The archive's propositions

Every page here is generated from the archive. Counts and quotations are read from the archive when the page is built, and a quotation the archive cannot supply fails the build rather than rendering something plausible.

• A coordinated slowdown is worth pursuing even at the cost of advantage, because uncoordinated speed is the greater risk.
  6 assert · 5 deny
• Safety and innovation trade off against one another: what makes a model safer makes it less capable or less widely available, so pushing on one goal costs ground on the other.
  4 assert · 7 deny
• Open models damage the business case of closed models by putting downward pressure on prices.
  3 assert · 2 deny
• Open models move the binding scarcity to hardware — compute, chips and the networking between them.
  3 assert · 2 deny
• Building national AI capacity on sovereign infrastructure is sound policy for smaller countries, not merely a way to sell deployments.
  14 assert · 2 deny
• Releasing model weights raises the risk of catastrophic misuse, because guardrails cannot be applied and usage cannot be monitored once the weights are out.
  2 assert · 4 deny
• Open models keep AI continually available, rather than dependent on any one provider's decisions.
  13 assert · 1 deny
• Open weights commoditise the model layer, moving durable value to distribution, data, and the infrastructure underneath.
  5 assert · 2 deny
• Open weights prevent AI from being controlled by a few actors, because anyone can run them.
  10 assert · 1 deny
• Architectural and training efficiency is what lets a latecomer reach the frontier, so scarce compute is not disqualifying.
  4 assert · 1 deny
• Compute, not algorithms or talent, decides who can compete at the frontier.
  6 assert · 1 deny
• Cheaper and more available models increase total consumption of intelligence rather than displacing it, and the volume creates economies of scale.
  6 assert · 1 deny
• Open models strengthen the country that hosts them, so restricting their release weakens the very position it means to protect.
  5 assert · 1 deny
• Capable models should be required to pass pre-release testing, binding open weights and closed APIs alike.
  6 assert · 0 deny
• The claim that advanced AI poses a catastrophic risk is overblown, and the fear does more damage than the technology.
  5 assert · 0 deny
• AI regulation in practice concentrates power in the largest labs rather than constraining them.
  2 assert · 1 deny · watched, thin evidence
• Distillation from closed frontier models, not independent invention, is what let Chinese labs close the distance to the frontier.
  2 assert · 1 deny · watched, thin evidence
• Demand for inference, not training runs, is what drives AI infrastructure spending and its economics.
  3 assert · 0 deny · watched, thin evidence
• Chinese labs release weights because openness is their competitive strategy against closed frontier labs.
  3 assert · 0 deny · watched, thin evidence
• Restricting chips and stopping industrial-scale distillation is the effective way to slow a rival's frontier, and licensing models is not.
  2 assert · 0 deny · watched, thin evidence