Summary of On Catastrophic Inheritance Of Large Foundation Models, by Hao Chen et al.
On Catastrophic Inheritance of Large Foundation Models
by Hao Chen, Bhiksha Raj, Xing Xie, Jindong Wang
First submitted to arxiv on: 2 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This position paper proposes a framework called UIM (Understand, Interpret, Mitigate) to address the neglected issue of catastrophic inheritance in large foundation models (LFMs). LFM performances have raised great concerns about their uninterpreted potentials not only in machine learning but also in various other disciplines. The authors identify weaknesses and limitations inherited from biased pre-training data, which can cause catastrophes on downstream tasks, such as bias, lack of generalization, deteriorated performance, security vulnerability, privacy leakage, and value misalignment. The proposed framework aims to unite the machine learning and social sciences communities for more responsible AI development and deployment. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large foundation models are getting incredible performances, but there’s a problem! These models inherit weaknesses from their training data, which can cause big problems when they’re used in other tasks. This paper proposes a way to understand these issues, figure out what they mean, and fix them. The authors want to make sure AI is developed and used responsibly. |
Keywords
* Artificial intelligence * Generalization * Machine learning
