I went to the NYC Secular Solstice this year. It was wondrous, but something stood out as I was compiling my thoughts about it on the way home. In one of the conversations after, someone expressed that up until now they had been assuming Solstice messages were directed at heroes but that they might have to update towards that no longer being true. Solstice spoke of heroism, and we cheered heroism, and then most people went back to the mundane lives they came from. That seemed unbearable sad to me.
I've been rereading HP:MoR recent (partially as a test drive of sequence reading on LessWrong 2.0 and partially due to a habit of rereading works that I find foundational or inspirational in my thinking) and I came back from NYC to find that I was on the tail end of the Azkaban section, specifically the refusal of the phoenix. The realization that this moment has passed, that it will not be offered again, is one of the two moments in HP:MoR that never fail to wrench a sob from me.
I think I might have wanted to be a hero, long ago, when I was a child and idolized Luke Skywalker and Mathias of Redwall and Nita Callahan. If that was true, it hasn't been true for a long time, that impulse was lost somewhere in my process of adapting to the world. The role that I wanted to fit instead was villain. I drove myself forward for years by looking to Grendel, Darth Maul, Peter Wiggin and other figures willing to ignore costs and convention to achieve what they wanted.
If Solstice is supposed to speak to heroes, then why does it speak to me?
Well, what do I mean when I call myself a villain?
I want the world to be different for my being in it. I want enough stocked power that when something in the world makes me sad or angry or upset I can reach out and make it be different. I would enjoy the thought that mankind as a whole or any given member of it owes me their lives. I like hurting people, exerting power over the world, and the feeling of winning. I’m more comfortable on the offense, causing things to happen and watching others react to me rather than waiting and reacting to what others do. I tend to act without default deference to social role- social norms form a sort of soft power that I've grown more conscious of as I've grown older, but I still can't bring myself to find "what will people think" at all a reason by itself to do or not do something.
It's only recently that I've found my first mental urge in response to idiocy is no longer wanting to cause them pain until they stop, that “how can I help them” is my true reaction to others instead of a mask that was useful to wear.
This is a change that took place when I wasn't looking. Several years of mostly successfully suppressing anger and fury, more years before that of not needing them on a weekly or daily basis, the slow dawning that I actually can be safe and warm and fed if I want to, and perhaps most of all pretending to be a mind much like my own but slowly bettering itself- these have let me relax my deathgrip on trying to maintain the ability to hurt anyone I need to. Empathy is still strange magic to me, but compassion isn't anymore. I find myself wanting to ease the pain of others as a part of my own utility function, not just for signalling reasons.
I am a strong proponent of words having meanings, and I have to admit I am not sure how well "villain" fits anymore. Using "hero" still feels strange when I think about it- perhaps I'm the sort of villain who shows up to Endbringer fights and keeps the peace while I build resources? I Am Samwise suggests the TV Tropes title of Dragon for the strong right hand of a larger figure, which I could very circumstantially find to be an ideal relationship, but there is a time for labels and a time for recognizing that “what you have, Mr. Potter, is freedom.” I am free. What do I want to be?
I just signed a lease on my current apartment for the next year. At the end of 2018, I'll have paid off my loans from the government and from family. I had been planning to study information security, since it seemed fun and well paid, and then settle down somewhere to garden my own little piece of the world and make fantastic wads of money and then retire early and do whatever the blazes I wanted. But when I thought about terrible circumstances that could happen, slippery slopes that the world around me could slide down, I drew lines around what could constitute a call to action for me. Then AlphaZero crossed one of those lines, demanding a response.
And this also connects to something I learned at Solstice. That someone else drew a line around an uncertain danger, a threshold to demand a response, and when that threshold was passed and travel into and out of the country that they lived in was limited, they found they did not actually take exceptional action to address the danger. They did not leave, they did not revolt, they did not even (I assume, though the world would look the same if they secretly did) make contact with underground elements to secure an escape route should the need arise. When they talked about this, they talked about learning that they wouldn't actually act even when they had previously decided they would and how that was a thing worth knowing about themselves. That reminded me of the subjects in the Milgram Experiment who wrote afterward to thank Milgram for the self-knowledge, and ten-year old!me sitting on the library floor reading about that and growing cold and declaring that I would never let my morality be set by an authority, that I would never do what I was told simply because I was told to. It made school into an even deeper level of hell, but I am glad of it now.
I need time to study, and I need the financial security the next year will bring if this is to be a marathon, but I am going to start attacking the FAI problem. I wish I could say this was due to rationally considering the expected cost and benefit, and the line was drawn this way, but I must confess it is an emotionally driven decision made by two thoughts.
I will be strong enough to act when I decide I will act
And to hear a phoenix song go unanswered is too sad to bear.
I’ll be studying machine learning and AI alignment this coming year. My goal is to be useful enough at the former that I could do it for a job, and useful enough at the latter to where any new progress would be comprehensible to me. Recommendations for study approaches or materials are appreciated. (I have read Bostrom's Superintelligance and other non-technical things, I am a decent programmer in general and in other areas of focus but haven't done any work in this one, I'm currently reading Functional Decision Theory.)