I think I figured out exactly how the “AI used a bottle of water per prompt” miscalculation happened open.substack.com/pub/andymasl...
Who's Who of AI
Andy Masley
When the going gets weird the weird turn pro.
Researching and writing about AI and other topics from a broadly effective altruist perspective. On a grant from Coefficient Giving.
andymasley.com
What they're sharing
open.substack.com
blog.andymasley.com
Articles & links
Just posted the first of a few posts on what I see as the history of the data center panic, this first one is on what was happening pre-ChatGPT blog.andymasley.com/p/a-history-...
Their own posts
Recent commentary
I've had a very nice back-and-forth email exchange with Shaolei Ren, who provided the original "AI uses a bottle of water per prompt" estimate with the Washington Post. (1/5)
TIL that everyone calling me a paid shill for defending AI water use is actually secretly funded by effective altruists. Nice try EAs but I'm onto you
To be clear I’m very happy we have an AI safety pope
There's a funny inconsistency in what people believe about how much energy AI uses: "AI uses 10x as much energy as a Google search" ✅ "10 Google searches use as much energy as a few seconds of Netflix" ✅ "AI prompts use as much energy as a few seconds of Netflix" ❌ Must be a lie, must be a shill
I followed EA for 7 years before getting more involved in the actual scene in DC 6 years ago. Over that time, I think Roko's Basilisk has come up maybe twice, both as jokes. This is not a serious part of EA thinking or what people are actually worried about with AI.
AI Czar Masley is taking a much slower approach to the data center buildout and is conspicuously only approving new construction atop Confederate grave sites
Weird aspect of Gebru's definition of AGI is it just makes AGI definitionally impossible. She says that normally in engineering when you're building something, you have clearly designed tests for inputs and outputs. "When you're building a bridge, you know what the bridge is supposed to do" (1/3)
If your AI safety plan involves summoning the Shrike to teach lessons to people it's probably not tractable