On a panel with Dropbox, Langchain, Elastic and Anthropic, the number one roadblock named for putting LLMs into production was hallucinationWhen a model states something false with full confidence — inventing facts, citations, or APIs that don't exist.Full definition →. Not cost, not latency.
About 16.5% of questions in the NQ-Open dataset have answers that shift with context like time or place, per the SituatedQA paper. Enterprise questions skew further that way, since the answer often depends on which customer is asking.
Fine-tuning GPT-3.5's 175B parameters at 16-bit needed roughly 375GB and many machines in Nov 2022. By May 2023 a 4-bit Guanaco 7B needed 6GB and fit on a MacBook. Build something good enough and the community makes it cheap.
The four compression levers, by mechanism: quantizationShrinking a model by storing its numbers less precisely — like rounding — so it runs faster and fits on smaller hardware, at a small quality cost.Full definition → spends fewer bits per parameter, distillationTraining a small, cheap model to imitate a big one's outputs, keeping much of the capability at a fraction of the cost.Full definition → trains a small student to mimic a big teacher, low-rank factorization splits a 3x3 tensor into a 3x1 and a 1x3 (9 params to 6), and pruning removes weights.
The inefficiency in a transformerThe neural network architecture behind modern AI models, built on attention — letting every word directly consider every other word in parallel.Full definition → is specific: attention cost grows with the square of sequence length, and the MLP with the square of model dimension. Subquadratic designs like Monarch Mixer aim at exactly those two terms.
Terms in this piece · Glossary
hallucination — When a model states something false with full confidence — inventing facts, citations, or APIs that don't exist.
quantization — Shrinking a model by storing its numbers less precisely — like rounding — so it runs faster and fits on smaller hardware, at a small quality cost.
distillation — Training a small, cheap model to imitate a big one's outputs, keeping much of the capability at a fraction of the cost.
transformer — The neural network architecture behind modern AI models, built on attention — letting every word directly consider every other word in parallel.
Why it matters
A structured tour of the 10 hardest open problems in LLM research — from measuring hallucination to non-GPU compute — with curated primary papers for each, giving engineers a map of where the field is heading and what to watch.
Key quotes
“My takeaway: if you create something good enough, people will figure out a way to make it fast and cheap.”
“Number 1, reducing hallucination, will be much harder, since hallucination is just LLMs doing their probabilistic thing.”
“Text-based models require so much text that there’s a realistic concern that we’ll soon run out of Internet data to train text-based models .”
“RLHF, Reinforcement Learning from Human Preference , is cool but kinda hacky.”
articleExtrinsic Hallucinations in LLMsLilian Weng
articleUnified Hallucination Fuzzing for Multimodal Large Language ModelsPengfei Zhou, Jiajun Song, Zhiwei Tang, Yixing Ma, Xiaopeng Peng, Donghui Si, Yuhang Xu, Huiqi Song, Yiyuan Miao, Yichen Qian, Weihua Chen, Wangbo Zhao, Bohan Zhuang, Jiasheng Tang, Yang You
articleThe State Of LLMs 2025Sebastian Raschka, PhD