3 ms·
Historical precedent: Quickselect, or any other discovery of a surprising algorithmic speed-up. LLMs haven’t been on the scene for very long. There’s still lot
by odyssey7 3mo ago
Historical precedent: Quickselect, or any other discovery of a surprising algorithmic speed-up.
LLMs haven’t been on the scene for very long. There’s still lots of room for efficiency discoveries. Plus, I keep hearing quantum is going to be a big deal in the next few years.
- _glass 3mo agoSuper interesting, when Quantum computing would enable really large models, or much larger contexts. But QRAM is even more behind than pure fault-tolerant gates.