Posts
Comments
This course looks excellent! I'm finishing up my ML master's degree in a few weeks, so it comes at a perfect time for me to get into AI safety in the interim period between graduation and starting at a job.
I'm a lifelong lurker on all things social media, which is something I'd like to change. In the past 20 minutes I've made various comments (Reddit, Steam review, etc.), sent some messages to strangers and wrote a Tweet. It may not be much, but it's more than I've done in the whole past year. I hope to continue this and grow into someone who enjoys contributing and interacting in online communities.
Thank you for this post!
Like someone else mentioned, this technique could work for other things. I don't use Facebook. I used a slightly modified version of this technique on my Youtube watch history and came to a ratio of about 1:20 (bad videos do tend to be shorter, so time wise it's probably 1:10 or so).
I've been trying to reduce my usage for a while, with moderate success. According to RescueTime I spend 32 hours on youtube.com in the past 30 days, and 49 hours in the 30 days before that. I'll keep in mind the poor good:bad ratio from now on and I'll report back in another 30 days with the new figure from RescueTime.
If it's under 25 hours I consider it a moderate success, under 20 a huge success.
EDIT:
Results are in, just 15 hours for the past 30 days. It seems to have worked quite well for me.
Suppose, as you say, some of this nonlinearity is already factored into the 70% estimate, that would imply that the 'real' number is even higher. For some interaction, like having a face to face conversation without any protection, the probability of an infection may have increased by 100% or even more.
I'm also not an expert. Intuitively this seems like a big step with just a handful of mutations.
rtnew=1.7 is an entirely different case. Suppressing it would require the sort of lockdown that would yield rt=0.6 for the old strain,
Is this valid reasoning? Intuitively I'd expect current measures to be more effective for a more infectious strain, so that it would require a lockdown that would yield something closer to 0.8 for the old strain.
I suspect that some measures will remain close to 100% effective, like not seeing family members and friends. Similarly some things physically can't become more infective, like being infected by your partner.
In that case I would expect that what currently are very high risk interactions (e.g. partner) and very low risk interactions (e.g. masked, 6ft+ grocery shopping) will still be almost as risky for a more infectious strain.
However if the bulk of infections happen because of moderate risk interactions, then moderate risk interactions becoming high risk interactions would significantly increase overal cases.
I wonder if the sudden increase is not just the result of some holiday or cold weather some time before. What are the chances that a new strain would dramatically increase daily cases in two countries within a few days of each other. Notably it started to increase in the Netherlands a few days before the UK. If anything this would point to it coming from a third country, yet it would still be odd that the outbreaks progress roughly in the same manner.