GPT-4 Takes the AI World by Storm 💥

Plus: Google launched new generative AI features 🪢 Microsoft follows up with Copilot for Work.

Hey there 👋

Welcome back to Unwind AI, your weekly dose of AI news and insights. We hope you are having a great week so far. We’re so thrilled to be featured in the 5 must-follow AI newsletters by Datacrunch! Can’t thank you enough for always supporting the publication. We promise to deliver the latest and best AI content to you to keep you on the edge.

The world of AI is constantly evolving, and this week is no different. We saw some groundbreaking developments like the release of GPT-4 by OpenAI, Google and Microsoft rolling out new AI-powered features, and much more.

Excited to explore? Buckle up and let’s get the ride rolling!

This issue covers:

1. Latest Developments 🌍

2. News from the Industry 🧑‍🏫

3. Tools of the Trade ⚒️

4. Hot Takes 🔥

5. AI Meme of the Week 🤡

Latest Developments 🌍

Our Pick 👌

Can’t be any other!

GPT-4 by OpenAI: OpenAI has released GPT-4, a large multimodal model that accepts image and text inputs and emits text outputs, outperforming other LLMs and SOTA models in various language and reasoning tasks, and demonstrating human-level performance on various professional and academic benchmarks. You can join the GPT-4 API waitlist to get access to this amazing model or play with it immediately with ChatGPT plus.

• Stanford’s Alpaca: A language model fine-tuned on instruction-following demonstrations, similar to OpenAI's text-davinci-003 but cheaper to reproduce.

• PyTorch 2.0: A faster, more Pythonic and Dynamic release with new features for improved performance and training optimization for GPUs and CPUs.

• Assembly AI’s Conformer - 1: A SOTA speech recognition model that leverages Transformer and Convolutional layers, achieving near-human level performance after training on 650K hours of English audio data.

• Midjourney V5: Latest Midjourney model with improved language understanding, accuracy, and aesthetic flexibility, producing photorealistic images.

• Visual ChatGPT by Microsoft: Enables users to communicate with ChatGPT using both language and images, incorporating various Visual Foundation Models.

• Meta AI’s MuAViC: First audio-video speech translation benchmark, containing roughly 1,200 hours of transcribed data spanning 9 languages.

• MathPrompter: Improves the accuracy and confidence of LLLMs in solving arithmetic problems.

• Preference Transformer: A neural architecture that models human preferences for RL, capturing temporal dependencies in human decision-making.

• Cones: Concept Neurons in Diffusion Models for generating related concepts in a single image with reduced storage consumption.

• Microsoft’s FoundationTTS: A speech synthesis system that uses a neural audio codec for discrete speech token extraction and waveform reconstruction.

• Self-Learning Agent for Performing APIs (SLAPA): A self-learning AI that can teach itself how to use APIs in real-time, improving its ability to write API calls.

• Adobe’s Video-P2P: A framework for video editing that uses Text-to-Set model to invert videos and a decoupled-guidance strategy to allow for detailed editing.

• GigaGAN by CMU and Adobe: A new GAN architecture that can successfully scale up text-to-image synthesis and offers faster inference time, high-resolution image synthesis, and supports latent space editing applications.

• 3D Cinemagraphy by Adobe: Generates video from a single still image by animating the scene in 3D space using layered depth images, motion estimation, and displacement synthesis.

• ViperGPT: Enables composition of vision-and-language models into subroutines to generate Python code for complex visual tasks, without additional training.

---

News from the Industry 🧑‍🏫

Our Pick 👌

Google has introduced new generative AI features into Google Workspace, and new APIs including PaLM API and MakerSuite, to make it easier and safer for developers and businesses to build with Google's best AI models through Google Cloud.

• Microsoft has launched Microsoft 365 Copilot, an AI-powered tool that combines LLMs with Microsoft 365 data to assist users across various applications.

• Microsoft's Bing chatbot now offers three response modes, creative, balanced, and precise for improved accuracy and creativity.

