大多数人认为扩散模型(diffusion models)是视频生成的核心技术，并将持续主导这一领域，但作者认为未来视频生成的发展将更多地依赖于语言模型和代理技术，而非单纯的扩散方法。这挑战了当前AI生成领域的技术共识，暗示了语言模型可能在视频生成中扮演更重要的角色。

This is only true if, and only if, you actually believe Sam Altman* is telling the truth and won't change his mind in the future. If you don't think you can win a a lawsuit against OpenAI, and I mean actually win it, not just be correct about them violating their contract with you, then this particular argument does not hold for you.

*or Dario Amodei, but they're cut from the same cloth

A small number of samples can poison LLMs of any size<br /> by [[Anthropic]] from 2025-10-09 accessed on 2025-12-30T15:26:42

Interesting position paper about how to have useful discussion about AGI and Foundation models.

Detailed explanation of what DeepSeek model is doing differently to improve performance and training time over ChatGPT.

for - indyweb dev - question - alternative to AI Large Language Models? - Is indyweb functionality the same as Freespeech functionality? - from TED Talk - YouTube - A word game to convey any language - Ajit Narayanan - data structure of thought - from TED Talk - YouTube - A word game to convey any language - Ajit Narayanan

https://metagov.org/projects/koi-pond

Metagov's KOI (Knowledge Organization Infrastructure) is a graph database that supports relationships between knowledge objects, users, and groups within Metagov. via JM

Introducing Llama 3.1: Our most capable models to date by [[Meta]]

https://research.ibm.com/blog/retrieval-augmented-generation-RAG

PK indicates that folks using footnotes in AI are using rag methods.

Hubinger, et. al. "SLEEPER AGENTS: TRAINING DECEPTIVE LLMS THAT PERSIST THROUGH SAFETY TRAINING". Arxiv: 2401.05566v3. Jan 17, 2024.

Very disturbing and interesting results from team of researchers from Anthropic and elsewhere.

• for: progress trap -AI, carbon footprint - AI, progress trap - AI - bias, progress trap - AI - situatedness

Introduction of the RoBERTa improved analysis and training approach to BERT NLP models.

Wu, Prabhumoye, Yeon Min, Bisk, Salakhutdinov, Azaria, Mitchell and Li. "SPRING: GPT-4 Out-performs RL Algorithms byStudying Papers and Reasoning". Arxiv preprint arXiv:2305.15486v2, May, 2023.

Zecevic, Willig, Singh Dhami and Kersting. "Causal Parrots: Large Language Models May Talk Causality But Are Not Causal". In Transactions on Machine Learning Research, Aug, 2023.

"The Age of AI has begun : Artificial intelligence is as revolutionary as mobile phones and the Internet." Bill Gates, March 21, 2023. GatesNotes

Minda Zetlin. "Bill Gates Says We're Witnessing a 'Stunning' New Technology Age. 5 Ways You Must Prepare Now". Inc.com, March 2023.

Feng, 2022. "Training-Free Structured Diffusion Guidance for Compositional Text-to-Image Synthesis"

Shared and found via: Gowthami Somepalli @gowthami@sigmoid.social Mastodon > Gowthami Somepalli @gowthami StructureDiffusion: Improve the compositional generation capabilities of text-to-image #diffusion models by modifying the text guidance by using a constituency tree or a scene graph.

Original Paper for discussion of the Reinforcement Learning with Human Feedback algorithm.

GPT-2 Introduction paper

Language Models are Unsupervised Multitask Learners A. Radford, J. Wu, R. Child, D. Luan, D. Amodei, and I. Sutskever, (2019).

"Attention is All You Need" Foundational paper introducing the Transformer Architecture.

GPT-3 introduction paper

"Are Pre-trained Convolutions Better than Pre-trained Transformers?"

"LaMDA: Language Models for Dialog Application" Meta's introduction of LaMDA v1 Large Language Model.

Benyamin GhojoghAli Ghodsi. "Attention Mechanism, Transformers, BERT, and GPT: Tutorial and Survey"

LLAMA 2 Release Paper

Daniel Adiwardana Minh-Thang Luong David R. So Jamie Hall, Noah Fiedel Romal Thoppilan Zi Yang Apoorv Kulshreshtha, Gaurav Nemade Yifeng Lu Quoc V. Le "Towards a Human-like Open-Domain Chatbot" Google Research, Brain Team

Defined the SSI metric for chatbots used in LAMDA paper by google.

