Events have become a great way to stay in touch with a space that is evolving so fast that in a blink of an eye might make you miss the next LLM advancement. That's why we're dedicating this newsletter to recap some of the amazing events we had the chance to attend.? But before we go there, did you read our recent announcement?
The last few weeks have been a whirlwind of preparing for, presenting, or learning at conferences. Each event we attended was packed with facts on the ground and led to very real discussions about AI deployments across various industries. Open vs closed source LLMs, various techniques to improve RaGs and LLM responses, LLM guardrails, costs of fine-tuning, database innovations, importance of data quality for good AI outcomes, AI agendas, practical use cases and real-world examples of how Gen AI efforts are being successful or struggling, were just some of the key topics across these different events. Given how rapidly AI tools, methods, and models are evolving, events like these make it manageable to keep up and become part of some innovations, especially when we get to hear straight from the founders at the forefront of the AI revolution.?
TL;DR - it’s all about multimodality, embeddings, knowledge graphs, LLMs, and chatbots!
- AI User Group Meetup in SF: Extracting the right embeddings to incorporate relevant metadata (Superlinked) and the ability to manage the embeddings along with the metadata (ApertureData) were the key highlights of the evening gathering organized by the AI User Group. It became clear that embeddings need to be paired with rapidly evolving metadata (e.g., product recommendations based on both visual features and popularity). There was also a much-needed deep dive into the challenges of LLM reliability and an impressive demo of Google Gemini's capabilities, including its ability to search through game actions and screen captures.
- WiDS Summit hosted by IBM SF: We attended the Women in Data Science (WiDS) San Francisco at IBM where the speakers covered a broad range of very relevant topics ranging from insights on building intelligent multi-agent systems, securing LLMs, entrepreneurial journeys in the light of rapidly evolving AI landscape, building AI safeguards for election integrity, and leveraging a combination of remote sensing, Machine Learning, and Causal Inference for climate solutions. Vishakha discussed how ApertureData has evolved to meet the growing demand for managing multimodal data and supporting complex AI-driven business use cases. She also illustrated real-world use cases of how ApertureDB combines vector database, knowledge graph, and multimodal data management into one interface to simplify building and deploying AI applications at scale (slides here).
- MLOps World + Generative AI World in Austin: Multimodal AI was one of the key trends highlighted in the kickoff keynote. There were plenty of talks and discussions throughout the conference around RAG workflows, vector database usage, the value of knowledge graphs, the need for simplifying multimodal data management, the importance of guardrails, LLM finetuning for domain-specific workloads, and the cost of breaking compliance with the new wave of question-answering systems. Lots of organizations are still trying to figure out the heads and tails of this new, rapidly evolving, post-ChatGPT world. Chatbots (simple to agentic ones) for question answering based on internal knowledge bases seem like the most common and accepted use case right now for Generative AI. Of course, there are some challenges like the correctness of responses, response times, how graphs can improve answers, and the ability to parse complex PDFs accurately while maintaining permissions structures. It was also exciting to explore several use cases for traditional machine learning and see how they are being enhanced with generative AI. (Stay tuned for a blog on this coming soon!)
The ApertureData Challenge submission deadline has been extended to Wednesday, November 20th, 2004 @ 11:59 PM PST. You now have extra time to finish your masterpiece. Have you seen the updated prizes and the very experienced judges who are looking forward to your innovative submissions? Check out the details on ApertureData Challenge: https://lu.ma/jrvh73zg
Start your project today with a free trial of ApertureDB Cloud.
