans have been advancing via technologies that were beyond their ability to design.
Engineering via search seems less novel than you indicate. The novelty lies more in the automation of search.
Selective breeding of plants and animals for food has been one of the most important classes of engineering projects over the past 10,000 years. It has involved so little design that I suspect most of the engineers were unaware that they were doing engineering until the past century or so.
The smallpox vaccine was developed before any clear concept of a virus, so it seems to have involved more search than design.
Drug development still seems to involve more search than design. That causes major problems for figuring out whether we should trust a drug. For a number of diseases, it leaves doctors with a need to perform a search over multiple drugs for each patient.
Engineering of new institutions is often too hard to do via design. I would summarize Eliezer’s description of CFAR’s initial steps as “we tried a design-heavy approach, it failed, we switched to a search-heavy approach”. The book Seeing Like a State documents more expensive versions of this pattern.
See also Henrich’s The Secret of Our Success for arguments that for several million years, humans have been advancing via technologies that were beyond their ability to design.