The Gradient of Generative AI Release: Methods and Considerations
By Irene Solaiman. 2023
141 citations recorded by the source corpus.
As increasingly powerful generative AI systems are developed, the release method greatly varies. We propose a framework to assess six levels of access to generative AI systems: fully closed; gradual or staged access; hosted access; cloud-based or API access; downloadable access; and fully open. Each level, from fully closed to fully open, can be viewed as an option along a gradient. We outline key considerations across this gradient: release methods come with tradeoffs, especially around the tension between concentrating power and mitigating risks. Diverse and multidisciplinary perspectives are needed to examine and mitigate risk in generative AI systems from conception to deployment. We show trends in generative system release over time, noting closedness among large companies for powerful systems and openness among organizations founded on principles of openness. We also enumerate safety controls and guardrails for generative systems and necessary investments to improve future releases.
Subfields: Health Informatics, Information Systems and Management, Safety Research, Natural Language Processing, Safety, Robustness & Interpretability
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Health Informatics
In Information Systems and Management
- Tri Dao
- Irene Solaiman
- Raymond Li
- Siva Reddy
- Dzmitry Bahdanau
- Margaret Mitchell
- Sara Hooker
- Muntasir Wahed
In Safety Research
- Irene Solaiman
- Siva Reddy
- Nicholas Meade
- Marie-Therese Png
- Zeerak Talat
- Yacine Jernite
- Alexandra Sasha Luccioni
- Su Lin Blodgett
In Natural Language Processing
- Siva Reddy
- Jason Aaron Edward Weston
- Daniel Fried
- Margaret Mitchell
- Yejin Choi
- Jianfeng Gao
- Yacine Jernite
- Thomas Wolf
In Safety, Robustness & Interpretability
- Su Lin Blodgett
- Irene Solaiman
- Dylan Baker
- Marc Marone
- Nicholas Meade
- Zeerak Talat
- Siva Reddy
- Andreas Engel
Citations in the corpus
- Evaluating the Social Impact of Generative AI Systems in Systems and Society (2023, background): “Increasing trends toward system closedness [227] is shifting access distribution.”
- StarCoder: may the source be with you! (2023, background): “In contrast, other LLM development projects have taken a fully open approach which aims to allow for community inputs into model development, release training data, and enable external audits throughout the full development process (Solaiman, 2023).”
- Regulating ChatGPT and other Large Generative AI Models (2023, background): “Additional factors, such as the balance of power among developers, must also be considered [149].”