Alexander Garcia’s Post

It's time to take your developer experience to the next level! Stack Overflow for VS Code brings together AI and validated knowledge from your Stack Overflow for Teams instance as well as the larger Stack Overflow community. Learn more about our OverflowAI modules now available: https://lnkd.in/eaSztckh

This year’s NeurIPS challenge was an important step for improving transparency around model evaluation and democratizing the usage of open-source, small language models. Along with PyTorch and Microsoft Research, Mozilla.ai helped evaluate the #NeurIPS efficiency challenge around training 1 LLM on 1 GPU in 1 day. Many state-of-the-art #LLMs remain gated behind expensive and often proprietary hardware. The 150+ submissions to the challenge proved that it is possible to obtain performant language models with open-source on something as little as a single GPU, given a limit of 24 hours.    To this end, the challenge also sought to address some of the following issues:   - Lack of transparency around model training, efficiency, and evaluation leads to issues of reproducibility around model performance.   - The lack of standardized evaluation methodologies for open-source models hinders meaningful comparisons.   - Insufficient access to large-scale hardware limits the open-source community to engage and leverage large models.      The event was sponsored by Lambda, Lightning AI, Amazon Web Services (AWS), and Mozilla.ai. Learn about this year's winners at: https://lnkd.in/dmiZ2thA #opensource #opensourcecommunity #neurips2023 #largelanguagemodels

#DBRX is now the best open model on AI2's WildBench, a benchmark of challenging tasks from real LLM users: https://lnkd.in/gqQT2jfk. If you haven't tried it, it's super easy to try it in this HuggingFace space: https://lnkd.in/gUZ-Ph4k or get started using the Databricks Foundation Model API (including with an OpenAI-compatible SDK): https://lnkd.in/gmvyF7Cz.

#DBRX is now the best open model on AI2's WildBench, a benchmark of challenging tasks from real LLM users: https://lnkd.in/eUuGFveg. If you haven't tried it, it's easy to try it in this HuggingFace space: https://lnkd.in/ePaWBXR9 Or get started using the Databricks Foundation Model API (including with an OpenAI-compatible SDK): https://lnkd.in/eVQjAquT.

Just got my email for Google Gemini pro 1.5 waitlist. So. Freaking. Excited! In their own words; You now have access to our experimental Gemini 1.5 Pro model in Google AI Studio You now have Private Preview access to our next generation Gemini 1.5 Pro model in Google AI Studio, with Gemini API support coming soon. Gemini 1.5 Pro uses a new architecture to improve efficiency for text and vision understanding. As early testers, you can experiment with an up to 1 million token context window and directly upload massive PDFs, whole code repositories, or even long videos into your prompt. YES THEY SAID 1 MILLION TOKENS!!! *Squeal*!!!

AI is not going to replace humans but humans with AI are going to replace humans without AI #ai #bard #genai

✨🤍 Some useful & Bardtacular tips of using #Bard Google

Beginner's Guide to Vector DBs: Introductory Overview https://lnkd.in/gVkRZnw9   Vector search and vector databases are the backbone of many modern LLMs. The vectors in question are the embeddings. Join this hands-on workshop for developers and AI professionals, on state-of-the-art technology. Recording and GitHub material will be available to registrants who cannot attend the free 60-min session.

Want to create advanced user experiences that harness multiple LLMs, scrape websites, and connect to databases? In this demo by Ania Kubow, you'll see how Langflow's drag-and-drop elements and chat box feature can build a seamless service for users. ⬇️ https://ow.ly/aH3s50SyMIe

We just shared our 2-hour long crash course on AI Engineering for free on YouTube. It'll teach you how to build and deploy a LLM-powered stock analysis app. Here's what you'll learn: * AI engineering basics * The OpenAI API * Models and instances * Prompting techniques * Few shot approach * Fine-tuning * Frequency penalties * Tokens * AI safety 101 * Prompt injections * Cloudflare Workers * AI Gateway * Cloudflare Pages ... and a lot more! Link in comment 👇