Chat GPT and Large Language Models Have Compositional Inference and Planning Capabilities

My guest today is a giant in the world of voice assistants and AI. Adam Cheyer is the co-founder of Siri, which was originally an app that, as you know, was acquired by Apple and launched as the key feature of the iPhone 4s in 2011. That is how Siri came to define expectations for modern-day voice assistants.   However, Cheyer and one of his Siri cofounders Dag Kittlaus, were disappointed that Apple decided to dramatically narrow the scope of the assistant to better align with its goal of making Siri a feature of the iPhone as opposed to a general purpose assistant.   Cheyer and Kittlaus founded Viv Labs, which eventually sold to Samsung and, after a few years, became the technology behind the new Bixby after Samsung's initial voice assistant flopped. A key part of the story here is that Cheyer was the visionary behind 50% of the four leading voice assistants to emerge over the past 15 years. You can imagine that gives him unique insights into what is happening today with the rise of generative AI and ChatGPT in particular.   Cheyer and I speak at length about his initial reaction to ChatGPT, what he thinks OpenAI got right, and foreshadows an emerging issue about how the company applies Plugins.    In addition, you will want to hear more about Cheyer's four critical elements for an assistant and the difference between a knowing and doing assistant. This creates a stark contrast between ChatGPT and Alexa. Sit back and get ready for some gems of insight.