I just wanted to chime in and say that Serbia doesn't have the best reputation but is actually really nice. I just spent a couple of weeks in Belgrade and it's one of the most underrated places I've been to.
The most beautiful church I've ever been to in my entire life is there. There are tons of old gorgeous buildings (interspersed with ugly appartment buildings). It feels just as safe as any other European city, including for women. There's a ton of graffiti, but that's about it in terms of crimey-ness. (Don't know the stats. Just know how I felt while I was there).
Not saying it's necessarily the best spot for an AI safety hub (why not Budapest, Prague, or Lisbon, if you're optimizing for cheap and European time zones?), but wanted to help fight against the reputation of Serbia. Belgrade is gorgeous and interesting and would highly recommend people living there for a few months.
Oddly enough, I haven't found any really good resources on this. Except for this one google doc I found ages ago that I can't seem to find again. I think the explanation I give here might actually be the best I've seen.
But also, the explanation is pretty simple, so it's less about understanding (which is relatively easy) and more about practicing (which is harder, but I find, still way easier than concentration practice).
The only pieces of instruction I'd add are that if you're finding it hard to transfer the feelings of lovingkindness to a new object, that means the object is too hard. You should find an object that's easier. It's the equivalent of lifting weights and jumping up to a weight that's too heavy. Gradual increasing weight is key.
Linking to the original post is the most requested feature! Unfortunately because of obscure technical reasons it's quite difficult. We did to figure out a way to do it for our channels on Spotify and some of the sub channels across all podcast players, but not the main channel.
We definitely did not choose to not link to it due to worries about audience engagement. It's just because podcast players are astoundingly bad. Also our particular setup makes it challenging
If you like that documentary, you might like Up as well. It's a documentary that follows 14 different kids in the UK, starting at the age of 7, then showing what their lives are like every 7 years.
They tried to make it representative, but based on what they thought was important in 1964 England, so mostly based on class.
It's really fascinating. One guy becomes homeless and ends up being a politician. Another is really successful but feels terrible because all of his friends are even more successful. There's a more or less happy family that seems content with a pretty average life. Etc.
Not even close to representative of the world's sentient beings, but nevertheless, way more representative than I ever get talking to my social circle. Also really cool to get a longitudenal sense of a person, as opposed to a snapshot.
Babies. Similar documentary on Netflix, but focused on what being a baby is like in different countries.
Dollar Street. Website by Hans Rosling showing people's lives at different income levels and countries. People take photos of standardized things, like their beds, bathrooms, favorite items, etc. Really gives you a sense of how much of a difference it makes to go from $1 a day to $4.
Also, thanks for sharing this! I love these sorts of documentaries and am so going to watch it.
In most endeavors, you expect to receive many nos before receiving a yes (eg applying to schools, jobs, publishing papers/books, startups, etc). In EA it's common to receive one no and for people to give up.
I think this would only make sense if it was in a field where talent / value was easy to spot and evaluate and there were good feedback loops. But AI safety is far more like evaluating startup founders than evaluating bridge-builders.
Except even more difficult to evaluate, because at least with for-profit founders, you find out years later if they made money or not! With ethics, you can't even tell if you're going in the right direction!
If that's the case, we should have more evaluators, so that there's less people who slip through the cracks.
I just wanted to chime in and say that Serbia doesn't have the best reputation but is actually really nice. I just spent a couple of weeks in Belgrade and it's one of the most underrated places I've been to.
The most beautiful church I've ever been to in my entire life is there. There are tons of old gorgeous buildings (interspersed with ugly appartment buildings). It feels just as safe as any other European city, including for women. There's a ton of graffiti, but that's about it in terms of crimey-ness. (Don't know the stats. Just know how I felt while I was there).
Not saying it's necessarily the best spot for an AI safety hub (why not Budapest, Prague, or Lisbon, if you're optimizing for cheap and European time zones?), but wanted to help fight against the reputation of Serbia. Belgrade is gorgeous and interesting and would highly recommend people living there for a few months.
Oddly enough, I haven't found any really good resources on this. Except for this one google doc I found ages ago that I can't seem to find again. I think the explanation I give here might actually be the best I've seen.
But also, the explanation is pretty simple, so it's less about understanding (which is relatively easy) and more about practicing (which is harder, but I find, still way easier than concentration practice).
The only pieces of instruction I'd add are that if you're finding it hard to transfer the feelings of lovingkindness to a new object, that means the object is too hard. You should find an object that's easier. It's the equivalent of lifting weights and jumping up to a weight that's too heavy. Gradual increasing weight is key.
Aw, thanks! And updated the link. :)
The Slack invite only works if you have a @ea-maastricht.org email address. Is there a link for people who don't have that?
Linking to the original post is the most requested feature! Unfortunately because of obscure technical reasons it's quite difficult. We did to figure out a way to do it for our channels on Spotify and some of the sub channels across all podcast players, but not the main channel.
We definitely did not choose to not link to it due to worries about audience engagement. It's just because podcast players are astoundingly bad. Also our particular setup makes it challenging
The URL only for some of the sub channels on some of the platforms. But always the title, author, and source.
Yeah. We do say at the beginning of every episode the title, author, and where to find it, and it's in the show notes, but not a link.
It does have a link on the sub-channels on Spotify, because for some weird arcane technical reasons, that was fine.
Aw. Thank you!
If you like that documentary, you might like Up as well. It's a documentary that follows 14 different kids in the UK, starting at the age of 7, then showing what their lives are like every 7 years.
They tried to make it representative, but based on what they thought was important in 1964 England, so mostly based on class.
It's really fascinating. One guy becomes homeless and ends up being a politician. Another is really successful but feels terrible because all of his friends are even more successful. There's a more or less happy family that seems content with a pretty average life. Etc.
Not even close to representative of the world's sentient beings, but nevertheless, way more representative than I ever get talking to my social circle. Also really cool to get a longitudenal sense of a person, as opposed to a snapshot.
Can watch it on youtube
You also might like:
Also, thanks for sharing this! I love these sorts of documentaries and am so going to watch it.
Good question! So, that's important, but I'm less worried about this because:
In most endeavors, you expect to receive many nos before receiving a yes (eg applying to schools, jobs, publishing papers/books, startups, etc). In EA it's common to receive one no and for people to give up.
I think this would only make sense if it was in a field where talent / value was easy to spot and evaluate and there were good feedback loops. But AI safety is far more like evaluating startup founders than evaluating bridge-builders.
Except even more difficult to evaluate, because at least with for-profit founders, you find out years later if they made money or not! With ethics, you can't even tell if you're going in the right direction!
If that's the case, we should have more evaluators, so that there's less people who slip through the cracks.
I discuss something similar in another comment thread here.