Microsoft recently hosted VentureBeat’s AI Impact Tour in New York City to discuss the risks and rewards of AI applications, as well as the ethics and transparency required. The event featured industry leaders from Microsoft, New York Presbyterian Hospital, and Citi, who shared their insights on how generative AI has impacted their organizations’ approach to industry challenges.
Sarah Bird, global lead for responsible AI engineering at Microsoft, highlighted the increased sophistication and understanding of AI among organizations. She emphasized the importance of identifying impactful and sophisticated use cases for AI, rather than applying it for the sake of it. Dr. Ashley Beecy from New York Presbyterian Hospital explained how they focus on balancing risks and rewards by prioritizing crucial use cases and urgent problems.
Dr. Promiti Dutta, head of analytics, technology, and innovation at Citi, discussed the paradigm shift brought about by the advent of generative AI. She highlighted the importance of considering internal policies and ensuring that use cases align with them, especially in highly regulated industries like finance and healthcare.
Bird emphasized the need for collaboration with customers and experts in specific domains to choose the right applications of generative AI. She mentioned how organizations like New York Presbyterian and Citi have combined immediate-win applications with use cases leveraging proprietary data to make a real difference for both the organizations and the users they serve.
The limitations and challenges of generative AI were also discussed. Bird addressed the term “hallucination” often used to describe the drawbacks of generative AI. She argued that the term personifies AI and can lead to ethical concerns. She explained that most generative AI applications are built with retrieval augmented generation, which can sometimes add information that doesn’t fit the context of the query. Microsoft is actively working to eliminate these grounding errors and improve the effectiveness of AI.
Looking ahead, Bird emphasized the importance of iteration in driving AI innovation. Microsoft’s partnership with OpenAI has allowed them to test the limitations of GPT-4 and find effective applications for the technology, such as scoring or labeling data with close to human capability. This progress enables continuous testing and training of AI models for better conversations and builds trust in the technology.
The VentureBeat AI Impact Tour will continue with stops in Boston and Atlanta, hosted by Microsoft.
In conclusion, the event highlighted the importance of balancing risks and rewards in AI applications and the need for responsible AI engineering. It also emphasized collaboration with experts in specific domains and the continuous iteration to drive innovation in generative AI.