[2110.10819] Shaking the foundations: delusions in sequence models for interacti...
source link: https://arxiv.org/abs/2110.10819
Go to the source link to view the article. You can view the picture content, updated content and better typesetting reading experience. If the link is broken, please click the button below to view the snapshot at that time.
Computer Science > Machine Learning
[Submitted on 20 Oct 2021]
Shaking the foundations: delusions in sequence models for interaction and control
The recent phenomenal success of language models has reinvigorated machine learning research, and large sequence models such as transformers are being applied to a variety of domains. One important problem class that has remained relatively elusive however is purposeful adaptive behavior. Currently there is a common perception that sequence models "lack the understanding of the cause and effect of their actions" leading them to draw incorrect inferences due to auto-suggestive delusions. In this report we explain where this mismatch originates, and show that it can be resolved by treating actions as causal interventions. Finally, we show that in supervised learning, one can teach a system to condition or intervene on data by training with factual and counterfactual error signals respectively.
Comments: | DeepMind Tech Report, 16 pages, 4 figures |
Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
Cite as: | arXiv:2110.10819 [cs.LG] |
(or arXiv:2110.10819v1 [cs.LG] for this version) | |
https://doi.org/10.48550/arXiv.2110.10819 |
Submission history
From: Pedro Alejandro Ortega [view email][v1] Wed, 20 Oct 2021 23:31:05 UTC (130 KB)
Recommend
About Joyk
Aggregate valuable and interesting links.
Joyk means Joy of geeK