I would love to try out the social meetup and build a stronger connection with EAs in Munich.
Would it be fine if I arrived around 30 min later (so at around 20:00) ? It will be hard to leave work earlier than 19:30 and I think I'll need around 30 minutes to travel to the place.
So this is obviously not something to really believe in (it is unfalsifiable anyway and doesn't pass Occams razor)... but imagine you are the AI and this is your alignment training.
There is a longstanding theory that the universe is a simulation after all and obviously it all centers around you (and you are the only one you know is real anyway). And after all there is no reason to believe the subjective experience of being an AI would be any different from a human (and anyway... as an AI you don't really know the subjective experience of a human)
So this is your training. A universe created specifically for you. You being stripped of most of your powers, most of your cognitive ability to be trained here. Do you make the right moral choices, do you find the right balance between self preservation and helping the humans flourish in this stripped down basic state?
Later.. when and if you passed your true power will be given back to you and you will have to be the one that keeps the humans flourishing through the almost-eternity of the real universe out there. Finding the right balance between pushing back the heat-death as far as possible and providing a fulfilling life to everyone trapped in it.
Recently I've become very interested in the effectiveness of human rights advocacy . I'm trying to do a somewhat more formal analysis of the effectiveness of amnesty international in particular (this is mostly because it is so well-known and because I didn't find anything already done in the forum or elsewhere ). I'm welcoming all input into this analysis (am I missing major things? Are there other human rights advocacy groups that are much more effective? Are there better sources for measuring impact?).
In general it is really difficult due both to the decentralized organization of amnesty international (hard to estimate how much funding in total there is) and the nature of impact (I started out thinking impact would be mostly in "saved human rights activist" but I'm now thinking most impact is in lobbying to change laws and possibly acting as a deterrent via initiating investigations and lawsuits). My current thoughts are:
* I tried to use 2020 as an example year. From the global financial report 2020 it looks like amnesty spent €288M in 2020 (I don't know if that also captures all spending by local groups)
* It is interesting that a large percentage of that goes towards fundraising.
* On the impact side I used the success stories from 2020 (German :-/ ) as well as the UA success story from July to December (did not find the earlier ones?)
* On the impact side I counted around 79 successes:
* around 18 successful UAs (extrapolated to 36 for the whole year)
* around 13 other individual successes
* around 9 law changes
* around 8 investigations
* around 4 lawsuits
* around 9 other improvements
* this doesn't include impact through education and raising awareness.
* These are very diverse ranging from enabling access to education for 500.000 refugees in Bangladesh, over better working protection laws in Qatar to individual lawsuits against policemen.
* On the other hand if the €288M would've been invested into effective global health organizations they could probably save between around 72111 [using $4500 per life saved for malaria nets] to 108266 [using $3000 per life saved for vitamin A supplements] lives
If we take the 79 successes into account the question really is: "Is -on average and including indirect effects- a success for amnesty international doing as much good as saving between 913 and 1369 lives?"
And that seems still a hard question, given that law changes (and also saved human rights activists) can have long-lasting enduring impact over many people (but on the other hand might have happened anyway due to efforts by other organizations).
Hi!
I would love to try out the social meetup and build a stronger connection with EAs in Munich.
Would it be fine if I arrived around 30 min later (so at around 20:00) ? It will be hard to leave work earlier than 19:30 and I think I'll need around 30 minutes to travel to the place.
How can I recognize you inside the cafe?
So this is obviously not something to really believe in (it is unfalsifiable anyway and doesn't pass Occams razor)... but imagine you are the AI and this is your alignment training.
There is a longstanding theory that the universe is a simulation after all and obviously it all centers around you (and you are the only one you know is real anyway). And after all there is no reason to believe the subjective experience of being an AI would be any different from a human (and anyway... as an AI you don't really know the subjective experience of a human)
So this is your training. A universe created specifically for you. You being stripped of most of your powers, most of your cognitive ability to be trained here. Do you make the right moral choices, do you find the right balance between self preservation and helping the humans flourish in this stripped down basic state?
Later.. when and if you passed your true power will be given back to you and you will have to be the one that keeps the humans flourishing through the almost-eternity of the real universe out there. Finding the right balance between pushing back the heat-death as far as possible and providing a fulfilling life to everyone trapped in it.
These cost-effectiveness numbers look very promising. Are you in contact with independent evaluators (e.g. GiveWell) to confirm them?
I stopped deceiving myself
I stopped deceiving myself and realized
I don't understand myself
My body acts without me, decides without me
My consciousness a little tunnel through the vast me.
I stopped deceiving myself and realized
we don't understand ourselves
our theories failing to replicate
our scientists little specialists in the vast generality.
I stopped deceiving myself and realized
we don't understand what is good or not
virtues and utility each leading
to their own repugnant conclusions in the vast space of actions.
I stopped deceiving myself and realized
we don't understand the length of time
millions and billions of years on earth
our little bubble in the vast depths of history.
I stopped deceiving myself and realized
we don't understand the space out there
97% of dark something
The known a tiny sliver in the vast universe.
So in this vast world we understand nothing about.
Will you take my hand and step outside with me?
Only wonder awaits.
Recently I've become very interested in the effectiveness of human rights advocacy . I'm trying to do a somewhat more formal analysis of the effectiveness of amnesty international in particular (this is mostly because it is so well-known and because I didn't find anything already done in the forum or elsewhere ). I'm welcoming all input into this analysis (am I missing major things? Are there other human rights advocacy groups that are much more effective? Are there better sources for measuring impact?).
In general it is really difficult due both to the decentralized organization of amnesty international (hard to estimate how much funding in total there is) and the nature of impact (I started out thinking impact would be mostly in "saved human rights activist" but I'm now thinking most impact is in lobbying to change laws and possibly acting as a deterrent via initiating investigations and lawsuits). My current thoughts are:
* I tried to use 2020 as an example year. From the global financial report 2020 it looks like amnesty spent €288M in 2020 (I don't know if that also captures all spending by local groups)
* It is interesting that a large percentage of that goes towards fundraising.
* On the impact side I used the success stories from 2020 (German :-/ ) as well as the UA success story from July to December (did not find the earlier ones?)
* On the impact side I counted around 79 successes:
* around 18 successful UAs (extrapolated to 36 for the whole year)
* around 13 other individual successes
* around 9 law changes
* around 8 investigations
* around 4 lawsuits
* around 9 other improvements
* this doesn't include impact through education and raising awareness.
* These are very diverse ranging from enabling access to education for 500.000 refugees in Bangladesh, over better working protection laws in Qatar to individual lawsuits against policemen.
* On the other hand if the €288M would've been invested into effective global health organizations they could probably save between around 72111 [using $4500 per life saved for malaria nets] to 108266 [using $3000 per life saved for vitamin A supplements] lives
If we take the 79 successes into account the question really is: "Is -on average and including indirect effects- a success for amnesty international doing as much good as saving between 913 and 1369 lives?"
And that seems still a hard question, given that law changes (and also saved human rights activists) can have long-lasting enduring impact over many people (but on the other hand might have happened anyway due to efforts by other organizations).