Embark on a decade-long journey exploring AI safety and trust in this 58-minute podcast episode featuring Petar Tsankov, Co-Founder and CEO of LatticeFlow AI. Delve into key areas such as the transition towards adversarial environments, challenges in model robustness and data relevance, and the necessity of third-party assessments. Explore current AI trends, emphasizing problems with biases, errors, and lack of transparency in generative AI and third-party models. Learn about LatticeFlow AI's mission to provide trustworthy solutions for new AI applications, their participation in safety competitions, and focus on proving neural network properties. Gain insights on data quality, robustness checks, emerging standards like ISO 5259 and ISO 40001, and the future of AI regulation and certifications. Essential listening for anyone passionate about trust and safety in AI.
A Decade of AI Safety and Trust - Challenges and Solutions
Overview
Syllabus
[] Petar's preferred coffee
[] Takeaways
[] Shout out to LatticeFlow for sponsoring this episode!
[] Please like, share, leave a review, and subscribe to our MLOps channels!
[] Expansion
[] Zurich ETH
[] AI Safety
[] Optimizing one metric, no fixed data sets
[] Trust life-changing issues
[] So much interest in GenAI
[] Explosion of GenAI Trust and Safety
[] Red Teaming
[] Trustworthy AI in Industry
[] DataOps Challenges
[] Trusting Third-Party Models
[] Testing Open Source Models
[] Specialized ML for Leasing
[] Regulation and Financial Incentives
[] Regulations Drive Innovation Balance
[] Regulations vs Certification: Voluntary Prove
[] Workflow Transparency: Trust & Efficiency
[] Engineers Balance Compliance Risks
[] Pushing Deep Learning Limits
[] Wrap up
Taught by
MLOps.community
