I also read nanosystems and Drexlers blog. The entire concept of nanotechnology is a series of mechanical catalysts.
A mechanical catalyst is a carved out niche in the catalyst. You use nano robotics to shove in the inputs—you probably need different bonding strategies—and you press together the multiple parts of the catalyst.
Internally the shape of the catalyst allows only one possible product, and by controlling the inputs you control what gets made. (vs solution in glassware in current chemistry)
Proteins as catalysts are in fact shapes held together by many forms of bonds. This is why they are so fragile and they have failure nodes where a misfolded proteins catalyzes peers to fail. Their floppiness also allows side products and limits the complexity of the molecules that nature is able to build.
You obviously need to use a strong material for a mechanical catalyst. Diamond or platinum, etc, and its covalently bonded.
This makes the side reactions happen less often, and it theoretically could allow for higher temperature operation.
I also disagree with Yudnowsky on the feasibility of advanced AI quickly developing nanotechnology. I suspect it would require a large scale effort and a large amount of intermediate steps that a fugitive or low resource ASI would be unable to accomplish. But I think he’s correct on this claim.
I also read nanosystems and Drexlers blog. The entire concept of nanotechnology is a series of mechanical catalysts.
A mechanical catalyst is a carved out niche in the catalyst. You use nano robotics to shove in the inputs—you probably need different bonding strategies—and you press together the multiple parts of the catalyst.
Internally the shape of the catalyst allows only one possible product, and by controlling the inputs you control what gets made. (vs solution in glassware in current chemistry)
Proteins as catalysts are in fact shapes held together by many forms of bonds. This is why they are so fragile and they have failure nodes where a misfolded proteins catalyzes peers to fail. Their floppiness also allows side products and limits the complexity of the molecules that nature is able to build.
You obviously need to use a strong material for a mechanical catalyst. Diamond or platinum, etc, and its covalently bonded.
This makes the side reactions happen less often, and it theoretically could allow for higher temperature operation.
I also disagree with Yudnowsky on the feasibility of advanced AI quickly developing nanotechnology. I suspect it would require a large scale effort and a large amount of intermediate steps that a fugitive or low resource ASI would be unable to accomplish. But I think he’s correct on this claim.