Responsible Generative AI: Ensuring Successful Real World AI Deployments

Current generative AI systems have already demonstrated value in terms of productivity, speed, and scales in many domains. However, current models suffer from bias, hallucination, and potential misuse. This panel will discuss some of the important AI and Generative AI application use cases (e.g., research, education, design, manufacturing, robotics, transportation, healthcare, cybersecurity) but also highlight vulnerabilities to attacks on their confidentiality, integrity and availability.  Adversary attacks include jailbreaking privacy or security information, poisoning training data, and denying access. In addition, adversaries can use LLMs to advance their goals by creating deep fakes or planning cyber-attacks. In this panel, leaders of real-world AI applications will share a range of important use cases along with practical methods participants are employing to counter risks (e.g., diversifying training data and teams, employing guardrails, establishing policies and training users) in order to safely deliver value to society.