Google Supercharges Bard with Programming Skills ⚡
Plus: Microsoft now has an AI Designer, Transformers with 1M+ tokens capacity, HuggingChat - ChatGPT's Rival, and more!
Hey there 👋
Buckle up! This issue is jam-packed with the latest and greatest in the world of artificial intelligence. We've got news on everything from video object tracking and segmentation to recurrent memory transformers scaling up to 1 million tokens and beyond 🤯
Watch out, OpenAI, because Google is stepping up its game with the impressive new Bard model that can generate code in over 20 programming language. As the battle for AI supremacy unfolds, we can't help but be excited to see the rapid advancements in technology and the potential for increased collaboration and competition in the AI arena.
🍿 So, grab some popcorn and hold onto your hats, because the race between the tech giants is gonna become much more interesting!!
This issue covers:
Latest Developments 🌍
News from the Industry 🧑🏫
Tools of the Trade ⚒️
Hot Takes 🔥
AI Meme of the Week 🤡
Latest Developments 🌍
Our Pick 👌
Track Anything: SAM Meets Videos: Video object tracking and segmentation that allows users to specify what to track and segment via user clicks.
Scaling Transformer to 1M tokens and beyond with RMT: Recurrent Memory Transformer architecture increases the capacity of LLMs to more than 1 million tokens.
Unleashing Infinite-Length Input Capacity for Large-scale Language Models with Self-Controlled Memory System (SCM): SCM enables Large-scale LLMs to process ultra-long texts without modification or fine-tuning.
AudioGPT: Multi-modal model that generates and understands speech, music, sound, and talking head content through spoken dialogue with humans.
Ask-Anything: A video Q&A tool to generate descriptive captions and support conversations in various language styles.
Phoenix: LLM aimed at making ChatGPT accessible across languages and in countries with restrictions on using ChatGPT.
Bark: A multilingual text-to-audio model that generates realistic audio, including nonverbal expressions like laughing, sighing, and crying.
DeepFloyd IF by Stability AI: An open-source text-to-image model that utilizes a frozen text encoder and cascaded pixel diffusion modules to achieve photorealistic results.
AutoNeRF: Trains NeRFs using autonomous agents for efficient exploration and improved downstream task performance.
Segment Anything in 3D with NeRFs: Uses NeRF and SAM to obtain 3D segmentation via one-shot manual prompting in a single rendered view.
HOSNeRF: Reconstructs dynamic human-object-scene neural radiance fields from a single video, rendering all scene details from arbitrary viewpoints.
TextMesh: Generating realistic 3D meshes from text prompts by extending NeRF and improving mesh extraction and texture finetuning.
TextDeformer: Manipulates triangle mesh based on text prompts and produces both large, low-frequency shape changes and small high-frequency details.
Towards Realistic Generative 3D Face Models: A 3D generative face model for detailed editing of 3D rendered faces, outperforms SOTA methods.
Factored Neural Representation for Scene Understanding: Encodes object movement and deformations from monocular RGB-D videos for efficient and editable analysis.
Learning Agile Soccer Skills for a Bipedal Robot with Deep Reinforcement Learning: Humanoid robot trained with Deep RL plays soccer with sophisticated and safe movement skills, exhibiting robust and dynamic behavior.
BadGPT: A backdoor attack that can compromise language models during reinforcement learning fine-tuning, and manipulate the generated text.
A Cookbook of Self-Supervised Learning (SSL): Guide to SSL training presented in a cookbook style, aimed at lowering the entry barrier to this complex area.
Learning to Program with Natural Language: Natural language programming and Learning to Program method improve LLMs' performance in complex tasks.
Enhancing Chain-of-Thoughts Prompting with Iterative Bootstrapping LLMs: An iterative bootstrapping approach that improves LLMs' reasoning performance.
Patch Diffusion: Patch-wise training framework that significantly reduces the training time costs while improving data efficiency for diffusion models.
Speed Is All You Need: Optimizations for fast on-device deployment of large diffusion models on GPU-equipped mobile devices.
Supporting Human-AI Collaboration in Auditing LLMs with LLMs: An auditing tool, AdaTest++, that leverages human-AI collaboration to rigorously audit LLMs.
Evaluation of GPT-3.5 and GPT-4 for supporting real-world information needs in healthcare delivery: Both LLMs can provide safe responses, but not always be useful in meeting specific information needs and require further research.
Adam Instability in Large-Scale Machine Learning: Adam optimization algorithm causes divergent behavior in LLM training by entering a state with uncorrelated parameter update vector.
News from the Industry 🧑🏫
Our Pick 👌
Microsoft has removed the waitlist on Microsoft Designer, and expanded its features including creating high-quality graphics, animations, and captions quickly, with or without design skills.
Google's Bard can now assist with programming and software development tasks, including code generation, debugging, and explanation in over 20 programming languages.
Cohere releases embedding archive of Wikipedia articles in multiple languages for language AI applications.
Hugging Face has launched an open-source alternative to ChatGPT called HuggingChat. You can play with it here!
