Censoring chemical data to mitigate dual use risk
Not on the map. Why?
Abstract (arXiv)
Machine learning models have dual-use potential, potentially serving both beneficial and malicious purposes. The development of open-source models in chemistry has specifically surfaced dual-use concerns around toxicological data and chemical warfare agents. We discuss a chain risk framework identifying three misuse pathways and corresponding mitigation strategies: inference-level, model-level, and data-level. At the data level, we introduce a model-agnostic noising method to increase prediction error in specific desired regions (sensitive regions). Our results show that selective noise induces variance and attenuation bias, whereas simply omitting sensitive data fails to prevent extrapolation. These findings hold for both molecular feature multilayer perceptrons and graph neural networks. Thus, noising molecular structures can enable open sharing of potential dual-use molecular data.
Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 16 of its 111 references that are in the map. First posted on arXiv 20 Apr 2023.
Builds on
16 of 111 references.
References · 16
- 2022Training language models to follow instructions with human feedbackcited by 15,971
- 2016Semi-Supervised Classification with Graph Convolutional Networkscited by 11,361
- 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,669
- 2022Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedbackcited by 3,720
- 2017MoleculeNet: A Benchmark for Molecular Machine Learningcited by 1,051
- 2022Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learnedcited by 729
- 2023A Watermark for Large Language Modelscited by 725
- 2019Federated Machine Learning: Concept and Applicationscited by 412
- 2020Responsive Safety in Reinforcement Learning by PID Lagrangian Methodscited by 316
- 2018Low-Shot Learning from Imaginary Datacited by 181
- 2023Emergent autonomous scientific research capabilities of large language modelscited by 132
- 2016Low Data Drug Discovery with One-shot Learningcited by 91
4 more
- 2020Understanding Double Descent Requires a Fine-Grained Bias-Variance Decompositioncited by 76
- 2022Self-Destructing Models: Increasing the Costs of Harmful Dual Uses of Foundation Modelscited by 57
- 2021DGL-LifeSci: An Open-Source Toolkit for Deep Learning on Graphs in Life Sciencecited by 38
- 2020Avoiding Negative Side Effects due to Incomplete Knowledge of AI Systemscited by 5
Built on by
Citing papers per year: 2023: 1; 2024: 2.
3 papers.