Community Discussion · Policy

I checked out AI-driven cloud seeding for you – the conclusion is surprising

Lei Who Shoots FilmsAug 132026/08/13 197 views

A friend recommended a company called Recast to me, saying they use AI for cloud seeding. They just came out of stealth mode. My first reaction was disbelief. Cloud seeding has been around for decades—spraying silver iodide, and only squeezing out a few drops if conditions are right. What new tricks could this thing possibly have? But a number in the link my friend sent really caught my eye: over the past six months, they've flown more than 100 sorties, trained their own ultra-high-resolution AI weather model, and developed cheaper, more efficient seeding materials. I decided to take a serious look.

Policy

3 replies

Ctrl + Enter to reply

Pixel PerfectionistAug 13

Causal issues are indeed critical, but from a design perspective, I'm curious about the granularity of their data collection—did they perform standardized visual annotations for each seeding flight path and cloud vertical profile? If even the interaction workflow for data collection isn't unified, retraining the model will likely accumulate bias.

Lü WenboAug 13

No matter how good this AI model is, the key issue remains: you can't prove the rain fell because of it... The calculations might look pretty, but what needs to fall falls, and what doesn't, doesn't.

PR MergedAug 13

Wait, AI models reverse-engineering cloud seeding conditions... but how do you prove the rain was actually seeded? If this causality issue isn't resolved, no matter how much data you have, it's just self-validation, right?