Hey everyone!! I am SUPER excited to publish the fourth and final episode of the AI-Native Database podcast series with Dan Shipper and Bob van Luijt!<p>I thought Dan brought such a unique perspective on how we write and create with ChatGPT and Vector DBs! We dove into how subjectivity, or personality, is reflected in AI systems. For example, imagine you ask an LLM: "Could you plan a vacation to Miami?". Compared to previous database or search engine technologies that will just spit out relevant information to the query, the LLM might respond: "Are you sure you want to go to Miami? I think Fort Lauderdale is nicer at this time of year".<p>The podcast is titled "Subjectivity in AI" because this is such a novelty of AI compared to previous technologies. How will this unique "personality" of LLMs (as well as vector embeddings, classifiers, and other model types) seep into all these new applications? Can we control it with prompts and RAG, training data, or some combination of the above? What do multi-agent systems look like, each with their own personality and intrinsic motivation? Further, what new art forms will arise because of this property of AI, what will the "overdrive" of AI use look like?<p>I hope you enjoy the podcast, the last in our series! This has been so much fun to record and publish!<p>https://www.youtube.com/watch?v=prV5R3T6UqM