Airbyte的动态

? Can LLMs be light-weight, open-source and grounded? Yep! And we're talking about #DataGemma! We're excited to announce DataGemma, the world’s first open models designed to help address the challenges of hallucination by grounding LLMs?in the vast, real-world statistical data of Google's Data Commons. DataGemma will expand the capabilities of Gemma open models by harnessing the knowledge of Data Commons to enhance LLM #factuality and #reasoning using two distinct approaches: - RIG (Retrieval-Interleaved Generation)? - RAG (Retrieval-Augmented Generation) ?? These DataGemma models are available to researchers and developers?starting now. ↘ Read more or try them out! https://lnkd.in/dSGYRXBM #DataGemma #GoogleCloud #OpenModel #AIGrounding #LLM

For the second time in a row, MongoDB Atlas Vector Search has been named “The Most Loved Vector Search Database” in Retool’s 2024 “State of AI” report. ?? ?? The survey took the pulse of 700+ tech professionals, including developers, data teams, leadership, and others in technical roles and industries, to explore their perspectives on AI's current state, future outlook, and challenges in innovation. To learn more about the latest report and the benefits of Atlas Vector Search, read this article. ?? https://lnkd.in/gwZapDxg

??How has Text-to-SQL evolved over time, and how can you use LLMs for various use cases? Here’s a nice survey that sheds light on the topic Text-to-SQL has come a long way in recent years, offering a powerful way to boost productivity and reduce the need for deep SQL expertise in many industries. This paper gives a solid overview of Text-to-SQL methods, starting with traditional approaches, moving through early deep learning models, and mainly covering the latest LLM techniques like prompting, fine-tuning, and agents for the task. ?? It’s a great read if you’re considering Text-to-SQL for your use-case and want ideas to optimize performance/cost etc. Link: https://lnkd.in/evRDiuNn

We're big fans of Postgres so we're starting a short series on how you can use AI directly in your Supabase Postgres db! ?? Check out how you can automatically translate any user generated content right in the Supabase dashboard using Postgres functions and the http extension ??

Spreadsheets just keep coming back to Machine Learning ?? this time, as one of the top choices for experiment tracking in the 2024 State of Prod ML Survey!! Super interesting insights already, only within a week after launching the survey - let's continue working together to map the broader ML ecosystem!! Survey Link: https://lnkd.in/eFhTYvqm This is a community initiative and the data will provide all of us in the community with actionable insights to improve the ecosystem - your input will help create a comprehensive overview of common practices, tooling preferences, and challenges faced when deploying models to production, ultimately benefiting the entire ML community ?? The survey should only take a few minutes, please help us spread the word: https://lnkd.in/eFhTYvqm It would be massively appreciated if you can share it around with your networks - we are all going to benefit from the insights as the data will be broadly available for all of us to derive meaningful and actionable insights from it ?????? Chip Huyen, Ben Lorica 罗瑞卡, Pascal Biese, Maria Vechtomova, Dr. Larysa Visengeriyeva, Tim Januschowski, Rachel Mc Glynn, Tim Kr?ger, Peter Barron, Bill de hóra, Dr. Alexander Borek, Demetrios Brinkmann, Eduardo Ordax, Elvis S., Daniel Svonava, Demetris Cheatham, Prashanth Chandrasekar, Ibrahim Haddad, Ph.D., Dmitry Petrov, Tania Allard, PhD, Nayur Khan, Alex Housley, Alex Ioannides, Andrei Paleyes, Stephen Batifol, Libby Kinsey, ?????? Mikiko B., Rapha?l Hoogvliets, Animesh Singh, Hamza Tahir, Jesper Dramsch, Dhruv Ghulati, Jacopo Tagliabue, Luigi Patruno, Oleksandr Stasyk, Hien Luu, Eugene Yan, Jacomo Corbo, Dattaraj Rao, Azul Garza -- If you liked this article you can join 60,000+ practitioners for weekly tutorials, resources, OSS frameworks, and MLOps events across the machine learning ecosystem: https://lnkd.in/eRBQzVcA #ML #MachineLearning #ArtificialIntelligence #AI #MLOps #AIOps #DataOps #augmentedintelligence #deeplearning #privacy #kubernetes #datascience #python #bigdata

Some fascinating insights coming from The State of Prod ML Survey posted by Alejandro Saucedo (Seldon alum and ML engineering legend). We're still early! The early data shows that spreadsheets for experiment tracking are still a thing. I wonder how many AI teams are still using Flask + FastAPI wrappers for ML inference?! AI practitioners, please take a few moments to share your feedback on tools and processes for managing the production lifecycle of your models (link below). Whilst you're at it, if you haven't already, subscribe to Alejandro's ML Engineer newsletter - probably the best weekly read on the topic. Congrats on the 300-issue milestone ??????

Explore massive datasets, identify top trends, and build custom visualizations – all through natural language prompts. That's the power of the new data canvas in Gemini for BigQuery, available for a first look at #googlecloudnext. Analysts, get excited! Get a first look at the data canvas, a powerful new tool from Google that uses natural language to revolutionize data exploration. ??Effortlessly exploring datasets with everyday language ??Transforming data with a few simple prompts ??Creating insightful visualizations in minutes ??Discovering hidden trends and patterns with ease See the tool in action in the video demo linked in the first comment! ? Don't miss out! ?? Get the insights you need to succeed – follow?Boris Mergold #datacanvas #dataanalytics #datavisualization #Google #AskBoris

We want databases to have AI first tools that work hand in hand with your data and we've found some interesting ways we can do this! See how we upgraded Supabase with have built it translation ?? using Postgres functions

?? ?? Get ALL Census public metadata on ALL datasets and ALL variables ?? (in one script.) Link to my git w/code and link to data files in comments. I didn't think this would be such a major pita, but it was... lots of error handling and debugging. But why would I want such data? My goal is to see if I can: 1. Create a knowledge graph 2. Train an LLM / graphRAG 3. Make Census data more accessible to answer questions via natural language 4. Allow LLMs to write more accurate code/do analysis #AI #LLM #Census #DataScience

Super important information for anyone contemplating AI based applications. It goes without saying (but I'll say it anyway), if you cannot trust the answers from the data, then the lifetime value of that application immediately goes to zero. Don't make the mistake of building your AI application on the wrong data foundation...please read Dan's and Jennifer's insightful comments!