• Anthropic raises $300 million with Spark Capital leading the round, to continue collaborations with companies such as Google and DuckDuckGo.

• Sam Altman, the CEO of OpenAI, invests $180 million in Retro Biosciences, a startup that aims to add 10 years to the average human life span through anti-aging biotech.

• ChatGPT is now available in Azure OpenAI Service for developers to integrate AI-powered experiences into their own applications.

• OpenChatKit is an open-source alternative to ChatGPT, with a 20B chat-GPT model, customization recipes, retrieval system, and moderation model.

• Grammarly launches GrammarlyGO, a generative AI tool for content creation that can adapt to user's writing style.

• Microsoft lays off its ethics and society team working for responsible AI.

• Adept raises $350M in Series B funding to further develop AI systems prioritizing user interaction and collaboration.

• French start-up Nabla launches Copilot for doctors, which uses GPT-3 to transcribe and repurpose patient consultation information.

• Microsoft has confirmed that the new Bing search engine is running on OpenAI's GPT-4, providing users with the most comprehensive search results.

• Anthropic has launched Claude, a next-generation AI assistant capable of a wide variety of conversational and text processing tasks, with two versions available, Claude and Claude Instant.

---

Tools of the Trade ⚒️

Our Pick 👌

Metaview automatically writes your interview notes for you so you can save time and focus on high-quality interactions with candidates. It can:

• Automatically generate concise, accurate bullet points from candidate’s responses to each question,

• Allow user to edit the generated notes to add comments and context,

• Save time and make the hiring process 10x more efficient.

• Radicalbit: Event stream processing and machine learning products, including Helicon, a codeless data platform for AI & MLOps, enabling real-time decision support systems and situational awareness for businesses.

• Knowmax: AI-powered knowledge management platform to simplify processes, enable self-service, and provide cost-effective training and seamless chat handling.

• LensAI: Contextual computer vision ad solution that matches ads with relevant visual content to offer an instant seamless sale experience.

• Listingcopy.ai: AI-based real estate listing description generator that creates accurate and persuasive ad copy for real estate professionals.

• Dispute AI: Credit repair tool that uses AI to automatically analyze and dispute credit report errors across all 3 bureaus, with coaching and tracking features to help you improve your credit score.

• Phenaki: A model that generates videos from text prompts and still images, with interactive examples.

• Chatfuel AI: Chatbot builder to automate customer communication, and boost marketing and sales.

• Trellis: AI-powered educational tool with a Socratic guide, quizzes, and simplified explanations for hard sciences, engineering, code, and math.

• Dimeadozen.ai: Business validation tool that uses AI to instantly assess the viability of your business idea and provides reports on market research, launch and scale strategies, and fundraising options, and more.

• KindleGPT: Search and chat with Kindle book highlights, and export Kindle notebooks as CSV files.

---

Hot Takes🔥

YK aka CS Dojo 📺🐦 @ykdojo

Googling is a skill, and so is ChatGPT-ing.

Andrew Mayne @AndrewMayne

I started as a prompt engineer at OpenAI in 2020. With ChatGPT & GPT-4, so much has changed. We’re now in the next phase. The term “prompt engineer” is feeling as antiquated as “typist” or “computer operator.” We’re all prompt engineers now.

Bojan Tunguz @tunguz

Hot take: Generative AI will commoditize *all* text, image, sound, and video data, which in turn will make *all* tabular data even more valuable.

Keerthana Gopalakrishnan @keerthanpg

@sama $23 is a lot of money in India. A chatbot that costs 1/17th of median monthly income is kinda expensive :)

---

AI Meme of the Week 🤡

Source: Ben Tossell

That’s all for this week!

Will see you next Saturday with more such content. Don’t forget to subscribe and give your feedback below.

BONUS 🎉

Share this newsletter with three other friends and stand a chance to win my book GPT-3: The Ultimate Guide to build NLP Products with OpenAI API. Winners will be selected on a monthly basis.

🎁 Every paid subscriber will also receive $39 USD worth of learning resources on trending topics like Python, Data Science, Machine Learning, and NLP!