I am a researcher at Coefficient Giving where I work on biosecurity and pandemic preparedness, and also think about other transformative technologies that powerful AI systems might enable.
djbinder
Karma: 200
The AI Industrial Explosion — Part 3: Going faster
The AI Industrial Explosion — Part 2: Transition Dynamics
The AI Industrial Explosion — Part 1: Maximum growth rates with current production methods
Thanks for the suggestion! No added a note that I did mostly write this, but put everything in a Opus 4.6 block to avoid spending time optimizing again Pangram.
Will whole brain emulation matter for the AI transition?
Hi—thanks for engaging with this! These points are discussed within the Technical Report on Mirror Bacteria, in Chapter 1 (which reviews opposite-chirality nutrient use) and Chapter 8 (which discusses predation specifically). Predation requires much more than simply the ability to catabolize small metabolites. From §8.5 (p. 184):
The evolution of protists and animals that prey on mirror bacteria appears more challenging. Protists and animals lack the enzymes required to degrade mirror proteins, sugars, nucleic acids, and lipids. They likely have limited or no ability to catabolize most mirror metabolites (Friedman & Levin 2012), and some D-amino acids are toxic to many organisms (Forsum et al. 2008; Friedman & Levin 2012; Yow et al. 2006; see also Box 1.2). These deficiencies could not be remedied through a handful of mutations, but would require the much slower evolution of novel proteins and catabolic pathways.
To give an illustrative example, C. elegans relies on many dozens of distinct lysozymes, glycosidases, proteases, phospholipases, nucleases, and other lytic enzymes to digest the macromolecules present in their bacterial prey (McGhee et al. 2007; Yilmaz & Walhout 2016). Digesting even a simple macromolecule like bacterial-derived glycogen requires intestinal amylase and α-glucosidase. Evolving a similar catabolic pathway for mirror glycogen would almost certainly require a similar or greater number of novel enzymes, and it seems questionable whether such adaptations could arise in nematodes even over millions of years.
For example, having a racemase that interconverts a D-amino acid into an L-amino acid isn’t enough, you also have to first breakdown mirror proteins into their constituent D-amino acids and then import the D-amino acids into the cells. Even beyond that there are a few challenges to overcome: D-amino acids can be toxic, and also racemases catalyze the interconversion of L/D-amino acids, so that would have to be carefully regulated to avoid causing a build-up of D-amino acids in the cell.
As a more minor point, the lipid membranes of bacterial cells don’t contain triglycerides (indeed, triglycerides are not capable of forming lipid membranes, rather, they are primarily used as specialized energy storage, usually in multicellular organisms; triglycerides are entirely absent from E. coli and most other bacterial species.) The constituents of bacterial lipid membranes are varied but chiral.
It’s worth distinguishing “how fast do things grow given the stuff around today?” vs “how fast could things grow with today’s technology but the stuff optimized to grow fast?” This post is about the latter, and there I don’t really see how semiconductor production would be the bottleneck; it’s just not that large a fraction of the physical capital required even with pessimistic assumptions about cost, and you can compensate for longer lead times by increasing sector production.
In the next post I look at the transition to rapid growth, planning to post sometime in the next week or so. Given my results there I’m skeptical of a semiconductor bottleneck, but I haven’t investigated in detail so it’s possible I’m missing them because they are hidden within a larger sector (this should be easy to fix though, I’ll add to my to do list).
Note I am assuming that the R&D required to create AI and robotic workers has been done and costs are comparable to the human brain wrt compute requirements. Obviously if compute costs are 1000x higher because AI systems aren’t very efficient that could be a strong drag on economic growth, but I expect that won’t last very long as inference costs continue to fall and the R&D needed for the transition gets done.