Published Feb 29, 2024

Are LLMs safe?

Suchin Gururangan delves into the safety of large language models by examining parameter-efficient adaptations, the role of decentralization in respecting user data privacy, and the socio-technical challenges in LLM training, highlighting innovative strategies to enhance model performance and ethical compliance.

Episode Highlights

Topics covered

Episode Highlights

Related Episodes

117 - Interpreting NLP Model Predictions, with Sameer Singh
Answers 383 questions
141 - Building an open source LM, with Iz Beltagy and Dirk Groeneveld
Answers 383 questions
40 - On the State of the Art of Evaluation in Neural Language Models, with Gábor Melis
Answers 383 questions
130 - Linking human cognitive patterns to NLP Models, with Lisa Beinborn
Answers 383 questions
44 - Truly Low Resource NLP, with Anders Søgaard
Answers 383 questions
109 - What Does Your Model Know About Language, with Ellie Pavlick
Answers 383 questions
63 - Neural Lattice Language Models, with Jacob Buckman
Answers 383 questions
81 - BlackboxNLP, with Afra Alishahi and Tal Linzen
Answers 383 questions
137 - Nearest Neighbor Language Modeling and Machine Translation, with Urvashi Khandelwal
Answers 383 questions
115 - AllenNLP, interviewing Matt Gardner
Answers 383 questions
114 - Behavioral Testing of NLP Models, with Marco Tulio Ribeiro
Answers 383 questions
64 - Neural Network Models for Sentence Pair Tasks, with Wuwei Lan and Wei Xu
Answers 383 questions
106 - Ethical Considerations In NLP Research, with Emily Bender
Answers 383 questions
136 - Including Signed Languages in NLP, with Kayo Yin and Malihe Alikhani
Answers 383 questions
80 - Leaderboards and Science, with Siva Reddy
Answers 383 questions

Dexa/NLP Highlights

Are LLMs safe?

Topics covered

Popular Clips

Language Model Data Curation

Customizing NLP Models

Surfing and Innovations

Dynamic Conversation

Model Merging Insights

Efficient Model Adaptation

Language Model Training

Modular Task Vectors

Adaptive Pre Training

Decentralizing Language Models

Data is Key

Decentralized Model Training

Language Model Retrieval

High School Newspapers

Episode Highlights

Model Adaptation Techniques

Decentralization in LLMs

LLM Training Challenges

Related Episodes

117 - Interpreting NLP Model Predictions, with Sameer Singh

141 - Building an open source LM, with Iz Beltagy and Dirk Groeneveld

40 - On the State of the Art of Evaluation in Neural Language Models, with Gábor Melis

130 - Linking human cognitive patterns to NLP Models, with Lisa Beinborn

44 - Truly Low Resource NLP, with Anders Søgaard

109 - What Does Your Model Know About Language, with Ellie Pavlick

63 - Neural Lattice Language Models, with Jacob Buckman

81 - BlackboxNLP, with Afra Alishahi and Tal Linzen

137 - Nearest Neighbor Language Modeling and Machine Translation, with Urvashi Khandelwal

115 - AllenNLP, interviewing Matt Gardner

114 - Behavioral Testing of NLP Models, with Marco Tulio Ribeiro

64 - Neural Network Models for Sentence Pair Tasks, with Wuwei Lan and Wei Xu

106 - Ethical Considerations In NLP Research, with Emily Bender

136 - Including Signed Languages in NLP, with Kayo Yin and Malihe Alikhani

80 - Leaderboards and Science, with Siva Reddy

Are LLMs safe?

Topics covered

Popular Clips

Episode Highlights

Model Adaptation TechniquesSuchin Gururangan explores innovative methods for adapting LLMs efficiently, focusing on parameter-efficient techniques and task vectors. These approaches aim to enhance model performance without the need for costly retraining.

Model Adaptation Techniques

Decentralization in LLMsSuchin Gururangan explores decentralized training and user data control in large language models (LLMs), emphasizing the importance of data privacy and ethical considerations. He discusses innovative approaches to empower data owners and improve model utility.

Decentralization in LLMs

LLM Training Challenges

Related Episodes