Top 5 Stories of the Week: Universal Forecasting

Friends, to wrap up this week and get ready for the weekend, we want to bring you the top 5 stories our readers clicked on to explore more about over the past 7 days. Was it OpenAI’s Sora video generation? Nope! As fascinating as that is, what you clicked to explore more of were the following…

As always, these are sourced from a data-driven list of the news that the AI community is most engaged with each day. Curation is our super-power over here, and if you find it valuable - we hope you’ll think about who else in your life might appreciate these stories and forward them this newsletter.

Enjoy digging into any of these hits you missed over the week, and we’ll see you next week.

-Marshall Kirkpatrick, Editor

#1. Salesforce's AI Team Announces 'Moirai', a Universal Forecasting Model

First impacted: Salesforce AI users, Data Scientists
Time to impact: Medium

Salesforce AI Research has announced the launch of Moirai, an AI model trained to do universal forecasting. They say it was trained on Large-scale Open Time Series Archive (LOTSA), a massive open time series dataset with 27 billion observations across nine domains, and has outshone other models in forecasting tests. Fields experimented in include forecasting in energy, transport, climate, sales, economics, and healthcare. [Unified Training of Universal Time Series Forecasting Transformers] Explore more of our coverage of: Salesforce AI, Moirai Model, Time Series Analysis. Share this story by email

#2. OpenAI's Andrej Karpathy Steps Down to Pursue Personal Projects

First impacted: AI professionals, AI enthusiasts
Time to impact: Short

Andrej Karpathy, founding team member at OpenAI and AI badass, says he has left the company to focus on personal projects. He mentioned that his departure is not due to any specific event, and he commends the OpenAI team and its future roadmap, hinting that those familiar with his work might have an idea of his upcoming projects. [via @karpathy] Explore more of our coverage of: Andrej Karpathy, OpenAI, Personal Projects. Share this story by email

#3. Hugging Face Assembles Anti-Deepfake Tools

First impacted: Content Creators, Cybersecurity Professionals
Time to impact: Medium

Hugging Face has compiled and shared a collection of tools aimed at identifying and preventing the proliferation of AI-generated "deepfake" content. Among these are watermarking tools and techniques known as "Photoguard" for protecting images from generative model manipulation, LLM Watermarking, and image "poisoning," which is a strategy intended to disrupt the development of facial recognition models. [via @mmitchell_ai] Explore more of our coverage of: Deepfake Prevention, AI Watermarking, Facial Recognition Disruption. Share this story by email

#4. Fractal Structure Found in Neural Network Training

First impacted: AI researchers, Neural network engineers
Time to impact: Medium

Jascha Sohl-Dickstein, a research scientist at Google, revealed his discovery of a fractal pattern in the hyperparameter landscapes of neural network training. (Parameters are derived from training data; the hyperparameters are externally imposed during configuration and optimization.) Sohl-Dickstein suggests that this fractal behavior can demonstrate the challenges of meta-learning, as minor adjustments in hyperparameters could lead to considerable changes in training dynamics. [Neural network training makes beautiful fractals] Explore more of our coverage of: Neural Network Training, Hyperparameter Optimization, Meta-Learning Challenges. Share this story by email

#5. StabilityAI Launches Text-to-Image Model: Stable Cascade

First impacted: Artificial Intelligence Developers, Graphic Designers
Time to impact: Short

StabilityAI has launched Stable Cascade, a new text-to-image model, which the company says significantly reduces costs by 16 times when compared to similar-sized Stable Diffusion models. The model was built on the Würstchen architecture, which was released in a research paper on Arxiv in September 2023. It reportedly excels in prompt alignment and aesthetic quality based on several tests against models such as Playground v2 and SDXL. [Introducing Stable Cascade — Stability AI] Explore more of our coverage of: StabilityAI, Text-to-Image Model, Cost Efficiency. Share this story by email

That’s it! Ok, keep the feedback coming and we’ll share more AI news next week. Have a good weekend!