“If people are open-minded creating a ‘human singularity’ should be pretty easy. We are connected individuals of 7.7 billion. We continuously improve, not incrementally like an AI singularity would imply. And we can improve exponentially.” — a dreamer
We believe everyone has limitless potential with two things: Zero external pressure. One hundred percent self-honesty. The latter follows the former.
Everything is constructed. Everything is an opinion. How do we even function? The answer is calm, honest, incremental communication.
So much of creation has two orthogonal parts. In writing you have to deliberately alternate between creating and critiquing, but not do both at same time.
Anxiety resulting in more reflection/reconsideration is actually a double-benefit: once it serves a purpose you can let it go.
Believing we are in a simulation is the result of being able to deal with both the infinite, the infinitesimal, the meta, and the lack of ego.
A reduced sense of wonder is key.
The initial friction against wanting to be immortal is the misconception one has limited energy. Moreover, everything can always be improved in the limit,
yet there is always more to explore. Like the Reinforcement Learning convergence conditions on epsilon or alpha parameters.
Bold hypothesis: A human can live their life without ever shutting their eyes to sleep.
One may have self-doubt even with a shitton of aphorisms, but that is because they think they’ll forget the lessons learned. Just like DRAM, refreshing is necessary. However, in humans, the refresh interval gets longer and longer exponentially (think spaced repetition. A good tool for this is Anki). Thus you can literally learn infinite with time, as power series converge. To make it concrete: given infinite time, if you can keep up with Anki, you can learn infinite. We will expand on this later, turning it into a mathematical proposition.
We learn and create new things because everyone who has ever existed has had access to what no one ever had before. Past scientists, for example, didn’t have Reinforcement Learning analogies to work with.
If you live forever you can literally do anything, as long as it follows the laws of physics. You can additionally teach anyone anything as long as they also live forever.
Always just take notes, unless you believe there is a very high probability you will remember it or come back to it later.
I never worry whatsover about writing anymore. It’s just a choice to no longer spend more time critiquing, qualifying, or adding more ideas. Similarly with decision making. My body generates anxiety but this is a benefit.
You don’t need to believe humans exist to prove stuff.
Starting to learn math is like an introduction to meta-thinking.
You need to write low-friction pieces if you want to be worth the time to your readers.
Don’t explain something to someone if they can explain it better to themselves.
Don’t overestimate your past self or underestimate your future self.
We are galaxies orbiting black holes. Naked loneliness is a highly productive feeling. One might think it is necessary to depend on others’ gravities, but really you must not fall under their influence.
Don’t just give risks thought proportional to their probability times impact. Give low probability risks zero thought. Humans cannot consider that many ideas simultaneuosly. We’re not neural networks.
No matter what, I have absolute belief I can keep on spinning.
Only the gradient matters.
Happiness is nowhere near as productive a concept as learning.
The goal is to align what you want with what you want, which is fully achievable in the limit.
Achieve certainty without external dependencies: You can gain confidence in the correctness of your hypotheses by telling yourself to do something and doing it. See Paul Erdos on amphetamines.
Most people wouldn’t even take the time to read a couple of these texts.
Maybe I was wrong and not only does real-life meta-thinking require the same techniques as the web-game Derivative Clicker, but it has the same mechanics. Compounding advantage.
When stressed about a specific external problem, dedicate discrete, incremental periods of time to analyze the problem from first principles.
Philosophically refine an idea until it’s near mathematical.
2 things we have dogs don’t: We can direct action. I can arbitrarily choose to lick the table. And we have the ability to reflect/meta-think at arbitrary levels (for a stupid example see Derivative Clicker).
We absolutely have wondrous superpowers and magical gifts compared to people 100 years ago. It’s not about a sense of wonder (I will elaborate on this). It’s about reflecting honestly on it.
If you’re honest with yourself about the effect of technology on you, you can begin to free yourself from its grasp.
We believe maybe half of people in similar situations to us share our same priors originating from childhood and evolution.
Force your priors into the ones you deeply want. In Reinforcement Learning, the value function is the prior from the perspective of an individual timestep.
The process of recall for memory is more helpful than one might think, because in addition to storing inputs, our memory also stores intermediate computations.
Never think the same thought twice in a row. You’re literally getting 0 new information. Spaced repetition can help with receiving new random seeds when revisiting thoughts.
Afraid to self-reflect? Feel like you’re biased? Afraid nothing’s real?. Try pretending you’re analyzing an alien, not a human.
Once you learn to optimize everything else, it’s trivial to optimize yourself.
We have a single anecdote of a person who (we believe) shows that anything is possible: our author Albert. He failed his classes with a GPA of less than 1.0. Along with his bro Aaron, he is writing this tweet.
Everyone has the potential for a self-singularity.