Adobe Fills the Generative AI Gap 🎥
Plus: Microsoft's Dev Conference Highlights, Videos Generated from Brain Activity, Meta's MMS Models with 1100+ languages, and more!
Hey there 👋
Microsoft took the center stage this week with its annal Microsoft Build developers conference 2023 where it unleashed a tidal wave of updates and new launches, especially some mind-boggling AI innovations! Here’s what went down in the AI space at Microsoft’s extravaganza:-
Microsoft Fabric: An end-to-end analytics platform that integrates various data and analytics tools, providing a unified experience for organizations.
Power BI: Latest updates include Copilot for faster data insights, Direct Lake storage mode for accessing massive data, and more.
Microsoft Q&A Assist: Obtain accurate answers to technical queries using AI that assists in formulating clear questions, detecting duplicate inquiries.
Integrated with Microsoft 365 Copilot, brings work-based capabilities to the browser's sidebar.
Microsoft Edge for Business - new browser that enhances privacy and security for organizations by separating work and personal browsing while providing enterprise controls.
Developers can enhance the browsing experience and increase app discoverability by adding existing PWAs to the sidebar in Microsoft Edge with just a few lines of code.
Developers can integrate their apps and services into Microsoft 365 Copilot using plugins, with over 50 partner plugins.
Microsoft Teams introduces AI libraries, Live Share enhancements, ISV improvements, Avatars, and immersive spaces for collaboration and app development.
Copilot in Power Pages: Revolutionizes the process of building and launching data-centric business websites, generates text content, building forms and chatbots.
Microsoft adopts OpenAI's open plugin standard, enabling plugin interoperability across ChatGPT, Bing, and other Microsoft platforms, with Bing being the default plug-in.
Microsoft Store powered with new features including an AI Hub for showcasing AI experiences and AI-generated review summaries.
Windows Copilot, a centralized AI assistant incorporating Bing Chat, web knowledge, and contextual information, available for preview in June.
Enhanced Windows for developers with Dev Home (streamline setup, GitHub integration, and central dashboard), smarter Windows Terminal with GitHub Copilot X, and additional developer-centric features.
This issue covers:
Latest Developments 🌍
News from the Industry 🧑🏫
Tools of the Trade ⚒️
Hot Takes 🔥
AI Meme of the Week 🤡
Latest Developments 🌍
Our Pick 👌
Mind-Video: A method that reconstructs videos from brain activity using techniques like masked brain modeling, multimodal learning and co-training with an augmented Stable Diffusion model to reconstruct high-quality videos from continuous fMRI data.
Text2NeRF: Generates realistic 3D scenes with complex geometries and detailed textures from text using NeRF and monocular depth estimation.
Video-ControlNet: A controllable text-to-video diffusion model that generates high-quality videos with fine-grained control.
RecurrentGPT: Language model that generates arbitrarily long text interactively by simulating the recurrence mechanism of RNNs.
Cross-Lingual Supervision improves LLMs Pre-training: Combining self-supervised language modeling with supervised machine translation improves LLMs' pre-training by incorporating cross-lingual supervision.
Language Model Hallucinations Snowball: Language models justify false statements by producing additional incorrect claims.
CoDi: Generative model enabling parallel generation and conditioning of multiple modalities.
PEARL: Enhances LLMs' reasoning over long documents via decomposition, planning, and execution of questions.
QLoRA: An efficient finetuning method that reduces memory usage, allowing to finetune a 65B parameter model on a single 48GB GPU.
Training Diffusion Models with Reinforcement Learning: Optimizes diffusion models without additional data or human annotation.
CRITIC: Allows LLMs to self-correct by interacting with external interactive tools, improving their performance through external feedback.
AudioToken: Adapts text-conditioned diffusion models for generating images from audio recordings, using a new token as an adaptation layer.
Less Is More for Alignment: Minimal fine-tuning data is sufficient for high-quality LLM output.
Pengi: Innovative audio language model that combines audio and text inputs to generate free-form text.
WebGUM: Instruction-following multimodal agent for autonomous web navigation with improved perception and reasoning.
Goat: A fine-tuned LLaMA model, outperforms GPT-4 on arithmetic tasks SOTA accuracy and near-perfect performance on large-number operations.
