Probability distribution
A set of non-negative numbers assigned to outcomes that sum to 1, specifying how likely each outcome is.
As defined in 174 | Tai-Danae Bradley on Algebra, Topology, Language, and Entropy
Other definitions
A mathematical representation of the likelihood of different outcomes; in language models, the output is a probability distribution over all possible next words rather than a single prediction
AI: Grappling with a New Kind of Intelligence | World Science Festival
Where it comes up · 10
- Markov Chains Clearly Explained! Part - 16 claims
- But what is the Central Limit Theorem?4 claims
- Howard Marks | The Knowledge Project #533 claims
- 174 | Tai-Danae Bradley on Algebra, Topology, Language, and Entropy3 claims
- AI: Grappling with a New Kind of Intelligence | World Science Festival2 claims
- Ruminating on Asset Allocation2 claims
- Human vs. Machine Consciousness | Imperial’s Murray Shanahan1 claim
- Geoff Hinton - Will Digital Intelligence Replace Biological Intelligence? | Vector's Remarkable 20241 claim
and 2 more
Most important claims · 10 of 24Search all claims for “Probability distribution” →
- A probability distribution on N+1 elements is precisely an N-simplex: the set of all such distributions forms geometric shapes (a point, line, triangle, tetrahedron, and higher), so probabilities are inherently geometric and topological objects, making entropy's connection to topology unsurprising.174 | Tai-Danae Bradley on Algebra, Topology, Language, and Entropy
- Investment decisions should be based on expected value (probability-weighted outcomes), but some possible outcomes under high expected value strategies may be unsurvivable, so expected value alone is an insufficient decision criterion without considering risk tolerance.Howard Marks | The Knowledge Project #53
- Risk exists because the future is uncertain and is a range of possibilities; sometimes we know what's in that range and sometimes we don't; managing risk requires developing a view of the probability distribution of future events.Howard Marks | The Knowledge Project #53
- The stationary distribution π is a special row vector that satisfies the equation π·A = π, meaning the probability distribution does not change after multiplying by the transition matrixMarkov Chains Clearly Explained! Part - 1
- One resource-rational strategy humans use is sampling: instead of representing a whole probability distribution, you consider only a few samples or possible instances, and instead of considering all possible outcomes when making a decision, you consider only a few—and the cases people consider tend to be the ones they should consider given limited resources.343 | Tom Griffiths on The Laws of Thought
- When simulating the restaurant Markov chain for 100,000 steps, the probability distribution converges to fixed values known as the stationary distribution or equilibrium state, meaning this distribution does not change with timeMarkov Chains Clearly Explained! Part - 1
- When a row vector π representing a probability distribution is multiplied by the transition matrix A, the result is the probability distribution in the next time step, and this process can be repeated iterativelyMarkov Chains Clearly Explained! Part - 1
- To find the stationary distribution of a Markov chain, one must solve two equations: the eigenvector equation π·A = π (with eigenvalue 1) and the constraint that the elements of π sum to 1 (because it is a probability distribution)Markov Chains Clearly Explained! Part - 1
- Large language models learn to produce grammatically correct, coherent text purely from algebraic and statistical input — seeing which words go with which at what frequencies — without any explicit grammatical labeling, demonstrating empirically that algebra plus statistics suffices to capture much of language.174 | Tai-Danae Bradley on Algebra, Topology, Language, and Entropy
- To train a world model by predicting what happens in video frames doesn't work using standard LLM techniques because videos have many plausible futures and we cannot train the system to predict all possible scenarios in the high-dimensional continuous space of video frames the way we can represent discrete word probabilities.AI: Grappling with a New Kind of Intelligence | World Science Festival