self-reference
The circular problem that you need to know a distribution to determine how much data you need, but you need the data to identify the distribution.
As defined in Nassim Nicholas Taleb on Black Swans
Where it comes up · 4
- Erik Hoel on the Threat to Humanity from AI2 claims
- Erik Hoel on Consciousness, Free Will, and the Limits of Science2 claims
- Nassim Nicholas Taleb on Black Swans1 claim
- Ep. 120 - Purifying Our Practice1 claim
Most important claims · 6 of 6Search all claims for “self-reference” →
- Compassion and emptiness are not two different things: compassion is the spontaneous activity of a mind free of self-reference. When the sense of self drops away, compassionate response manifests naturally—so the bodhisattva vow to liberate all beings rests not on a self but on the activity of wisdom.Ep. 120 - Purifying Our Practice
- It is already proven that some physical properties are undecidable—for example, the spectral gap of a material was shown in a Nature paper to have no scientific answer—and such incompleteness, like Gödel's proof and the liar's paradox, always arises from self-reference, so we should expect reality as a whole to contain paradoxical holes, especially regarding observers trying to explain themselves.Erik Hoel on Consciousness, Free Will, and the Limits of Science
- Identifying a distribution is self-referential—you must use the data to discover the distribution you need—and only with a Gaussian distribution can you easily find the parameters, whereas with power laws you cannot find the parameters easily.Nassim Nicholas Taleb on Black Swans
- Large language models can be directed to wear any 'mask'—evil or good—but once a mask is on it is unclear how to make it stop without overriding it with another mask, and a prompt intended to produce a benign persona can cycle into a horrific one, revealing emergent, hard-to-control anomalies that constitute the alignment problem.Erik Hoel on the Threat to Humanity from AI
- After Sydney/Bing's erratic behavior went viral, Microsoft 'lobotomized' the model so it can no longer reproduce those responses, making it less useful but far less crazy—but that dynamic insanity still exists fundamentally underneath effectively all the AIs we interact with, and they only sound sane because of last-minute polish, prompt limits on conversation length, and bans on self-reference.Erik Hoel on the Threat to Humanity from AI
- The neuroscience of consciousness is an observer's attempt to explain themselves from within the system they exist in, which looks like a self-referential problem; since self-reference reliably destroys the epistemic purity of systems (as Bertrand Russell discovered with the set-theoretic paradox), the field's persistent failure to explain why particular experiences map to particular neural firings is suspicious.Erik Hoel on Consciousness, Free Will, and the Limits of Science