mlsanityData to 7 Oct 2026

On the Limitations of Compute Thresholds as a Governance Strategy

arXiv 2407.05694 ↗First posted 8 Jul 2024cs.AI cs.CL cs.ETcited by 17

Not on the map. Why?

Abstract (arXiv)

At face value, this essay is about understanding a fairly esoteric governance tool called compute thresholds. However, in order to grapple with whether these thresholds will achieve anything, we must first understand how they came to be. To do so, we need to engage with a decades-old debate at the heart of computer science progress, namely, is bigger always better? Does a certain inflection point of compute result in changes to the risk profile of a model? Hence, this essay may be of interest not only to policymakers and the wider public but also to computer scientists interested in understanding the role of compute in unlocking breakthroughs. This discussion is timely given the wide adoption of compute thresholds in both the White House Executive Orders on AI Safety (EO) and the EU AI Act to identify more risky systems. A key conclusion of this essay is that compute thresholds, as currently implemented, are shortsighted and likely to fail to mitigate risk. The relationship between compute and risk is highly uncertain and rapidly changing. Relying upon compute thresholds overestimates our ability to predict what abilities emerge at different scales. This essay ends with recommendations for a better way forward.

Key facts. Cited by 17 papers in the map (data through 7 Oct 2026). Builds on 90 of its 201 references that are in the map. First posted on arXiv 8 Jul 2024.

Builds on

90 of 201 references.

References · 90

  1. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,491
  2. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,160
  3. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  4. 2020Scaling Laws for Neural Language Modelscited by 6,411
  5. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  6. 2018GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understandingcited by 4,854
  7. 2014Going Deeper with Convolutionscited by 4,231
  8. 2019Unsupervised Cross-lingual Representation Learning at Scalecited by 3,907
  9. 2021Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsitycited by 3,014
  10. 2022Emergent Abilities of Large Language Modelscited by 2,483
  11. 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 2,081
  12. 2024Mixtral of Expertscited by 1,785
38 more
  1. 2020GShard: Scaling Giant Models with Conditional Computation and Automatic Shardingcited by 1,601
  2. 2017Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inferencecited by 1,548
  3. 2022SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Modelscited by 1,338
  4. 2019Energy and Policy Considerations for Deep Learning in NLPcited by 1,242
  5. 2020RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Modelscited by 1,235
  6. 2023AWQ: Activation-aware Weight Quantization for LLM Compression and Accelerationcited by 1,234
  7. 2021Scaling Language Models: Methods, Analysis & Insights from Training Gophercited by 1,199
  8. 2017A Closer Look at Memorization in Deep Networkscited by 936
  9. 2023A General Theoretical Paradigm to Understand Learning from Human Preferencescited by 906
  10. 2022LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scalecited by 881
  11. 2022Quantifying Memorization Across Neural Language Modelscited by 755
  12. 2021Scaling Vision with Sparse Mixture of Expertscited by 740
  13. 2023Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizescited by 710
  14. 2019Green AIcited by 620
  15. 2017Learning Sparse Neural Networks through L0 Regularizationcited by 571
  16. 2019Defending Against Neural Fake Newscited by 553
  17. 2016Learning Structured Sparsity in Deep Neural Networkscited by 538
  18. 2021Deep Learning on a Data Diet: Finding Important Examples Early in Trainingcited by 529
  19. 2022Beyond neural scaling laws: beating power law scaling via data pruningcited by 519
  20. 2016Quantized Neural Networks: Training Neural Networks with Low Precision Weights and Activationscited by 413
  21. 2020Closing the AI Accountability Gap: Defining an End-to-End Framework for Internal Algorithmic Auditingcited by 380
  22. 2015Deep Learning with Limited Numerical Precisioncited by 368
  23. 2021Scaling Laws for Transfercited by 265
  24. 2019Massively Multilingual Neural Machine Translation in the Wild: Findings and Challengescited by 253
  25. 2024Semantic Entropy Probes: Robust and Cheap Hallucination Detection in LLMscited by 246
  26. 2024Foundational Challenges in Assuring Alignment and Safety of Large Language Modelscited by 193
  27. 2020The Computational Limits of Deep Learningcited by 188
  28. 2024Aya Dataset: An Open-Access Collection for Multilingual Instruction Tuningcited by 174
  29. 2018Mesh-TensorFlow: Deep Learning for Supercomputerscited by 168
  30. 2019What Do Compressed Deep Neural Networks Forget?cited by 164
  31. 2020ProtTrans: Towards Cracking the Language of Life's Code Through Self-Supervised Deep Learning and High Performance Computingcited by 162
  32. 2020Characterising Bias in Compressed Modelscited by 162
  33. 2022Compute Trends Across Three Eras of Machine Learningcited by 161
  34. 2020The Hardware Lotterycited by 160
  35. 2023Generative Language Models and Automated Influence Operations: Emerging Threats and Potential Mitigationscited by 157
  36. 2020Challenges in Deploying Machine Learning: a Survey of Case Studiescited by 151
  37. 2023Pushing Mixture of Experts to the Limit: Extremely Parameter Efficient MoE for Instruction Tuningcited by 148
  38. 2022Lifting the Curse of Multilinguality by Pre-training Modular Transformerscited by 145

50 most cited of 90 All 90 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 4; 2025: 12; 2026: 1.

17 papers.

Sort:

By area:

  1. 2025Can We Trust AI Benchmarks? An Interdisciplinary Review of Current Issues in AI Evaluationcited by 57
  2. 2025One Tokenizer To Rule Them All: Emergent Language Plasticity via Multilingual Tokenizerscited by 11
  3. 2025The Multilingual Divide and Its Impact on Global AI Safetycited by 10
  4. 2025When Life Gives You Samples: The Benefits of Scaling up Inference Compute for Multilingual LLMscited by 9
  5. 2024The Reality of AI and Bioriskcited by 8
  6. 2024Nexus: Specialization meets Adaptability for Efficiently Training Mixture of Expertscited by 7
  7. 2024MEG: Medical Knowledge-Augmented Large Language Models for Question Answeringcited by 6
  8. 2025Red Teaming AI Policy: A Taxonomy of Avoision and the EU AI Actcited by 6
  9. 2025International Agreements on AI Safety: Review and Recommendations for a Conditional AI Safety Treatycited by 3
  10. 2025A Framework for the Private Governance of Frontier Artificial Intelligencecited by 3
  11. 2025NeurIPS should lead scientific consensus on AI policycited by 3
  12. 2025Practical Principles for AI Cost and Compute Accountingcited by 2
5 more
  1. 2025AI Agents Should be Regulated Based on the Extent of Their Autonomous Operationscited by 2
  2. 2024Countering Autonomous Cyber Threatscited by 1
  3. 2025Local Differences, Global Lessons: Insights from Organisation Policies for International Legislationcited by 0
  4. 2025From Turing to Tomorrow: The UK's Approach to AI Regulationcited by 0
  5. 2026Governance at the Edge of Architecture: Regulating NeuroAI and Neuromorphic Systemscited by 0