On the Dangers of Stochastic Parrots - Can Language Models Be Too Big?

Explore a thought-provoking keynote and panel discussion on the potential risks and ethical considerations surrounding large language models. Delve into Professor Emily M. Bender's presentation of her co-authored paper, examining the trend towards increasingly massive language models and questioning their implications. Gain insights into the history of language models, associated risks, mitigation strategies, and the distribution of benefits. Engage with topics such as unmanageable data, value lock, bias, and the concept of "stochastic parrots." Participate in a stimulating panel discussion featuring experts Dr. Anjali Mazumder, Dr. Zachary Kenton, and Professor Ann Copestake, addressing challenges, research trajectories, and the balance between academic and corporate interests in AI development.

Introduction
Presentation
Guiding Questions
Brief History
Graph
Risks
Mitigation
Who is getting the benefits
Unmanageable data
Value lock
Bias
How big is too big
Research trajectories
What is stochastic parrot
Risk management strategies
Challenges
Questions
Questions and comments
Can language models be too big
Building specificity
Incentives
Companies vs academia
Copilot
Comments
Are stochastic parrots agents