Controlling the Extraction of Memorized Data from LLMs via Prompt-Tuning: Strategies to control extraction rates and privacy-utility trade-offs.
PandaGPT: Model that can process and connect information simultaneously from various modalities such as text, images, videos, audio, depth, and thermal.
CodeCompose: AI-powered code authoring tool deployed at large-scale at Meta, with improved code generation, increased documentation, and API discovery.
News from the Industry 🧑🏫
Our Pick 👌
Adobe has introduced Generative Fill in Photoshop that allows users to create, modify, and replace images using simple text prompts, powered by Adobe Firefly.
Meta has introduced Massively Multilingual Speech (MMS) models that support speech-to-text and text-to-speech for over 1,100 languages, aiming to preserve language diversity.
Anthropic secures $450M in Series C funding to scale its AI assistant and research, with support from tech giants like Google, Salesforce, and Zoom.
Neeva, company that built ad-free and private AI-powered search engine, is shutting down its search product due to difficulties in acquiring users and creating a sustainable business.
OpenAI leaders propose the establishment of an international regulatory body, similar to the IAEA, to oversee the development of AI and ensure safety standards are upheld.
Elon Musk believes AI could potentially become humanity's all-powerful caretaker, imposing strict controls on humans and computing and weapon systems as an "uber-nanny."
Apple is actively recruiting generative AI experts to join its teams, indicating the company's interest in leveraging generative AI for its products.
Google agrees to collaborate with EU lawmakers on an "AI Pact," a voluntary set of rule/standards for AI while formal regulations are being developed.
Google DeepMind's visual language model, Flamingo, can now generate video descriptions of YoutTube shorts based on the initial frames, improving categorization and search results for viewers.
OpenAI-backed startup, 1X, beats Tesla, deploys AI-enabled humanoid robots in real world as security guards and planning further deployment in hospices and assisted living facilities.
China's AI bot Ernie refuse to address controversial topics such as COVID-19's origin and ban users for asking “bad” questions about President Xi Jinping.
Opera has unveiled Aria, an AI-integrated browser, with features such as generating text or code, searching the web, and answering product queries.
Google introduces Product Studio on Merchant Center to help small businesses create unique product imagery using AI and simplify listing on Google.
Tools of the Trade ⚒️
Our Pick 👌
Nexus by Clay: Network navigator powered by AI, saving time and enhancing relationships with personalized assistance and seamless networking solutions.
WizAI: AI-powered WhatsApp chatbot with unlimited messaging, (coming soon) GPT-4 integration, plugins, image creation/recognition.
FineVoice: Enhance your live streams, podcasts, meetings, and online teaching with voice effects, lifelike voiceovers, and customizable soundboards.
Video Highlight: Video summarization and note-taking, accelerates research by removing transcription and allowing effortless exploration and analysis.
SupportGuy: Always-on, always-available AI-powered chatbot for 24/7 customer support, never miss a customer inquiry again.
Collider AI: Boosts digital sales with an AI that generates personalized ads, webpages, and emails, continuously learning and adjusting itself.
Waitlyst: Accelerate software company's growth with autonomous AI agents that drive revenue, retention, and personalized customer engagement.
KAI: Supercharge your iPhone keyboard with AI assistance to enhance writing, creativity, and save time.
AI Diary: Your personal writing companion that understands and engages with you, offering smart suggestions, and insightful features.
TaxGPT: Automate your tax filing, maximize deductions, save time and money.
Shortform: Instantly summarize web and YouTube content, gain context, counterarguments, and grasp the best ideas faster.
LLM Report: Monitor OpenAI API usage with a hassle-free dashboard, direct data retrieval without any installation required.
Pigro: Smarter text chunking and generative indexing for accurate document retrieval.
Modly AI: Your all-in-one assistant for saving time, increasing efficiency, and achieving success with AI-powered tools across multiple roles.
Steve AI: Create live action and animation videos using text effortlessly, with AI selecting the perfect media assets for your video.
Wudpecker: Streamline meetings, automate notes, and collaborate seamlessly for productive teamwork.
ParallelGPT: Streamline ChatGPT tasks with bulk processing, low-code workflows, and secure collaboration on Google Cloud.
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.