AI/ML Terms You May Not Know, Battle of the LLMs, Unstructured Data Meetups, and More!
In this issue:
AI/ML Terms You May Not Know
Which LLM Knows Vector Databases the Best?
Voyage AI Embeddings and Rerankers for Search and RAG
ICYMI: Unstructured Data Talk Recaps from Around the World
Upcoming Events
AI/ML Terms You May Not Know
Why is it relevant?
All forms of data are constantly coming into the world, so the need for tools to navigate it has become more crucial. BERTopic employs neural network-based techniques to uncover themes and patterns in large text data with unprecedented accuracy and depth.
What is Topic Modeling? AI VS Bread 🤖🍞
Topic modeling: a method for unearthing the latent themes or "topics" within a collection of documents. Examine the text within these documents to detect patterns and relationships that indicate the presence of these topics. 🔎
🍞Example: a document focused on artificial intelligence will likely contain terms like "large language models" and "ChatGPT," unlike a document centered on baking bread.
What is BERTopic?
BERTopic: a novel topic modeling technique that simplifies the topic modeling process. It uses various embedding techniques and class-based TF-IDF (c-TF-IDF) to create dense clusters, allowing for easily interpretable topics while keeping important words in the topic descriptions. 💡
It approaches topic modeling in 4 steps on a high level:
▶️ Document embedding: Convert documents into embeddings using Bidirectional Encoder Representations from Transformers (BERT).
▶️ Dimensionality Reduction: Compresses embeddings into a lower-dimensional space.
▶️ Clustering: Group these embeddings to gather similar documents in one category.
▶️ Topic Extraction: Extract topic names using a class-based variation of TF-IDF.
Which LLM Knows Vector Databases Best?
We asked LLMs to explain what a vector database is to someone who isn’t familiar with it. Which LLM do you think “won” this response? Let us know in the comments below and see who we picked! ⬇️
Here are the results from various LLMs using the same prompt:
Anthropic Claude (Claude 3.5 Sonnet):
ChatGPT (GPT 4o):
Google Gemini:
Vector databases are super smart librarians! 🤓
Voyage AI Embeddings and Rerankers for Search and RAG
Voyage AI provides various customized embedding models across many domains to carry out efficient RAG techniques. These models are connected with vector databases like Milvus by Zilliz to store and retrieve vector embeddings related to the generated query.
Zilliz partnered with Voyage AI to streamline the conversion of unstructured data into searchable vector embeddings on Zilliz Cloud. Voyage AI embedding models integrated in Zilliz Cloud Pipelines are voyage-code-2, voyage-law-2, and voyage-large-2-instruct. These models are for specific code, law, finance, and multilingual domains.
See the step-by-step on using this integration and Cohere (the LLM) to build a RAG application: Read Blog
Watch the meetup talk with Voyage AI CEO Tengyu Ma on “cutting-edge embeddings and rerankers for search and RAG”
ICYMI: Unstructured Data Talk Recaps from Around the World
We had a packed schedule in July 🌞with fun AI events all over the world. See the recaps and register for next month’s events! ⬇️⬇️⬇️
▶️ Garbage In, Garbage Out: Why poor data curation is killing your AI models (and how to fix it)
▶️ It's your unstructured data: How to get your GenAI app to production (and speed up your delivery)
▶️ Multimodal Embeddings
Recap: July 17 Berlin @ Google
▶️ Gemini with Advanced Function Calling
▶️ RAG using PGvector
▶️ Docker tools & Open Source libraries for productivity
▶️ How to Load Test an LLM API using Gatling
▶️ Improving Analytics with Time Series and Vector Databases
▶️ Scaling Vector Search: How Milvus Handles Billions+
Save your spot for August events
Upcoming Events
August 1: Unstructured Data Processing from Cloud to Edge (virtual)
Learn why you should add a Cloud Native Vector Database to your Data and AI platform. Tim Spann 🥑 will cover a quick introduction to Milvus, vector databases, and unstructured data processing. By adding Milvus to your architecture, you can scale out and improve your AI use cases through RAG, real-time search, multimodal search, recommendation engines, fraud detection, and many more emerging use cases.
August 2-4: AI-focused Touring Hackathon & Tech Fair (in-person)
🌟 TechStars Startup Weekend Personal AI Hackathon: Fri-Sat with the pitches 🗣️ and a Tech Fair 🎪 on Sunday showcasing local AI startups.
Christy Bergman will be there on Friday to answer your Milvus questions and kickstart the hackathon!
Register to be a Hacker, Designer, Product Manager, Growth Hacker, DevOps Engineer, or Sales & Business Dev
August 5: San Francisco Unstructured Data Meetup (in-person)
Join us in San Francisco for a meetup on August 5! There will be food, refreshments, networking, and cool AI talks.
▶️ Using Ray Data for Multimodal Embedding Inference with Christy Bergman , Developer Advocate at Zilliz
▶️ Building the Future of Neural Search: How to Train State-of-the-Art Embeddings with Aamir Shakir , Co-founder, Mixedbread
▶️ A Different Angle: Retrieval Optimized Embedding Models with Marqo
Save your spot: https://lu.ma/3q2brqp8
August 8: Building an Agentic RAG locally with Milvus, Ollama, and Llama Agents (virtual)
With the recent release of Llama Agents, we can now build agents that are async first and run as their own service. During this webinar, Stephen Batifol will show you how to build an Agentic RAG System using Llama Agents and Milvus.
August 13: South Bay Unstructured Data Meetup (in-person)
Stay tuned for the speaker line up coming soon!
Save your spot: https://lu.ma/tutkha5k
-
In the next newsletter, we’ll cover Softmax and Local Sensitivity Hashing (L.S.H.). Start reading now:
▶️ Decoding Softmax: Understanding Its Functions and Impact in AI
▶️ Local Sensitivity Hashing (L.S.H.): A Comprehensive Guide
Thanks for tuning in and see you next month! 👋
Co-Founder of Altrosyn and DIrector at CDTECH | Inventor | Manufacturer
4moUnderstanding vector databases is crucial for optimizing LLM performance, especially when dealing with high-dimensional data and semantic search. Models like OpenAI's GPT-4 and Google's PaLM are known for their robust integration with vector databases, facilitating efficient similarity searches and context retrieval. This capability enhances their utility in various applications, from recommendation systems to advanced information retrieval. Considering the rapid advancements in vector embeddings and indexing techniques, how do you envision the evolution of LLMs influencing the optimization and scalability of vector databases in the near future? What innovative approaches might emerge to further enhance the efficiency of these integrations?