I am concerned at reports of the enormous energy use by AI systems. How much energy does iNat use and is it sourced from renewable energy?
This was answered a few years ago. iNat’s no longer at Cal Academy, but the training machines we use to train our models are in a place that claims to be 100% sourced from renewable energy. Training does take up a fair amount of energy, but unlike querying LLMs (eg Chat GPT), querying the iNat model for ID suggestions is very low energy.
Here’s an answer from Alex Shepard, one of our engineers, about energy used for training a natural language search model: https://www.inaturalist.org/comments/15470892
Thanks for your reply, which is somewhat reassuring, but what about the energy required just to store all the images?
Some hopeful news from the ‘think smarter’ side of the data power management side:
https://www.livescience.com/technology/computing/surprisingly-simple-coding-trick-can-slash-data-center-energy-usage-by-30-percent
Photos and sounds are stored on servers from Amazon Web Services, you can read about their sustainability efforts here.
As to how much energy is used to store and serve iNat’s photos and sounds, I’m not sure that’s really calculable. I believe those servers are also in Washington State, which derives most of its energy from hydroelectricity.
You haven’t provided any insight as to what you’d find reassuring, so I’m not sure what you’re looking for, but any sizable website will incur energy usage. I’d imagine that whatever energy iNat photo and sounds storage and bandwidth consume, it’s significantly less than platforms like Instagram, YouTube, Facebook, Netflix, etc.