This post is well over a decade old, yet no one noticed that in 1984, Winston actually writes that "Freedom is the freedom to say that two plus two makes four. If that is granted, all else follows."
And in being tortured, he grants that 2+2 = 5
Eliezer has introduced a different sum, which makes the same point, but it's not from Orwell's 1984
Sharing Aaron Swartz's review from 2011
"Harry Potter and the Methods of Rationality by Eliezer Yudkowsky
This is a book whose title still makes me laugh and yet it may just turn out to be one of the greatest books ever written. The writing is shockingly good, the plotting is some of the best in all of literature, and the stories are simply pure genius. I fear this book may never get the accolades it deserves, because it’s too hard to look past the silly name and publishing model, but I hope you, dear reader, are wiser than that! A must-read.
As it says at the beginning, you really need to give it a couple chapters to get started before passing judgment — the first bunch are quite silly and it doesn’t seem worth sticking with until you’ve gotten past them."
So glad you've come over to share your updated experience.
And I was one of the 30MM who shared the huge chart of biases way back.
I was a student in the tradition of Heuristics & Biases when I studied at Stanford, so long ago that Amos Tversky was still alive. Both Kahneman & Tversky explicitly doubted that knowing about biases could enable us to overcome the 'fast thoughts' that generate an immediate interpretation. I've heard both of them compare biases to optical illusions, which don't go away when you realize they are illusions.
Hamming Questions are core to some exercises in CFAR workshops.
Personally, I've never been motivated by setting goals. Once they are fixed, the removal of exploration and the single mindedness of optimization are fatal to sustaining my interest.
I don't know if CFAR has ever clicked into the resistance that comes up when people are confronted with the question of what is the work of greatest significance that one could possibly do.
At least in my dissertation research, I found people were more reluctant to set goals for the things that most mattered to them. My interpretation was that it was a way to evade the possibility of discovering you've failed at something really meaningful. This was called "The Delmore Effect", as it was robustly observed that people had explicit and well-structured goals for lower priority ambitions, but less articulate, more sketchy ideas for pursuing the activities most important to their identity
I'd never heard the phrase "fist pump" before, but apparently it is an alternative with a slightly different hand arrangement than "fist bump." Given how close the sounds b/p are, I may have been collapsing numerous occasions of hearing "pump" mapping it to "bump"
https://www.nytimes.com/2012/10/28/magazine/who-made-that-fist-bump.html
Excellent ideas. I (hormetic) just re-tweeted my admiration for "pick directions, not goals"
My dissertation examined why people hesitate to explicitly articulate goals for the domains that matter most. Because a goal has a definite failure state, it can motivate exertion to avoid failure.
Upstream of that motivation, however, we don't want to expose ourselves to the risk of failure. The more important some domain is, the more threatening it will be to encounter evidence of having failed. I observed that people (erm, Stanford undergrads) were least articulate about goals when the domain was very important. I labeled this "The Delmore Effect," to describe the robust decay in goal articulation as something was of increasing importance. Source: https://web.archive.org/web/20160317034318/http://www-psych.stanford.edu/~wit/PhDraft.pdf
Unless the identical twin sisters marry identical brother twins, the preferential love for their own husband would distinguish their feelings for their own children
I'm in! One suggestion. The URL for registering has de minimis info, https://www.havenbookings.space/events/Duncon
Ideally, that page should link to this one to help contextualize what's in the can
RatBabble, for prototyping text layouts with something better than fake Latin Lorum Ipsem
Bayesian update epsilon prior heuristic decoherence acausal simulacrum UDT corrigibility instrumental convergence mesa optimizer cognitive prior Tarski epistemic coherence orthogonality mindspace Srisk metaethical anthropic bias bounded utility risk gradient decision theoretic Nash bargaining iterative amplification qualia inversion adversarial epistemology fractal uncertainty dynamic utility heuristic threshold convergence margin probabilistic inference variance tradeoff model alignment signal extraction noise reduction emergent property modularity recursion value stability meta update robust calibration uncertainty quantification prior adjustment likelihood estimation hypothesis space inference chain stochastic gradient loss minimization error correction systemic risk Bayesian calibration parameter sweep decision node utility function bounded rationality incentive compatibility strategic abstraction paradox resolution epistemic control consistency check reflective equilibrium alignment protocol instrumental strategy entropy maximization risk factor simulation fidelity synthetic insight heuristic mapping