Stability AI has released its Image Upscaling API that increases image size without compromising their sharpness, providing more detail and texture.
Russia's Sberbank has launched its own chatbot called Gigachat which is said to excel among its peers, including ChatGPT.
Yelp has launched AI-powered search updates, and features like "Surprise Me" for highly rated recommendations, the ability to add videos to reviews.
Google has introduced Security AI Workbench, powered by a its security LLM, Sec-PaLM, to address top security challenges and prevent new infections.
Google is adding generative AI tools to its advertising business, allowing customers to input content that will be "remixed" by AI to generate ads.
OpenAI's ChatGPT now allows users to turn off chat history, is introducing a ChatGPT Business subscription, and has added a chat export option.
Meta plans to introduce AI agents to billions of people through its apps, while also working on generative AI products for release in the coming months.
Atlassian is adding AI capabilities, powered by OpenAI's LLMs, to Jira and Confluence, including features like meeting summarization, tweet drafting, and Jira code writing.
EU lawmakers are proposing a tiered approach to regulating generative AI, with specific requirements for the different types of AI.
OpenAI is trying to trademark "GPT" but faces a long and uncertain process due to queue for similar trademarks and the need to prove that "GPT" is proprietary and not just a descriptive acronym.
Meanwhile, OpenAI has also released Brand Guidelines for using the brand in marketing and communications.
UK government announces £100m funding for a new taskforce to accelerate the development and adoption of safe and reliable foundation models of AI.
PwC is investing $1 billion in generative AI with Microsoft and OpenAI to automate aspects of its services and train staff in AI capabilities.
Pinecone raised $100 million on a $750 million valuation as demand for vector databases grows in AI-driven semantic search and interest in LLMs.
Robust.AI has raised $20M in a funding round led by Prime Movers Lab to expand its autonomous warehouse cart and software offering.
Biden plans to limit American investment in China's high-tech industries, seeking G-7 support at the summit in Japan.
The Department of Homeland Security will create a task force to investigate AI use in protecting the country, including screening for forced labor goods and detecting fentanyl shipments.
Yokosuka, Japan has become the first municipality in the country to use ChatGPT for administrative tasks, to free up humans for one-on-one interactions.
Tech companies are aggressively recruiting rare talent from top AI university programs like Stanford, MIT, and Cornell, decreasing enrollment for AI Ph.D.s.
Greywing has developed SeaGPT powered by GPT-4, simplifies crew changes for maritime crew managers by automating communication and information extraction.
TikTok is reportedly testing a generative AI avatar feature allowing users to create custom profile pictures from a range of styles.
Artifact launched a Summaries Tool that generates AI-powered article summaries, with different styles available for entertainment purposes.
RunwayML has launched its iOS app with Gen-1, its popular tool for video and image transformation.
Tools of the Trade ⚒️
Our Pick 👌
Neural Frames: Animation generator that creates videos, digital art, music videos, object placement and more from text prompts.
Superus: AI chatbot that maps out ideas and creates interactive visual storytelling for better comprehension.
ChainIntelGPT: AI-powered real-time crypto data analysis with natural language search and leading crypto platform integration.
ShopWithAI: AI chat-based shopping assistant that understands your personal style and suggests items.
Sketch AI: Creates digital art based on sketches and text prompts.
DoMyShoot: Studio-quality guided product photography on your smartphone, with photo editing and marketing content generation.
Chat Prompt Genius: Generate high-quality prompts and content ideas for chatbot conversations.
Wodka.ai: Build custom AI chatbots trained on your data sources for 24/7 support and efficient sales assistance.
RecurPost: Social media management tool for content creation, scheduling, analytics, and team collaboration.
Aomni: Uses AutoGPT to plan queries and extract relevant trustworthy information from the internet, without generating false content.
My Approach: Offers tailored expert answers to business questions using AI-powered analysis.
Octie: Write marketing copy 10x faster and generate copyright-free images.
Prowriting: Generate clear, concise and consistent UX copies, saving time and budget and better customer experience.
Aim: Open-source AI metadata tracker that allows for easy observation and comparison of metadata.
Purple Wave: AI-powered digital marketing tools for businesses, like email automation, course creation, funnel building, and page designing.
Alphy: Transcribe, summarize, and question YouTube videos or Twitter Spaces.
Unicorn Platform: Landing page builder with drag & drop functionality and easy customization.
Aistote: Analyzes courses and generates personalized quizzes to help better understand and remember information.
Pictory: Creates short, branded videos from long form content, turns scripts into sales videos, blog posts into videos, and more.
PinkLion: Investment copilot to help manage your investments with features like simulations, asset forecasts, and portfolio analytics.
RealLife3D: Efficiently convert video and still images into 3D at a reasonable cost, using AI.
Plicanta: Simplify job search with a powerful portfolio website, effortless application tracking, and innovative tools.
SpellPrints: Create and monetize AI-powered applications without coding skills, offering creative freedom and various AI models and tools.
Hot Takes 🔥
AI Meme of the Week 🤡
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 FREE learning resources on trending topics like Python, Data Science, Machine Learning, and NLP!