Paul Mills’ Post

Ever wish you could live chat with Wikipedia? Then say hello to Wikichat — a RAG chatbot built using Astra DB, LangChain, and Vercel. 🤖💬 What if you could use GenAI to converse with Wikipedia and ask it about the very latest updates on your favorite topics? Let’s go! In this YouTube livestream, you will learn how to use Next.js along with Vercel’s AI SDK and LangChain to build and deploy a production-ready RAG (retrieval augmented generation) chatbot that can discuss the latest real-time changes in thousands of the most popular Wikipedia pages. Join us on January 24th at 1130PM India time for a live, 45-minute session where we’ll show you how to: Ingest real-time data using the Wikipedia API Chunk and embed the data using LangChain and Cohere Build a simple chatbot using the Vercel AI SDK and Next.js Build intelligent RAG prompts to utilize real-time Wikipedia data Create relevant vector queries of Wikipedia articles and updates To wrap up, we’ll host a live Q&A where you can ask technical questions. Don’t miss this chance to elevate your skills in building state-of-the-art RAG apps. #DataStax #VectorDB #RetrievalAugmentedGeneration https://ow.ly/4SlT50QtTpK

Have you been thinking of using AI within your business? A simple, effective way to start is reducing the time required to gain valuable insights through your documentation. Take a task that requires lots of manual search, using Retrieval Augmented Generation (RAG) get your Language Model of choice to pull the info required. Using AstraDB, LangChain and Vercel we've built a WikiChat to store the vector embeddings to demonstrate just how powerful this can be. 10,000 recent Wiki pages are available to query, siting the page source that the information was retrieved from. Now lets talk about why this is so powerful for business... Replace the wikipedia resources with: - Legal documents 📚 - Medical documents 🏥 - Candidate information 👩🔧 - Technical specifications 🚁 Improve your workflow and reduce the time it takes to gain valuable insights from pools of complex resources (not just documents, but videos and audio files too). Gaining the edge with AI requires you to take to start with simple tasks. We're here to reduce the friction and get you into production. Message me direct to find out more, or register with AstraDB and get started today. https://lnkd.in/eZWCXFNi

#Topics Midjourney Alpha is here with AI image generations on the web [ad_1] Are you ready to bring more awareness to your brand? Consider becoming a sponsor for The AI Impact Tour. Learn more about the opportunities here. Midjourney, among the most popular image-generating AI services in the world today with more than 17.5 million users in its server on the separate messaging app Discord, has begun testing an “alpha” version of its website, midjourney.com, that will allow users to generate imagery directly there instead of using Discord. The Midjourney Alpha website with an AI image generation field was first spotted yesterday by avid users such as Nick St. Pierre (@NickFloats), who shared screenshots and videos of it on X. Midjourney Alpha is officially here!!If you've generated 10k or more images you should have access. This is still the alpha, so things may change a bit. But here is a quick look at image creation on the web. BYE BYE DISCORD ? https://lnkd.in/giT28Rbj Nick St. Pierre (@nickfloats) December 12, 2023 Midjourney today confirmed today in its Discord announcements channel that it had “begun testing an alpha version of the website which includes image creation,” however, the company noted the alpha release was limited to “people w...

#Topics Midjourney Alpha is here with AI image generations on the web [ad_1] Are you ready to bring more awareness to your brand? Consider becoming a sponsor for The AI Impact Tour. Learn more about the opportunities here. Midjourney, among the most popular image-generating AI services in the world today with more than 17.5 million users in its server on the separate messaging app Discord, has begun testing an “alpha” version of its website, midjourney.com, that will allow users to generate imagery directly there instead of using Discord. The Midjourney Alpha website with an AI image generation field was first spotted yesterday by avid users such as Nick St. Pierre (@NickFloats), who shared screenshots and videos of it on X. Midjourney Alpha is officially here!!If you've generated 10k or more images you should have access. This is still the alpha, so things may change a bit. But here is a quick look at image creation on the web. BYE BYE DISCORD ? https://lnkd.in/giT28Rbj Nick St. Pierre (@nickfloats) December 12, 2023 Midjourney today confirmed today in its Discord announcements channel that it had “begun testing an alpha version of the website which includes image creation,” however, the company noted the alpha release was limited to “people w...

Agents to do tasks autonomously using your browser. MultiOn is an excellent example of how agents can do tasks using your browser with a simple goal. The agent will understand the request, go to the web page, understand the structure and start breaking down tasks to achieve the goal. These agents work using a mix of vision models, DOM, and browser automation. This agent is different from other browser automation tools in that it is generic and not trained on a single website or platform. It can be used for any task. Have you tried any browser automation agents before? What was your experience? #ai #agents #llm

