What do you care about in this world? I claim that this doesn't matter much, due to convergent instrumental goals of rational agents (for a more technical account, see here). For any set of values humans are likely to have, there are a set of instrumental goals that will be necessary for achieving them. I don't need to know whether I want there to be a revolution in the arts, or whether society should have three genders or none, to be able to claim more broadly that to get there we will need things such as "not being extinct" and "there existing scientific and technological progress in society".
Figuring out which pivotal events are going to occur in your future light cone, and getting into a position to effect positive change on them, is the main goal.
Early on in life, you should not expect to have good answers to these questions, so following some convergent instrumental strategies of your own would be well advised.
Suppose we discretise your life; say you have ~3 periods per year of Deep Work, where you can make significant progress on a project. That's 60 tokens of project-time in your first 20 years. Then how to spend these token? Which projects will teach you about the world, and which will give you the ability to change the world once you've picked more concrete goals?
What to optimise for?
Recently, I was at a conference, and I met a researcher from the human brain project. I asked this researcher about what they were working on, and they explained that they were working on a five year project. I don't recall the specifics of it, but I do recall asking them "When will you know whether this was a good project to work on?"They replied that they would find out in five years, when the project either failed or succeeded. I probed further: "And then will you have enough information to tell whether you made a good choice, as opposed to, for example, just having good luck?" and they replied that they had no way of knowing whether it was a good project to work on.
I think it's important to consider this a failure to plan well. This researcher, in five years time, will have learned nothing about how to choose research projects, and will be in no better a position to evaluate their future projects, toward whatever aims they have.
So the first step, is that you should always be testing a hypothesis. What current hypothesis to you believe about the world that would change your plans significantly if you found out you were wrong? Or, in a more gears-based framing, what node in your model of the world change your plans the most, if you were to expand the causal machinery inside of it?
Moving on from this, there's an issue whereby a lot of your ability to move through the world is dependant on how people perceive you. Most people who make decisions that significantly effect your life do not know you very well. Even those who will get to know you very well, often have to make a decision to work with you, when they don't know you very well. Robin Hanson argues that identity is a set of signals that tell people what enduring traits to expect of you, and I think this is right. Figure out what signals are the highest leverage throughout the world and then spend time and effort attaining them. Many of them are very generalisable - for example, a successful blog, business or research paper can be impressive to large swathes of people that aren't experts in your field. Importantly, where possible, figure out a narrative that suits you best, and design your CV / Resume / Linkedin profile around it.
And then finally, while you do not know what domain-specific skills you will need to work on the problem you eventually find, there are a number of domain-general skills that are very useful in a wide number of areas. Do you know how to build a team / organisation, and lead them to achieve an ambitious goal? Do you know how to enter a new domain and quickly figure out the useful abstractions / key variables that will allow you to succeed? Do you know how to encase your insights in clear prose that you can communicate to a broader intellectual community / members of your team? Certain skills are useful in many domains, and so building these skills as you test your hypotheses and gain signals will be useful.
What do you care about in this world? I claim that this doesn't matter much, due to convergent instrumental goals of rational agents (for a more technical account, see here). For any set of values humans are likely to have, there are a set of instrumental goals that will be necessary for achieving them. I don't need to know whether I want there to be a revolution in the arts, or whether society should have three genders or none, to be able to claim more broadly that to get there we will need things such as "not being extinct" and "there existing scientific and technological progress in society".
Figuring out which pivotal events are going to occur in your future light cone, and getting into a position to effect positive change on them, is the main goal.
Early on in life, you should not expect to have good answers to these questions, so following some convergent instrumental strategies of your own would be well advised.
Suppose we discretise your life; say you have ~3 periods per year of Deep Work, where you can make significant progress on a project. That's 60 tokens of project-time in your first 20 years. Then how to spend these token? Which projects will teach you about the world, and which will give you the ability to change the world once you've picked more concrete goals?
What to optimise for?
Recently, I was at a conference, and I met a researcher from the human brain project. I asked this researcher about what they were working on, and they explained that they were working on a five year project. I don't recall the specifics of it, but I do recall asking them "When will you know whether this was a good project to work on?"They replied that they would find out in five years, when the project either failed or succeeded. I probed further: "And then will you have enough information to tell whether you made a good choice, as opposed to, for example, just having good luck?" and they replied that they had no way of knowing whether it was a good project to work on.
I think it's important to consider this a failure to plan well. This researcher, in five years time, will have learned nothing about how to choose research projects, and will be in no better a position to evaluate their future projects, toward whatever aims they have.
So the first step, is that you should always be testing a hypothesis. What current hypothesis to you believe about the world that would change your plans significantly if you found out you were wrong? Or, in a more gears-based framing, what node in your model of the world change your plans the most, if you were to expand the causal machinery inside of it?
Moving on from this, there's an issue whereby a lot of your ability to move through the world is dependant on how people perceive you. Most people who make decisions that significantly effect your life do not know you very well. Even those who will get to know you very well, often have to make a decision to work with you, when they don't know you very well. Robin Hanson argues that identity is a set of signals that tell people what enduring traits to expect of you, and I think this is right. Figure out what signals are the highest leverage throughout the world and then spend time and effort attaining them. Many of them are very generalisable - for example, a successful blog, business or research paper can be impressive to large swathes of people that aren't experts in your field. Importantly, where possible, figure out a narrative that suits you best, and design your CV / Resume / Linkedin profile around it.
And then finally, while you do not know what domain-specific skills you will need to work on the problem you eventually find, there are a number of domain-general skills that are very useful in a wide number of areas. Do you know how to build a team / organisation, and lead them to achieve an ambitious goal? Do you know how to enter a new domain and quickly figure out the useful abstractions / key variables that will allow you to succeed? Do you know how to encase your insights in clear prose that you can communicate to a broader intellectual community / members of your team? Certain skills are useful in many domains, and so building these skills as you test your hypotheses and gain signals will be useful.
ExpectedValueOfProject=DecisionRelevantInfo+RareSignals+CrossDomainSkills