tensor
A mathematical object generalizing scalars and vectors, often described circularly as 'something that transforms like a tensor'; a rank-2 tensor can be split into symmetric and antisymmetric parts.
As defined in Ask Me Anything | July 2024
Where it comes up · 4
- 211 | Solo: Secrets of Einstein's Equation4 claims
- Etched - Building AI Hardware to Make Inference Faster and Cheaper - [Invest Like the Best, EP.480]1 claim
- Ask Me Anything | July 20241 claim
- You need to learn AI in 2024! (And here is your roadmap)1 claim
Most important claims · 7 of 7Search all claims for “tensor” →
- Riemann's breakthrough was realizing that if you can specify the length of every infinitesimal line segment at every point—via a generalization of Pythagoras using an N×N array of numbers (the metric tensor, with N² components for N dimensions)—then you know everything about the geometry of a space intrinsically, without referring to any external embedding space.211 | Solo: Secrets of Einstein's Equation
- Although Newton treated mass as the source of gravity, E=mc² shows mass is merely one form of energy (rest energy), so the true relativistic source of gravity is the energy-momentum tensor—a two-index 4×4 array whose components encode energy density (TT), pressure (spatial diagonal), and stress, strain, and heat flow (off-diagonal)—unifying many pre-relativity quantities into one compact object.211 | Solo: Secrets of Einstein's Equation
- Curvature is characterized by how initially parallel lines fail to remain equidistant—diverging, converging, or twisting—and all this information is summed up in the Riemann tensor, a 4×4×4×4 array (256 components in four dimensions) with four indices, which is calculated by taking derivatives of the metric tensor to capture how the metric changes from point to point.211 | Solo: Secrets of Einstein's Equation
- Tensors are usually unhelpfully described as something that transforms like a tensor, but a better geometric explanation is possible, where a rank-2 tensor may encode two geometrical objects corresponding to its symmetric and antisymmetric parts.Ask Me Anything | July 2024
- A tensor is a generalization of a vector: where a vector requires a column of N numbers (its components) to specify in N-dimensional space, a tensor adds more indices, so a two-index tensor needs an N×N array and a four-index tensor needs N×N×N×N, with the metric tensor being a two-index example and the Riemann tensor a four-index example.211 | Solo: Secrets of Einstein's Equation
- Understanding PyTorch fundamentals like tensors and automatic differentiation (autograd) is worth working through because PyTorch does a lot behind the scenes, and if you skip this foundation, your code might run but you won't understand what's happening, making debugging harder.You need to learn AI in 2024! (And here is your roadmap)
- At a SEMI speakers' dinner, Gavin, then 22 and the youngest speaker, sat next to a senior TSMC VP who shared a math background and discussed AI tensor-level computation, and the next day TSMC asked to work with Etched.Etched - Building AI Hardware to Make Inference Faster and Cheaper - [Invest Like the Best, EP.480]