*loads* of people saw the title and thought “oh, this is a book about how AI is Good, Actually”. For anyone who doesn’t know, the full quote is Eliezer’s: “The AI does not hate you, nor does it love you, but you are made out of atoms which it can use for something else.”. I much preferred the old title but I guess I shouldn’t be surprised people didn’t get it!
Paul_Crowley
Karma: 136
“ultimately I made offers to two candidates both of which I had had strong gut feelings about very early, which was rewarding but also highly frustrating.”—I hope this comment doesn’t come across as incredibly mean, but, are you getting that from notes made at the time? When I find myself thinking “this is what I thought we’d do all along”, I start to suspect I’ve conveniently rewritten my memories of what I thought. Do you have a sense of how many candidates you had similar strong positive gut feelings about?
Thank you for a very helpful comment!
When I applied to Google I did a phone interview and a full day of in-person interviews, plus a 1-hour conference call about how to do well in the second round. Lots of people devote significant time brushing up their coding interview skills as well; I only didn’t because things like Project Euler had brushed up those skills for me.
Of course, the one who writes the post about it is likely to be the outlier rather than the median.
If you can’t afford it, doesn’t that suggest that earning to give might not be such a bad choice after all?
Could you comment specifically on the Wayback Machine exclusion? Thanks!
Nitpick: “England” here probably wants to be something like “the south-east of England”. There’s not a lot you could do from Newcastle that you couldn’t do from Stockholm; you need to be within travel distance of Oxford, Cambridge, or London.
You have a philosopher’s instinct to reach for the most extreme example, but in general I recommend against that.
There’s a pretty simple counterfactual: don’t take or promote the pledge.
I went to a MIRI workshop on decision theory last year. I came away with an understanding of a lot of points of how MIRI approaches these things that I’d have a very hard time writing up. In particular, at the end of the workshop I promised to write up the “Pi-maximising agent” idea and how it plays into MIRI’s thinking. I can describe this at a party fairly easily, but I get completely lost trying to turn it into a writeup. I don’t remember other things quite as well (eg “playing chicken with the Universe”) but they have the same feel. An awful lot of what MIRI knows seems to me folklore like this.
I think being too nice is a failure mode worth worrying about, and your points are well taken. On the other hand, it seems plausible to me that it does a more effective job of convincing the reader that Gleb is bad news precisely by demonstrating that this is the picture you get when all reasonable charity is extended.
I strongly suspect that the group photo is of very high value in getting people to go, making them feel good about having gone, and making others feel good about the conference. However, it sounds like trying to optimize to shave a few minutes off would be pretty high value.
What is remarkable about this, of course, is the recognition of the need to address it.
I agree with your second point but not your first. Also it’s possible you mean “optimistic” in your second point: if x-risks themselves are very small, that’s one way for the change in probability as a result of our actions to be very small.
Where the survey says 2014, do you mean 2015?
Yes, I’d treat the ratio of brain masses as a lower bound on the ratio of moral patient-ness.
Tax complicates this. If I’m in a higher tax band than you, I can make a donation to charity more cheaply than you can, so you will “receive” more than I “give”, and vice versa.
It seems a bit like the question behind the question might be “I’d like to help, but I don’t know formal logic, when will that stop being a barrier”. In which case it’s worth saying that I’m attending a MIRI decision theory workshop at the moment, and I don’t really know formal logic, but it isn’t proving too much of a barrier; I can think about the assertion “Suppose PA proves that A implies B” without really understanding exactly what PA is.
Thanks for the encouragement!
I wonder if you can do something with a different kind of disaster? Maybe make it a coach that can get people out of the danger zone? Or is that cheating because people don’t want seats to be “wasted”?
I’ve been trying to work out how to sell EA in the form of a parable; let me illustrate with my current best candidate.
In a post-apocalyptic world, you’re helping get the medicine that cures the disease out to the people. You know that there’s a truck with the medicine on the way, and it will soon reach a T-junction. The truck doesn’t know who is where and its radio is broken; you’re powerless to affect what it does, watching with binoculars from far away. If it turns left, it’ll be flagged down by a family of four and their lives will be saved. If it turns right, it’ll be flagged down by a school where dozens of families with the disease have taken refuge.
Don’t you find yourself fervently wishing the truck will turn right? It’s not because the family’s lives aren’t worth saving; they are, and they all deserve to live. But it’s clear that the better outcome is that it turn right.
So here’s some things I like about this: it’s not totally unfair. It’s not just a choice between “save A” and “save A and B”; if you make the most effective choice, then some people die who you could have chosen to save. And weirdly, I think the reframing where you can’t choose who gets saved, you can only will the truck to make the right decision, might help people see more clearly; you’re not worried about guilt about not saving the family or anger at someone making the wrong moral choice, just looking at a flip of a coin and discovering how you want it to land.
What I’d like to improve is to somehow make it more like an everyday situation rather than a super contrived one.
Any improvements? Does this seem like a useful exercise?
>> The group teaches or implies that its supposedly exalted ends justify means that members would have considered unethical before joining the group (for example: collecting money for bogus charities).
> Partial (+0.5)
This seems too high to me, I think 0.25 at most. We’re pretty strong on “the ends don’t justify the means”.
>>The leadership induces guilt feelings in members in order to control them.
> No
This on the other hand deserves at least 0.25...