The Annotated S4 Efficiently Modeling Long Sequences with Structured State Spaces Albert Gu, Karan Goel, and Christopher Ré.

A new approach to transformers

Efficiently Modeling Long Sequences with Structured State SpacesAlbert Gu, Karan Goel, and Christopher R ́eDepartment of Computer Science, Stanford University

Bowman, Samuel R.. "Eight Things to Know about Large Language Models." arXiv, (2023). https://doi.org/https://arxiv.org/abs/2304.00612v1.

Abstract

The widespread public deployment of large language models (LLMs) in recent months has prompted a wave of new attention and engagement from advocates, policymakers, and scholars from many fields. This attention is a timely response to the many urgent questions that this technology raises, but it can sometimes miss important considerations. This paper surveys the evidence for eight potentially surprising such points: 1. LLMs predictably get more capable with increasing investment, even without targeted innovation. 2. Many important LLM behaviors emerge unpredictably as a byproduct of increasing investment. 3. LLMs often appear to learn and use representations of the outside world. 4. There are no reliable techniques for steering the behavior of LLMs. 5. Experts are not yet able to interpret the inner workings of LLMs. 6. Human performance on a task isn't an upper bound on LLM performance. 7. LLMs need not express the values of their creators nor the values encoded in web text. 8. Brief interactions with LLMs are often misleading.

Found via: Taiwan's Gold Card draws startup founders, tech workers | Semafor

This isn't true. The Stochastic Parrots paper outlines other avenues for reining in the harms of language models like GPT's.

Ganguli, Deep, Askell, Amanda, Schiefer, Nicholas, Liao, Thomas I., Lukošiūtė, Kamilė, Chen, Anna, Goldie, Anna et al. "The Capacity for Moral Self-Correction in Large Language Models." arXiv, (2023). https://doi.org/https://arxiv.org/abs/2302.07459v2.

Abstract

We test the hypothesis that language models trained with reinforcement learning from human feedback (RLHF) have the capability to "morally self-correct" -- to avoid producing harmful outputs -- if instructed to do so. We find strong evidence in support of this hypothesis across three different experiments, each of which reveal different facets of moral self-correction. We find that the capability for moral self-correction emerges at 22B model parameters, and typically improves with increasing model size and RLHF training. We believe that at this level of scale, language models obtain two capabilities that they can use for moral self-correction: (1) they can follow instructions and (2) they can learn complex normative concepts of harm like stereotyping, bias, and discrimination. As such, they can follow instructions to avoid certain kinds of morally harmful outputs. We believe our results are cause for cautious optimism regarding the ability to train language models to abide by ethical principles.

Google translation:

Every Egyptology student learns in their first semester that the Egyptian word pt (pronounced pet) means "heaven". The collection of documents in the dictionary archive comprises around 6,000 document slips. In the order of this material one learns that the Egyptian heaven has gates and ways, waters and banks, sides, pillars and chapels. This makes it tangible that the Egyptians had something completely different in mind when they heard the word "heaven" than modern Westerners do, namely a mythical space in which gods and spirits of the dead dwell.

This is a fantastic example of context creation for a dead language as well as for creating proper historical context.

In looking at the uses of and similarities between Wb and TLL, I can't help but think that these two zettelkasten represented the state of the art for Large Language Models and some of the ideas behind ChatGPT

Is there? By whom? Why industry only and not government, academia and civil society?

Bender, Emily M., Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜” In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610–23. FAccT ’21. New York, NY, USA: Association for Computing Machinery, 2021. https://doi.org/10.1145/3442188.3445922.

Would the argument here for stochastic parrots also potentially apply to or could it be abstracted to Markov monkeys?

"Talking About Large Language Models" by Murray Shanahan

tutorial article on BERT

DePonySum ask us to consider what you would need to program to be able to answer a wide range of eight grade science level questions (e.g. What can magnets be used for.) The answer is you would need a whole slew of separately trained and optimized models.

Language, they say, is a way to compress intellectual capacities.

It is then no surprise that common sense reasoning, and solving a wide range of problems, is first discovered through language models. Words and their relationships are probably a very efficient way of representing knowledge.