고정된 트윗
The second stage of the AI economy is here: Bottlenecks and Red Queen Races
- Datacenter buildout is an infinite sector that will extend to the Dyson spheres somewhere in the future
- The ramp creates shortages in many areas, one by one, first in chips, then in memory, electricians , photonics etc.
- Each bottleneck cause an increase in prices
- then capacity in that area has to be ramped over 2-5 years
- The first Red Queen Race is in cybersecurity
- Frontier models have to be used for defense. Waiting till open source exposes you to hackers who also get open source tools at the same time
- there will be other winner take most markets , from financial trading to cola marketing
- this is likely where the surplus profits
from AI will be lost. Many companies have no choice but to engage in these zero sum games.
- this is how the model companies will come to dominate the economy
- in search Google and Facebook became rent collector, one on all information request, the other on social information
- AI is now a pay-to-win game. If you do not pay, catastrophe may strike.
Prakash@8teAPi
looks like we have an answer on what the early AI economy will look like: - deflationary - unemployment increases - Fed decreases rates - cost of debt is cheaper - borrowing increases - stock market zooms upwards - companies keep building AI - creates some jobs at much higher productivity - best companies go from $1 mil/employee to $10 mil/employee revenue - more AI building -> more job turnover, with smaller job creation echo Challenge is to expand the hyper productive job creation fast enough to absorb the job turnover. I guess that’s where the Fed will come in running the economy hot.
English
