Presented by Sendbird, this VB Spotlight is all about how generative AI is revolutionizing customer engagement, improving CX at scale, and driving business growth. In this event, industry experts share real-world examples, discuss challenges, and offer practical insights to help you develop your organization’s gen AI strategy. You can watch it on demand here!
According to Jon Noronha, co-founder of Gamma, the biggest advantage of large language models (LLMs) is also their biggest drawback – their creativity. While this can be wonderful, it also means that LLMs can be unpredictable. Asking the same question of an LLM may yield different answers depending on slight differences in phrasing. This poses a challenge for companies building apps around LLMs, as they have to rethink their software development process and find ways to debug and monitor these models at scale. It’s a whole new ball game that requires infrastructure tools to help development teams understand how their LLMs perform in real-world scenarios.
Irfan Ganchi, CPO at Oportun, agrees that working with LLMs is not like working with traditional software. It’s not deterministic and even slight changes in inputs can lead to vastly different outputs. Additionally, tracing back through an LLM to understand why it produced a certain output is not easy. Crafting the perfect LLM for production requires a lot of trial and error, and the existing tooling for updating LLMs, test automation, and CI/CD pipelines is lacking. It’s a challenging process that demands caution.
Shailesh Nalawadi, head of product at Sendbird, adds that there are misconceptions surrounding generative AI in production environments. Many people believe that LLMs have real-time access to indexed information like Google search, but in reality, they are often trained on outdated data. To get the desired response from an LLM, users need to provide specific and relevant information. Prompt engineering becomes crucial in a business setting. Nalawadi also highlights the misconception that generative AI will automate everything. In reality, it enhances productivity and works in partnership with humans, rather than replacing them.
When leveraging generative AI, being intentional and having a fundamental strategy is key, says Ganchi. Organizations must incrementally test the value of generative AI and show its productivity. Before deploying it, there should be infrastructure in place to measure performance and evaluate outputs. Noronha emphasizes the importance of selecting the right problems to apply generative AI to. Finding cases where generative AI can tackle tasks that nobody wants to do or improve efficiency can lead to unexpected benefits, such as inspiring people to create things they wouldn’t have otherwise.
To delve deeper into the current state and future of generative AI, as well as learn from industry leaders and discover concrete ROI through real-world case studies, don’t miss this VB Spotlight event. You can register to watch it for free now! The agenda includes discussions on how generative AI is leveling the playing field for customer engagement, how different industries can harness its power, challenges and solutions related to large language models, and a vision of the future powered by generative AI. The presenters include Irfan Ganchi, Jon Noronha, Shailesh Nalawadi, and the moderator is Chad Oda from VentureBeat.
