ChatGPT Continues to Amaze as GPT-4 Looms on the Horizon
Salesforce Bulletin Board...oops 🙈 Latest developments in AI
Hey there 👋
It’s that time of the week and your favorite newsletter is here, bringing you some piping hot updates directly from the field.
Can you believe how quickly this technology is evolving? Every week there's something new and exciting to learn about! From new models and methods to innovative applications, there's always something to keep us on our toes.
Well, you don’t have to worry about frisking the internet for your AI scoop, we’re here to keep you in the loop!
This issue covers:
Latest Developments 🌍
News from the Industry 🧑🏫
Tools of the Trade ⚒️
Hot Takes 🔥
AI Meme of the Week 🤡
Latest Developments 🌍
Our Pick 👌
Scaling Robot Learning with Google’s ROSIE: ROSIE uses text-to-image diffusion models for data augmentation to obtain large-scale datasets for robot learning and shows improved robustness and generalization of high-level robot learning tasks.
![[optimize output image]](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F451d5846-1e15-43fb-b17a-c1e322f58ac9_600x338.gif "[optimize output image]")
Stable Diffusion with Brain: A new method to reconstruct high-resolution brain images using functional magnetic resonance imaging.
Google’s USM: A multilingual pre-trained Universal Speech Model achieving SOTA performance in speech recognition and speech-to-text translation tasks.
Microsoft’s BiomedCLIP: A domain-specific adaptation of CLIP achieving SOTA results on biomedical imaging tasks.
Adobe’s CoCs: A new visual representation paradigm that views an image as a set of unorganized points to extract features.
Interactive Text Generation by Microsoft: A new approach that allows training text generation models interactively using user simulators to provide edits, and guide it towards a target text.
OpenAI’s New Family of Generative Models: Consistency models achieving high sample quality without adversarial training.
MobileBrick: A novel data capturing and 3D annotation pipeline using LEGO models for research on 3D reconstruction for mobile AR applications.
Textured Mesh Recovery from NeRF: A new approach for creation of high-quality textured meshes from images using an efficient and adaptive surface refinement process based on a NeRF initialization.
Google’s new Opensource Model Flan 20B: An improved version of the UL2 20B checkpoint model.
PaLM-E by Google: An embodied language model that incorporates real-world sensor modalities into language models, enabling it to perform multiple embodied reasoning tasks.
First Opensource Implementation of LLaMa: ChatLLaMA is fine-tuned for instruction tasks with RLHF, enabling faster and cheaper training and inference processes compared to ChatGPT.
KOSMOS-12 by Microsoft: A Multimodal LLM that can perceive general modalities, learn in context and follow instructions, achieving impressive performance on language understanding, perception-language, and vision tasks.
LLM AUGMENTER System by Microsoft: Improves LLM's ability to use external knowledge and reduce hallucinations.
Google’s InfoSeek: A Visual Q&A dataset to evaluate pre-trained vision and language models' ability to answer visual information-seeking questions.
Google’s Research on LLMs’ In-Context Learning: LLMs can override their pre-existing knowledge to learn new things in context, and their ability to do so improves with scale.
News from the Industry 🧑🏫
Our Pick 👌
Salesforce and OpenAI have launched the ChatGPT app for Slack, which provides instant conversation summaries, research tools, and writing assistance directly in Slack.
GPT-4, the highly anticipated language model, is rumored to be released next week as per the CTO of Microsoft Germany, giving users a sneak peek of AGI and its multimodal abilities.
Salesforce Ventures has launched a $250 million generative AI fund to support responsible AI development, with initial investments in Anthropic, Cohere, Hearth.AI, and You.com.
Microsoft has launched Copilot for its business apps, with capabilities such as automating repetitive sales and customer service tasks, generating content ideas and assisting with e-commerce product listings.
Salesforce’s Einstein GPT is the first generative AI CRM technology that offers personalized and real-time responses using AI-generated content.
Discord is experimenting with OpenAI-powered chatbots, conversation summaries, and an AI-powered AutoMod for server moderation.
Microsoft's Florence AI system generates alt text for Reddit images and offers image and video capabilities for Azure Cognitive Services.
MIT researchers have developed Sybil which can assess a patient's risk of developing lung cancer based on CT scans, achieving strong predictive power.
Brave Search has launched Summarizer, a new feature that provides plain-written summaries at the top of search results, powered by its own LLMs.
Marc Andreessen's venture capital firm leads over $200m investment in Character.ai, a generative AI chatbot company, valuing it at $1bn.
Ford has established Latitude AI, a subsidiary dedicated to developing a hands-free, eyes-off automated driving system for next-generation Ford vehicles.
Stability AI, the parent company of Stable Diffusion, is seeking to raise funds at a valuation of about $4 billion after just reaching unicorn status in a funding round last year.
Tools of the Trade ⚒️
Our Pick 👌
Multi Model Playground: All the big language models under one roof for the very first time. Compare the output of OpenAI’s ChatGPT, Anthropic's Claude and Cohere's language model in a single playground.
ChatLLM by Abacus AI: Train a custom chat AI model to remember context and provide accurate responses at a low cost.
Lasso AI: Process automation solution for streamlining tasks using attended and unattended bots with platform integrations.
Adrenaline: AI-based code debugging tool to instantly fix errors and assist with debugging.
Infomail: AI-powered email marketing tool for generating engaging content in over 50 languages and creating unique templates quickly.
Hubble: No-code platform for creating AI-powered applications quickly using advanced AI models like GPT-3 and connecting to data sources.
AI Lawyer: Provides instant legal help, document creation, and review with accessible, private, and cost-effective solutions.
Socialvar: An automated marketing solution offering a full-stack social media platform, WhatsApp automation, and email/SMS marketing.
Autopia: Streamlines agile workflows with AI, allowing teams to efficiently manage sprints, tickets, and stakeholder meetings in JIRA.
Wordmetrics AI: An SEO writing platform that helps content marketers improve their content's visibility by offering real-time guidance.
Hot Takes 🔥
ChatGPT FTW 🏎️
Free Knowledge - Accessible for ALL 📚
Race to AGI 🏃♂️
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 $39 USD worth of learning resources on trending topics like Python, Data Science, Machine Learning, and NLP!