Cross-language voice cloning, Plugins for Bing Chat, Building No-code LangChain-based semantic PDF search app, Chatbots by Stability AI and Inflection and more
Greetings and Welcome to this week's AIBrews - your thoughtfully curated guide to the most promising AI products, learning resources and a concise roundup of the week's impactful news. The aim is to provide a balanced selection in the rapidly changing AI landscape, ensuring our readers stay informed without feeling overwhelmed. Please let us know how we can further optimize your experience and save you time.
Without further ado, let's dive in.
In today’s issue:
AI Pulse: : News, Insights and Social Spotlight of the Week
AI Toolbox: Product Picks of the Week
AI Skillset: Learn & Build
🗞️ AI Pulse: News, Insights and Social Spotlight of the Week
🔥 News & Insights
Play.ht has launched its latest machine learning model that supports multilingual synthesis and cross-language voice cloning. This allows users to clone voices across different languages to English, retaining the nuances of the original accent and language [Details].
A new programming language for AI developers, Mojo, has been developed by Modular, the AI developer platform co-founded by Chris Lattner ( he cofounded the LLVM, Clang compiler, Swift). Mojo combines the usability of Python with the performance of C. Up to 35,000x faster than Python, it is seamlessly interoperable with the Python ecosystem [Details | Twitter Link].
Stability AI released StableVicuna, the first large-scale open source chatbot trained via reinforced learning from human feedback (RHLF) . There’s also an upcoming chat interface which is in the final stages of development [Details].
Eleven Labs introduced new speech synthesis model that supports seven new languages (French, German, Hindi, Italian, Polish, Portuguese, and Spanish). This makes it possible to generate speech in multiple languages using a single prompt while maintaining each speaker's unique voice characteristics [Details | Demo video].
New features for AI-powered Bing Chat: richer visuals, long-form document summarization, broader language support, visual search, chat history, sharing options, AI-assisted Edge actions, and contextual mobile queries.
Third-party plugins in Bing chat with more details coming at Microsoft Build later this month [Details].
Debut of ‘Pi’ chatbot by Inflection (founded by co-founders of Google DeepMind and LinkedIn). It’s designed for relaxed, supportive and informative conversations. Pi is free for now without any token restrictions [Details | Chat].
Sal Khan, Khan Academy founder, discusses AI's potential to transform education in a TED Talk, highlighting personal AI tutors, teaching assistants, and new features of their chatbot, Khanmigo [Video].
Salesforce announces Slack GPT - generative AI for Slack. It includes:
An AI-ready platform to create custom workflows and automate tasks via simple prompts, without coding. Users can integrate language models of choice: ChatGPT, Claude, or custom-built ones.
Built-in AI features in Slack, such as conversation summaries and writing assistance.
The Einstein GPT app for AI-powered customer insights from Salesforce Customer 360 data and Data Cloud [Details].
Replit’s new 2.7B params code LLM, ReplitLM is now open-source. It outperformed Codex and LLaMA despite being smaller in size [GitHub | Hugging Face Demo].
Nvidia will present 20 research papers at SIGGRAPH, covering generative AI models for personalized images, inverse rendering tools for 3D objects, neural physics models for realistic simulations, and neural rendering models for real-time, AI-driven visuals. [Details].
Snap plans to show sponsored links to users during chat with its My AI chatbot [Details].
IBM is set to pause hiring for around 7,800 positions that could potentially be replaced by AI and automation [Details].
Box is introducing generative AI tools across its platform, allowing users to obtain document summaries or key points and create content in Box Notes [Details].
Stability AI released DeepFloyd IF, a powerful text-to-image model that can smartly integrate text into images [Details].
Sam Altman and Greg Brockman from OpenAI on AI and the Future in this podcast [YouTube Link]
Researchers at The University of Texas at Austin have developed a non-invasive AI system, known as a semantic decoder. It can convert brain activity while listening to a story or silently imagining telling a story, into coherent text using fMRI scans and transformer model [Details].
HackAPrompt: The first ever prompt hacking competition, with $37K+ in prizes, starting May 5th. Sponsored by OpenAI and others. [Details | Prompt Hacking Tutorial ].
New plugins added in ChatGPT plugin store:
🔦 Social Spotlight
A GPT-4 AI Tutor Prompt for customizable personalized learning experiences [GitHub Link].
Portfolio Pilot: A verified ChatGPT plugin for investing that analyses your portfolio for actionable recommendations [Twitter Link with Demo]
Baby AGIs interacting in the real world via phone using vocode (Open source library for building voice conversations with LLMs) [ Twitter Link]
Data visualization in ChatGPT with code interpreter plugin [Twitter Link]
ThinkGPT, a Python library for LLMs, enables chain of thoughts, reasoning, and generative agents. It addresses limited context, improves one-shot reasoning, and integrates intelligent decisions [GitHub Link].
🔍 🛠️ AI Toolbox: Product Picks of the Week
guidde is an AI-powered tool that creates video documentation with AI-generated narration, storyline, and screenshots. Simply use the browser extension to capture your actions and stop when finished. It records and edits your workflow, producing shareable how-to videos. The extension is available for Google Chrome and Microsoft Edge.
Tavus, an AI video personalization platform to create numerous customized AI-generated videos by recording just one video. Users can add unique variables to their templates, such as company names or personalized intros, to create distinct videos for every viewer. The AI cloning captures emotion as well replicates the way you speak. Another feature is variable backgrounds: automatically includes a scrolling capture of a submitted URL, such as customer website, LinkedIn profile etc., adding a unique personal touch to each video.
FileGPT lets users ask questions in natural language across a range of content types, from text and audio to video, web pages, and scanned handwritten documents. Another useful feature is that it lets you ask questions across multiple files at once and integrates the information in a single comprehensive answer.
📕 AI Skillset: Learn & Build
No-Code: Build a LangChain-based semantic PDF search app with no-code tools Bubble and Flowise. Flowise is an open-source UI visual tool to build your customized LLM flow using Langchain:
GPT-4 - How does it work, and how do I build apps with it? - Harvard University’s CS50 Tech Talk:
Prompt injection explained, with video, slides, and a transcript from a webinar organized by LangChain [Link].
The Full Story of Large Language Models and RLHF by AssemblyAI [Link].
Getting started with generative AI on AWS using Amazon SageMaker JumpStart [Link].
Thank you for reading AI Brews! If you have any thoughts, questions or just want to say hello, please don't hesitate to hit reply. Mariam
Thanks for reading AI Brews! Subscribe for free to receive new posts.