You can get PMF and then lose it. We did. Our market changed A LOT since last September. I haven't factored that in yet. Here's what happened: Feb 2023: ChatGPT Plus launches Mar 2023: GPT-4 launches, the world goes crazy for Prompt Engineering May 2023: most Plus subscribers have access to Plugins Nov 2023: Custom GPTs announced, the world goes crazy for "new apps" Jan 2024: GPT store launched April 2024: Plugins ded May 2024: GPT-4o launched, now free for all We launched our existing marketing strategy in September. Novelty was still huge. Plugins were abundant and people were going crazy with the possibilities. The market was in early adopter mode. Our entire market was about 1million Plus subscriber worldwide and we had a direct access to them through any plugin developer. We were the first company to run an ad natively in a GPT Plugin response. Then OpenAI killed the Plugin market and moved over to GPTs...but it's not that simple anymore. User number went from 1million to 200million. Everyone has access to the full set of features now, not just Plus subscribers. Imho plugins were superior for custom GPTs in market fit. Custom GPTs are still superior from a technical standpoint. I get it. Yet, now we're not in early adopter mode anymore. We're in the chasm. (There's a brilliant book by Geoffrey Moore on it called Crossing the Chasm, read it) It's not an isolated, unique market with crystal clear reach anymore. We're still looking for the partners that allow the same reach plugins allowed us before. Not to mention, that the novelty is wearing off. Prompting in itself is not that interesting anymore. You need use cases. Tell me what I can do with this thing that I haven't tried before. We've been talking about how to use AI not what to use it for. That's probably a shift in comms we need to make. (and solve reach, obviously)

We recently introduced the "RequestBot - An Ask AI" functionality powered by Writesonic, which has changed our product-building process. Read 👇 We are discovering - Ideas around the feature requests - Some unknown bugs - Gaps in documentation ------ Here are some popular feature requests we got 1. Ability to manipulate cookies in HTTPs requests 2. Support for Intercepting HTTP Traffic from Android & iOS Apps 3. HTTP Traffic Inspector in the Chrome extension 4. Block Requestly on some websites 5. Breakpoint support --------- Here is a sneak peek of the designs of the "HTTP Traffic Inspector" feature. - Go to Network Inspector - Select Tab(s) to inspect the traffic - Filter traffic by tab/domain URL - Supports graphQL operationName in the traffic table Github Issue - https://lnkd.in/gXki3Pmv ------- We're already working on a 1-Click setup to Inspect & Modify traffic in Android Emulators. Launching by the end of this week! https://lnkd.in/ggyuykew -------- The development of "Blocking Certain Websites in Requestly" is underway and should be available by the end of this week. https://lnkd.in/gWdpZiJu ----- These are all incredibly powerful signals for building a good product, and AI-based conversations make it possible. Users can ask anything with AI without any sense of judgment. Here's the process we are following ↓ 1. Use Botsonic (or a similar tool or build your own CustomGPT) 2. Integrate into your product 3. Get the feedback and feed it back into ChatGPT 4. Ask for Insights 5. Manually Review the answers & Improve the source of your content Hope it helps everyone build a good product 🚀

Check out our GenAI based notification tool!

🤖 Introducing the Wikipedia Chatbot, your new go-to resource for instant information and knowledge exploration! 📚💡 Tired of sifting through endless search results for answers to your burning questions? With our innovative Wikipedia Chatbot, you can now access accurate and reliable information in a snap. 🔥 What's more, our Chatbot offers the convenience of pre-prompts, allowing you to select from a range of common queries or topics to streamline your search process even further.🚀 Whether you're a student conducting research, a professional seeking quick facts, or simply a curious mind hungry for knowledge, our Wikipedia Chatbot is here to make your quest for information effortless and efficient.🌟 Try it out now : https://lnkd.in/gFv-2p9m For more details : https://shorturl.at/gtQ45 Visit Lyzr : https://www.lyzr.ai/ Embrace the future of information access and discovery with the Wikipedia Chatbot. Join us in revolutionizing the way we learn! 🚀🌐 #Wikipedia #Chatbot #Knowledge #AI #Innovation #InformationAccess #Learning #Technology #GenAI #LyzrAI

I wanted to share this excellent 4-box that my friend & former colleague, Leo Ryan recently published. The distinction between "Intelligent Augmentation" and "Intelligent Automation" feels healthy. His model tallies with a conversation I recently had with Maximilian. What I took from that: #1. Only use GenAI for stuff that other tools in the workflow can’t perform. So, f’rinstance, if you’ve a good scraper for data extraction, or a good ML model for topic identification or clustering or text recognition, use those instead. Minimise your exposure to GenAI to the bits where it brings value — the unknowns will always be too unknown. Instead, look at what it can genuinely add, and use it as a component, never the whole machine. #2. Don’t use it for generating “content” — if anything, this is its flashiest, but weakest use case. Can't recall Max's exact words, but he was at his most scathing. This chimes with something we should all know: the most successful algorithms over the past 20 years have all been about 𝘳𝘦𝘥𝘶𝘤𝘪𝘯𝘨 our content exposure. For the Toffler fans out there, we’ve been in Information Overload since before 1970. Meta, Google, TikTok the iOS App Store etc. are all about limiting our exposure to the vast oceans of news and crap so we only see the info/crap that’s most likely to stimulate/reward our attention. Creating “more crap” in the hope we’ll get free carry in email, search and social is an arms race* against the best-funded, best-staffed crap filter organisations in the world. (*It works, but only in the way that occasional zombies in zombie swarms manage to breach the perimeter. Stop building zombies).