fractal pattern recursive structure optimization landscape epistemic frontier decision boundary variance reduction signal detection information formalizable idea uncertainty propagation noise floor calibration curve epistemic horizon integration test convergent series discontinuous jump scaling law tractability analytic solution rational approximation strategic update adversarial signal probability density utility maximization nonlinear dynamics feedback loop algorithmic bias stochastic resonance meta level alignment risk cognitive architecture convergence heuristic asymptotic limit dynamic equilibrium prior probability likelihood ratio causal inference synthetic model computational theory epistemic trace divergence measure simulation offset margin error calibration constant inference update signal noise paradigm shift alignment envelope rational framework strategic equilibrium iterative process adversarial equilibrium update vector noise model risk horizon convergence path fractal dimension structural inference bounded update decision tree coherence metric meta parameter signal integrity value alignment uncertainty heuristic convergence rate prior extraction likelihood function decision threshold heuristic calibration gradient descent error term probabilistic model signal inference entropy measure instrumental pivot fractal algorithm update scheme stochastic iteration prior refinement calibration feedback dynamic model equilibrium state convergence criterion risk model probabilistic simulation inference vector signal fluctuation meta balance noise variance decision gradient strategic recursion adversarial function probabilistic threshold convergence point parameter tuning update matrix uncertainty update risk profile strategic parameter meta abstraction calibration error stochastic block variance parameter inference loop signal a locally entropy function update module calibration loop prior distribution heuristic structure signal threshold convergence analysis risk escalation decision filter meta aggregation bounded strategy error propagation utility matrix fractal recursion uncertainty balance iterative simulation calibration shift update impulse strategic noise convergence burst prior vector likelihood measure stochastic drift decision switch epistemic update error distribution meta feedback bounded inference signal process risk adjustment convergence regulator heuristic balance probabilistic inference update cycle calibration step fractal resonance strategic feedback adversarial iteration utility ‘valid’ if convergence signal meta structure error analysis stochastic update bounded calibration signal gap likelihood adjustment prior alignment heuristic coherence uncertainty factor update algorithm decision span risk inertia convergence algorithm probabilistic formalizable idea meta parameter bounded vector error metric signal rhythm calibration constant update variable strategic mathematics where adversarial parameter utility threshold convergence core iterative update heuristic bias probability flux update quantum calibration pulse signal vector bounded true statements error bandwidth likelihood pulse prior algorithm epistemic drift uncertainty core decision amplitude meta burst strategic recursion adversarial step is utility to a convergence density iterative fractal heuristic impulse update rhythm calibration flow signal is simplest bounded impulse error quotient likelihood fractal prior channel epistemic chain uncertainty impulse decision fractal meta iteration strategic chain adversarial rhythm utility model theory convergence quotient iterative chain heuristic iteration update ‘In math’ calibration iteration signal semantically true bounded iteration error of a likelihood iteration prior coherence epistemic iteration uncertainty in model decision step is meta coherence strategic iteration adversarial The notion utility iteration convergence step is iterative iteration heuristic iteration update a formalizable calibration iteration signal semantically true bounded iteration error evaluated argument likelihood iteration prior coherence epistemic iteration uncertainty step is meta iteration strategic iteration adversarial a formalizable utility iteration convergence iterative coherence heuristic formalizable idea update iteration calibration coherence signal to a bounded iteration error true statements likelihood iteration prior coherence epistemic iteration uncertainty if it decision ‘valid’ if meta iteration strategic iteration adversarial true statements utility iteration convergence iterative coherence heuristic formalizable idea update iteration calibration coherence signal to a bounded iteration error true statements likelihood iteration prior coherence epistemic iteration uncertainty of a decision statements relative utility iteration convergence math a iterative coherence heuristic formalizable idea update iteration calibration coherence signal to a bounded iteration error true statements likelihood iteration prior coherence epistemic iteration uncertainty type of decision relative to it only other semantically statements from theory In produces semantically a given true statements math a formalizable idea mathematics where model theory type of true statements ‘valid’ if utility iteration convergence
Inspired by @arithmoquine on X