Battle of the AI Video Models: Luma AI vs Runway; Taxonomy of Prompting; AI Robot Drivers … and more

Welcome to this week's episode of AI Weekly Breakthroughs, hosted by Jan Štihec, Shelf's head of AI. Join us as we cover the latest innovations and updates in AI tech and research.
In this Episode:
Battle of AI Video Models:
Luma AI: Generates 120 frames in 120 seconds using text and image inputs. It’s free and impressive, though early reviews are mixed.
Runway Gen-3 Alpha: Trained for large-scale multimodal tasks, offering robust tools for text-to-video, image-to-video, and text-to-image generation. Featured videos include diverse scenes like a woman in traffic in a Japanese city and a balloon on a deserted street in Europe.
Advances in Image Generation:
Stable Diffusion 3 Medium by Stability AI: Their most advanced model to date can run on both consumer and enterprise GPUs. Available via Stability.AI’s API, it offers high-quality image generation.
Prompting Research:
New Taxonomy Presented: Collaboration between Stanford, OpenAI, and Microsoft introduces a structured view of prompting, including a taxonomy of prompting techniques.
AI Robot Drivers:
Autonomous Humanoid Robots: Groundbreaking study showcases robots driving vehicles by learning and recognizing objects, ensuring safety and operational efficiency.
Other Announcements:
NVIDIA: Announces an open synthetic data pipeline for training large language models.
Mistral AI: Secures $640M in Series B funding, achieving a $6B valuation.
OpenAI: Welcomes former NSA head to its board and safety committee.
Databricks: Expands Mosaic AI to boost enterprise-level LLM-powered application development.
Catch all these insights and more in this episode of AI Weekly Breakthroughs.
Subscribe, like, and comment to stay on top of the AI frontier!