✨ Issue #37

Clark Valberg
February 20, 2024 • Estimated Reading Time: 11 minutes

Your weekly briefing on all things AI

This week we're spotlighting a breakthrough that's more than just a technological marvel—it's a foundational step toward Artificial General Intelligence that paves the way for AI to learn about the world more like we do. This week we also see AI code-writing tools coming from Apple, Google, & others. Let’s dive into the latest!

📰 AI in the News

OpenAI’s Revolutionary Leap in Text-to-Video Generation. Exhibit A:

Prompt: A stylish woman walks down a Tokyo street filled with warm glowing neon and animated city signage. She wears a black leather jacket, a long red dress, and black boots, and carries a black purse. She wears sunglasses and red lipstick. She walks confidently and casually. The street is damp and reflective, creating a mirror effect of the colorful lights. Many pedestrians walk about.

Click here for full video & more examples

OpenAI's Sora, a groundbreaking text-to-video model, has been unveiled, representing a major advancement in AI-driven content creation. This technology promises to revolutionize industries by streamlining video production, making it more efficient and less resource-intensive.

Sora is currently in a feedback and testing phase with selected experts and creatives, focusing on creating complex scenes and addressing deepfake concerns. While it showcases OpenAI's progress in AI with its ability to generate detailed videos, Sora still faces challenges with physical realism and causal relationships.

Key Features and Capabilities

Realistic Scenes: Sora can generate complex scenes with multiple characters, specific types of motion, and accurate details of the subject and background
Physical World Simulation: The model understands how things exist in the physical world, enabling it to simulate people, animals, and environments with believable real-world interactions
Long-Range Coherence: Sora maintains consistency and object permanence, allowing characters and objects to persist even when they are occluded or leave the frame
Prompt Adherence: The model generates videos that adhere to the user's prompt, creating compelling characters that express vibrant emotions
Video Extension: Sora can take an existing still image and animate it, or extend an existing video by filling in missing frames

(Full Story)

Meta Open Sources Model for AI Learning via Video

Meta has released V-JEPA, an AI model advancing machine intelligence, especially in video understanding. V-JEPA excels in predicting missing parts of videos, focusing on abstract representations rather than pixel details. This self-supervised model, trained on unlabeled data, is efficient in learning and adapting to various tasks. Meta is releasing this model under a Creative Commons NonCommercial license for researchers to further explore.

V-JEPA's approach represents a significant step towards achieving Artificial General Intelligence (AGI), as it mimics human learning by observing and understanding the world, a crucial capability for AGI development. This model could pave the way for AI systems that process and interpret video content more like humans, enhancing their ability to learn from and interact with their environment. (Video Explanation)

📈 Venture Deals

Sierra raised $110M to bring AI agents to business (link)
Zededa, an enterprise computing company, raised $72M ($400M valuation) to power AI tools (link)
Unlearn raised $50M to optimize clinical research with AI-powered digital twin technology (link)
Rasa raised $30M to help enterprises build better chatbots (link)
Scribe raised $25M Series B to expand access to workplace knowledge, prioritizing mission over money (link)
Mechanical Orchard raised $24M to help enterprises get on the cloud (link)
Magic.dev is a code generating platform that raised $23M to compete with Github’s CoPilot (link)
Ezra raised $21M to develop AI-powered technology for full-body MRI scanning (link)
Daedalus raised $21M to build precision-manufacturing factories powered by AI (link)
Contents raised $18M to enable businesses to create high-quality multilingual content using AI (link)
Podcastle raised $13.5M to scale its AI content creation platform (link)
Marqo raised $12.5M for technology that helps companies search the vast amounts of unstructured data in their systems (link)
Attentive raised $7M to boost automation in landscaping, construction services (link)
MediConCen raised $6.85M to automate insurance claims using AI (link)
Tortus raised $4.2M to speed up medical admin with AI assistants (link)
Upliance raised $4M ($17M valuation) to create home appliances that integrate machine learning and AI (link)

💡 Industry Impacts

🏠Real Estate

Matterport Launches ‘Property Intelligence’: This technology provides automated property analysis, measurements, layouts, and reporting, revolutionizing property analytics across various industries. It eliminates the need for multiple site visits, automates manual workflows, and enables better decision-making based on rich property data. The technology is currently being piloted with select customers and is expected to reach general availability soon. (link)

⚖️ Legal

No ‘GPT’ Trademark: The U.S. Patent and Trademark Office has denied OpenAI's application to trademark "GPT," ruling it as "merely descriptive." This decision impacts OpenAI's branding but doesn't prevent others from using similar terms for AI models. The term GPT, short for "generative pre-trained transformer," was deemed too generic for trademark protection, with the USPTO noting its widespread use by various companies. (link)

⚡Energy/Utilities

AI Improved Critical Facility Management: Neara, an Australian startup, uses AI to strengthen utility networks against extreme weather. Its technology simulates weather impacts on utilities, enhancing restoration and safety. This innovation in digital modeling and analysis illustrates AI's significant role in improving critical infrastructure management. (link)

🛍️ Politics

Tech Giants Agree on Something: Major tech companies, including Microsoft, Meta, Google, Amazon, Adobe, and IBM, along with 13 others, have signed a voluntary pledge at the Munich Security Conference to combat election-related deepfakes. Although the measures are voluntary and lack legal force, the agreement highlights a growing effort to address the challenges posed by AI-generated content in political contexts. (link)

🛠️ Tool of the Week

Tome.app is an AI slide deck generator that just recently added some data visualization capabilities. They’ve got an intuitive app that lets users upload documents, media, and now CSVs to generate beautiful slides in less than a minute. They also integrate with Figma, Airtable, and more.

With teams becoming more and more visually focused, slides have practical applications everywhere. Using tools like Tome for pitch decks, sales presentations, media kits, onboarding docs, helps leaders get rid of the “blank page syndrome” and spend more time on the core message.

More Tools

Pika.art is a free text to video generator that is worth playing with, if you haven’t already.
AutoCut is a powerful set of tools for automated video editing.
Favikon is revolutionizing influencer marketing with its AI-driven platform that merges discovery & management on one platform.