The US National Weather Service releases updated 84-hour forecasts every 6 hours. Even with supercomputers at their disposal, due to the computational complexity of simulating physics, that is their best possible effort.
Google, meanwhile, is “developing a machine learning model that it says can accurately predict weather in seconds – not hours – and outperforms 90% of the targets used by the world’s best weather prediction systems.” Using a single desktop computer, they can generate a highly accurate 10-day forecast in under a minute.
Yes. Search generally pulls data from databases. It doesn’t compute weather forecasts. The addition of AI results is net addition computation. In the worst case scenario where the generation of the AI results happens on-the-fly, that’s a lot more computation. I’m sure they pre-compute a lot of them so they’re not in the worst case scenario. However in the best case scenario they still have to do this new additional heavy (check LLM compute usage) computation once per result. So the profit margin for search is very likely lower than it used to be when isolating for this variable. If they’re somehow increasing their revenue from these results, that’s another variable that might offset it. I’ve no idea. What I’m certain about is the cost is higher after AI results were introduced because more energy is used.
The US National Weather Service releases updated 84-hour forecasts every 6 hours. Even with supercomputers at their disposal, due to the computational complexity of simulating physics, that is their best possible effort.
Google, meanwhile, is “developing a machine learning model that it says can accurately predict weather in seconds – not hours – and outperforms 90% of the targets used by the world’s best weather prediction systems.” Using a single desktop computer, they can generate a highly accurate 10-day forecast in under a minute.
More information:
https://www.weforum.org/stories/2023/12/ai-weather-forecasting-climate-crisis/
Given this information, and given the enshittification of Google search, would you still make the same guess about their profit margins?
Yes. Search generally pulls data from databases. It doesn’t compute weather forecasts. The addition of AI results is net addition computation. In the worst case scenario where the generation of the AI results happens on-the-fly, that’s a lot more computation. I’m sure they pre-compute a lot of them so they’re not in the worst case scenario. However in the best case scenario they still have to do this new additional heavy (check LLM compute usage) computation once per result. So the profit margin for search is very likely lower than it used to be when isolating for this variable. If they’re somehow increasing their revenue from these results, that’s another variable that might offset it. I’ve no idea. What I’m certain about is the cost is higher after AI results were introduced because more energy is used.