LLM
Large language model—an AI trained on vast text corpora to predict and generate language.
As defined in cwt_alison-gopnik
Where it comes up · 113
- Facing Superintelligence (with Ben Goertzel)23 claims
- 258 | Solo: AI Thinks Different17 claims
- Metas AI Boss Says He DONE With LLMS...12 claims
- 434 - Can We Survive Ai11 claims
- Yann LeCun: Special Lecture on AI and World Models10 claims
- Is RAG Still Needed? Choosing the Best Approach for LLMs9 claims
- Theprimeagen8 claims
- Conversations with Tyler 2023 Retrospective7 claims
and 105 more
Most important claims · 10 of 336Search all claims for “LLM” →
- An AI trained to predict and imitate humans does not thereby become human inside, just as an actress who learns to perfectly predict and imitate every drunk in a bar does not become drunk - becoming drunk would actually impair the cognitive acuity needed to predict the drunks - so learning to predict human text does not give an alien architecture human values.434 - Can We Survive Ai
- Historical papers from Shane and Hassabis (2009 and 2016) anticipated key developments: Shane's thesis identified compression of Wikipedia as a test for AI (which became the loss function for LLMs), and Hassabis's 2016 neuroscience-AI paper identified attention as a critical capability (which was later crystallized in transformers).Demis Hassabis — Scaling, superhuman AIs, AlphaZero atop LLMs, AlphaFold
- Unlike pure quantitative strategies where backtesting uses out-of-sample testing and established historical testing methodologies, LLM-based systems know all historical data simultaneously, potentially negating traditional out-of-sample validation techniques that have been foundational to quantitative strategy development.Redefining Value Investing in a Magnificent Seven Dominated World | Jacob Pozharny
- LLMs combined with enterprise technology will build application companies whose business models sell work rather than seats, so they compete for operating expenditure budgets instead of software tool budgets, a different framework from the history of software.Peter Fenton & Victor Lazarte - Purpose and Partnership - [Invest Like the Best, EP.354]
- Today's big ideas are things people build — large language models, Ethereum, potentially superconductivity — rather than books; the big-idea books were always a poor substitute for the actual big ideas, so the shift toward technological ideas is preferable.Marginal Revolution 20Th Anniversary
- For decades, Moravec's paradox held that tasks easy for humans were hard for computers and vice versa; modern AI reversed expectations by becoming good at fluid human tasks like writing coherent essays and conversation - which the field thought would be hardest - while still stumbling on simpler-seeming tasks, and only later excelling at math and science.434 - Can We Survive Ai
- Because LLMs are probabilistic and predict the most likely next item, an image model asked for a doctor might default to the majority demographic and thereby produce even more under-representation than actually exists, which is one legitimate reason to deliberately curate training data toward greater diversity.What Does "Unbiased" Mean in the Digital World? (with Megan McArdle)
- Roblox's auto-translation relies on early machine learning models such as BERT and other lightweight filters rather than very large language models, because translation and filtering tasks can be done effectively and quickly without a trillion-row LLM.Conversations with Tyler - Dave Baszucki
- Contrary to a decade-old expectation that the most powerful AI models would be built air-gapped from the internet and only released after careful safety assurances, in reality the most powerful models are connected to everything and released to hundreds of millions of people more or less the moment they are built—making the assumed safety step a complete mirage.427 - Ai Friends Enemies
- The default path of the technology is that the high-capacity security state will eventually run an LLM across the passively-produced dataset of every electronic transaction, making the implicit invasion of privacy far worse than what was baked into the system in 1980 when surveillance required physically examining microfiche in archives.Patrick McKenzie — Money laundering, big tech censorship, SBF & Japan