Language models are incredible engineering breakthroughs but require auditing and risk management before productization. These systems raise concerns about toxicity, transparency and reproducibility, intellectual property licensing and ownership, disinformation and misinformation, supply chains, and more. How can your organization leverage these new tools without taking on undue or unknown risks? While language models and associated risk management are in their infancy, a small number of best practices in governance and risk are starting to emerge. If you have a language model use case in mind, want to understand your risks, and do something about them, this presentation is for you! We’ll be covering the following:
– Studying past incidents in the AI Incident Database and using this information to guide debugging.
– Adhering to authoritative standards, like the NIST AI Risk Management Framework.
– Finding and fixing common data quality issues.
– Applying general public tools and benchmarks as appropriate (e.g., BBQ, Winogender, TruthfulQA).
– Binarizing specific tasks and debugging them using traditional model assessment and bias testing.
– Engineering adversarial prompts with strategies like counterfactual reasoning, role-playing, and content exhaustion.
– Conducting random attacks: random sequences of attacks, prompts, or other tests that may evoke unexpected responses.
– Countering prompt injection attacks, auditing for backdoors and data poisoning, ensuring endpoints are protected with authentication and throttling, and analyzing third-party dependencies.
– Engaging stakeholders to help find problems system designers and developers cannot see.
– Everyone knows that generative AI is going to be huge. Don’t let inadequate risk management ruin the party at your organization!
Talk by: Patrick Hall
Here’s more to explore:
LLM Compact Guide: https://dbricks.co/43WuQyb
Big Book of MLOps: https://dbricks.co/3r0Pqiz
Connect with us: Website: https://databricks.com
Twitter: https://twitter.com/databricks
LinkedIn: https://www.linkedin.com/company/databricks
Instagram: https://www.instagram.com/databricksinc
Facebook: https://www.facebook.com/databricksinc
Add comment