3 Comments
User's avatar
Leonidas Tam, PhD's avatar

Google is definitely counterpunching well with their version of Shipmas

Expand full comment
The AI Architect's avatar

Solid roundup. That Shopify SimGym concept is interesting becuase it sidesteps the core problem with A/B testing, wich is sample size. If the sims actually model customer behavior patterns well enough, the risk is everyone optimizes for the same local maxima. Real conversion bumps come from edge cases and tail behaviors that synthetic customers wont replicate unless they're trained on way more variance than current datasets.

Expand full comment