Mozilla.ai’s Post

Stack Overflow is selling your answers whether you like it or not 🤷🏽♂️ Stack Overflow announced its partnership with OpenAI earlier this week to improve its generative AI models' performance on programming-related tasks. This means that all of you who have contributed to the platform to help other devs, your content will be sold and used to train data. That's not even the worst part - 'right to forget' has been removed from the T&Cs, and you can't even delete your own content 💀 This is a great way to undermine and disrespect your contributors and probably not just the end of Stack Overflow, but could be the end of the open internet as we know it. I can imagine we'll be seeing more platforms or open-source resources realising that the data on their platform is worth a lot more than they had initially thought. This raises a really important topic of data ownership and content ownership, but that's for a different post. Read more here:

#oss #llm #engineering #platform #observability #metrics #tracing #evals #prompts #ai #datasets #langchain #openai Open source LLM engineering platform: Observability, metrics, evals, prompt management, playground, datasets. Integrates with LlamaIndex, Langchain, OpenAI SDK, LiteLLM, and more https://lnkd.in/gwWrTRC9

I have been trying to get up to speed on the state of open source in GenAI -- not in the model space, but in applications and platforms. A few notable discoveries lately include: - Flowise, a low-code app that appears to do in a lovely UI what LangChain does in python (possibly using LangChain behind the scenes) https://flowiseai.com/ - LibreChat, a chat frontend with model selection (and history search!) https://docs.librechat.ai/ - OpenRouter, an amazing project that brokers just about every model and service under the sun in a unified API (I swear I had this idea...gotta move fast!) https://openrouter.ai/ Thanks to Peter Salvitti and David LaPorte for the recs! Finally, I'm excited about this repo from AWS (h/t to James Smith for the heads-up), which seems to check a lot of boxes that currently interest me: a multi-modal, multi-model chat frontend with the ability to create distinct RAG bots. Looking forward to giving it a try. https://lnkd.in/gAiicQqJ #ai #highered #artificialintelligence #genai #rag

During our discussion with Nathan Wrigley about AI and how it will affect content creators, I said that some questions do not have a single answer and the right decision may depend on many factors. That's why even after ChatGPT and other tools became a part of our toolset, it's still important to hear real people's opinions, especially when these opinions are based on their real experience. It seems that Stack Overflow tries to focus on this specific advantage of the platform. Now, they're experimenting with a new feature called Discussions. As the official statement says, "Discussions can provide a fluid, conversational way for users of all skill levels to tap into the collective knowledge of the community. They can explore technical approaches, share different perspectives, and make more informed technical decisions." However, it looks like the new feature has not become too popular yet. I just checked the recent Discussions and most of them don't even get more than 100 views. Anyway, I'm looking forward to seeing how everything turns out. Did you already hear about Stack Overflow Discussions? Would you consider using this new feature? https://lnkd.in/dFmnH4mV #stackoverflow #developertools #development #webdevelopment #programming #coding #computerscience

Imagine a JavaScript module for easily interacting with AI chat models like Google Gemini, Anthropic Claude & Cohere command-r all using the same familiar OpenAI format. No dependencies. Just a simple interface and adapters for each platform. I made it for Smart Connections. It adapts the API requests at runtime. This way ancillary components can be designed to work with a single API format and still work with multiple platforms. I'm thinking about making it open-source. Is this something that you'd be interested in? Let me know 🌴

🔥 Excited to share my latest project: YTVideoAnalyser! 🎉 I've built a comprehensive application that analyzes YouTube videos using advanced Large Language Models (LLM) and Retrieval-Augmented Generation (RAG). The frontend is developed in React.js, while the backend is powered by Django, interacting seamlessly with OpenAI APIs and Pinecone for vector embeddings. 👨💻 How it works: 1. Input your desired YouTube video link. 2. The app analyzes the video. 3. Redirects you to a chat page where you can ask detailed questions about the video. The AI responds based on the video content, ensuring in-scope responses only. For the demo, I’ll be showcasing the analysis on the latest Windows CrowdStrike issue. 🔗 Special thanks to Santiago Valdarrama for his invaluable LLM videos and @underfitted on YouTube. Check out the project and watch the demo video here: https://lnkd.in/dw_PdFVr #AI #MachineLearning #SoftwareDeveloper #YouTubeAnalyzer #FullStack #OpenAI #Pinecone #ReactJS #Django

It's the biggest Universe yet, and GitHub made a series of huge announcements this morning, led by partnerships with Google and Anthropic, bringing new models to GitHub Copilot, and the technical preview of GitHub Spark, a new AI-powered natural language app builder. Check out CNBC's piece for a great overview, with more news in the comments!

GitHub expands AI suite: Copilot goes multi-model, Spark debuts for web dev GitHub is marking a significant evolution of its developer tools with two major announcements: a multi-model upgrade to Copilot and the introduction of Spark, a natural language web development platform.  The code hosting giant is expanding Copilot beyond its OpenAI-only architecture to include Anthropic’s Claude 3.5 Sonnet and Google’s Gemini 1.5 Pro models while simultaneously launching GitHub Spark to simplify web application development through natural language instructions. The expansion includes additional integration news as Copilot arrives on Apple’s Xcode in public preview. Since Copilot’s initial release, this three-pronged update represents GitHub’s most ambitious AI toolkit expansion. Source- https://lnkd.in/gp7tNU23

Right now I'm tinkering with Next.js and watching some tutorials. The first one is Javascript Mastery's tutorial, Promptopia. Here you can sign in with your google account, create new AI prompts give them a tag and post it. In your profile you can edit and delete them. The original project doesn't cover loadings and if you refresh the page meanwhile you are on your own profile page, the prompts won't load until you visit the home page. But still, it's an interesting first start. Also this is my very first attempt to create a serverless project. Until now I created my own database and server in a separate docker container. https://promptopia.jez.hu/

I've initiated a project to build an AI article summarizer using React and OpenAI's GPT-4 API. Starting with React, I've completed the interface, focusing on a minimalist and intuitive interface. The UI features an input field for article URLs, designed for simplicity and usability. Next, I'll integrate the GPT-4 API for article summarization, aiming for seamless functionality. Stay tuned for further updates on this exciting project! Check out the progress of my project till now on my GitHub- https://lnkd.in/gRPnUEhV #webdevelopment #React #OpenAI #GPT4 #AI #Summarizer #vite #reduxtoolkit