Try swapping in “Biosecurity” or “nuclear proliferation” as established important-but-wonky topics. We’re in a bubble for thinking about AI stuff regularly and I think it’s easy to overestimate how much other people are going to search for these relatively abstract phrases, even as they worry about jobs and other more concrete things

‍

Out of 4 times that "AI Alignment" has exceeded the global minimum of "Light Pollution", three have happened since mid-March. Plenty of time for stars to align either in 2023 or 2024 IMO.

Biggest worry: What's up with the seemingly relatively consistent rise of Light Pollution in 2023? If this is the new baseline then it'll be much harder to beat than the old "global minimum" of Light Pollution.

@HenriThunberg global minimum looks like it was on Christmas

the one day when light is considered holy, and not a polluting nuisance

@Lily So everybody who bet yes needs to create a global meme connecting Christmas day to “AI alignment”, to get a timely spike. If you can pull that off, I won’t even be mad to lose

if public interest in this spikes, I think "AI safety" is much more likely to catch on as the mainstream term than "AI alignment"

@Lily "AI alignment" is a search 'topic' on Google Trends. It's a collection of search terms related to the same thing. Trends doesn't say what search terms are included in a topic, but it does kind of look like "AI safety" is not included, from this

https://trends.google.com/trends/explore?date=today%201-m&geo=US&q=%2Fg%2F11bzrg63f7,Ai%20safety&hl=en

@HarlanStewart interesting, wonder what (if any) category "AI safety" would fall under, if not alignment - doesn't look like the more left-coded "AI ethics" is a topic either

@Lily I agree but I don't think this is about what term "catches on". Since it only has to be for a single day I think the most likely scenario is a public figure/major publication uses the phrase which causes a bunch of people who don't know what it is to go and look it up.

Google Trends only aggregates on a weekly basis, not daily?

My p(daily) is significantly higher than p(weekly).

@TobiasHaeberli there's an option for the past hour, past 4 hours, past day, past week, past month, etc etc. So each day's results will be a part of tomorrow's aggregate

@TobiasHaeberli also if you want, you can just download the data... as a csv, and it will allow you to see each individual day's data

@firstuserhere huh, weird. I don't have this option and the .csv was aggregated on a weekly basis.

@TobiasHaeberli top right (1st of the 3 options)

@firstuserhere Oh yeah, I'm stupid and got confused with the date format. Thank you!

@TobiasHaeberli yeah no worries!