- The Rundown AI
- Posts
- Godmother of AI unveils World Labs
Sign Up | Advertise | Tools | AI University
Welcome, AI enthusiasts.
Fei-Fei Li, known as the “Godmother of AI,” just launched a startup to create AI that understands and generates 3D worlds.
With a fresh $230M in funding, could this spatial intelligence push AI beyond language models and revolutionize AR/VR, robotics, and more? Let’s get into it…
In today’s AI rundown:
World Labs aims to create 3D worlds with AI
AI creates open-world video games from text
Prompt tips and tricks for OpenAI’s new o1
OpenAI o1 is smarter than most humans
6 new AI tools & 4 new AI jobs
More AI & tech news
Read time: 4 minutes
LATEST DEVELOPMENTS
WORLD LABS
Image source: World Labs
The Rundown: Fei-Fei Li, renowned AI researcher known as the “Godmother of AI”, has launched World Labs, a spatial intelligence company developing AI models capable of understanding and generating 3D environments.
The details:
World Labs is pioneering "Large World Models" (LWMs) to perceive, generate, and interact with 3D worlds, incorporating physics and semantics beyond 2D imagery.
Co-founded by Li, Justin Johnson, Christoph Lassner, and Ben Mildenhall, the startup has secured over $230 million in funding from high-profile investors, including Andreessen Horowitz.
Initial applications focus on creating and editing virtual 3D spaces for creative professionals and general users.
The team aims to bridge the gap between AI's current text-based understanding and the complex spatial relationships of the real world.
Why it matters: Right now, AI relies heavily on text descriptions provided by humans. But figuring out spatial intelligence and moving AI beyond LLMs and language could greatly impact and accelerate the development of intelligent AR/VR, robotics, architecture, game design, and much more.
TOGETHER WITH DEFINED.AI
The Rundown: Defined.ai provides the most diverse and ethically-sourced marketplace for AI training data, allowing businesses to create smarter, more cost-effective AI solutions at scale.
Defined.ai delivers:
Customizable, bias-free datasets that adhere to strict privacy and ethical standards
Industry-specific data for sectors including insurance, banking, and retail
Instant access to pre-built datasets to accelerate your AI projects
Claim your 5 free hours of data today, exclusively for The Rundown readers!
TENCENT
Image source: GameGen-O
The Rundown: Tencent just unveiled GameGen-O, a groundbreaking AI model capable of generating open-world video game content from text prompts and allowing interactive gameplay simulation.
The details:
GameGen-O can create diverse game elements like characters, environments, actions, and events from simple text prompts.
It also allows for interactive control of the generated video game content it creates to build worlds that gamers can interact with.
The AI model was trained on OGameData, a new dataset of over 4,000 hours of high-quality video game footage and uses a two-stage training process: foundation pretraining and instruction tuning.
The training data segments were annotated by GPT-4o for validation purposes.
Why it matters: Following the impressive accomplishments of Google’s GameNGen, Tencent’s GameGen-O raises the stakes for AI-generated video games. This technology will dramatically reduce the time and cost of building open-world games, leading to higher quality games created by smaller teams of game developers.
AI TRAINING
The Rundown: OpenAI's latest release, o1, unlocks new reasoning capabilities, but there’s a catch: prompts should be fundamentally different than the way you prompt GPT-3 and GPT-4, due to the new Chain-of-Thought (CoT) architecture.
Here’s our top tips:
Log into ChatGPT (paid subscription required) and select o1-preview or o1-mini from the model dropdown.
Keep your prompts simple and direct, avoiding Chain-of-Thought instructions.
Use XML tags for clarity and test with challenging questions or philosophical queries.
Analyze its thinking process by clicking on the "thinking" indicator after each response to understand the model's reasoning path.
Pro tip: Experiment with different query types to fully grasp o1's advanced reasoning abilities and check out OpenAI’s new ‘Advice on prompting‘ guide.
PRESENTED BY ASSEMBLYAI
The Rundown: AssemblyAI delivers the most accurate speech-to-text model with advanced audio intelligence – allowing teams to build beyond transcription and transform words into actionable insights.
AssemblyAI offers:
Top performance rankings across major industry benchmarks
Fully-featured speech understanding models
Comprehensive developer docs and reliable SDKs
Sign up today and get $50 in free credits.
AI RESEARCH
Image source: TrackingAI.org
The Rundown: OpenAI’s latest model, “o1”, has reportedly scored an IQ of around 120 on the Norway Mensa IQ test, potentially marking the first time an AI model has surpassed the average human IQ.
The details:
o1 correctly answered 25 out of 35 questions on the Norway Mensa IQ test.
It demonstrated proficiency in solving complex visual and logical puzzles, both on publicly available tests and new, unpublished questions.
This performance suggests the improvement is not due to the questions being in the training data.
While OpenAI hasn't officially confirmed these results, ChatGPT Plus members can test o1-preview themselves using the model dropdown inside ChatGPT.
Why it matters: If confirmed, these results would indicate that OpenAI has achieved Stage 2 performance on their 5-tier roadmap towards AGI. This milestone suggests the model can stop, think, and reason about problems to arrive at correct answers more frequently — a crucial prerequisite for Stage 3 (AI agents).
NEW TOOLS & JOBS
🎬 Runway Gen-3 Video-to-Video - Video style transfer with the Gen-3 AI video generation model
🎵 Suno Cover - Reimagine the music you love with AI covers
👩🎤 HeyGen Avatar 3.0 - AI-generated avatars with emotions and tones that match your message
📊 Infography - Convert blogs into engaging infographics
🚀 FlutterFlow - Quickly generate code and styles for app development
📞 AIPhone - Translates and transcribes live calls seamlessly
👨💼 Siena - Product Manager
💻 Hippocratic AI - Senior Full Stack Engineer
📊 Horizon3 - Marketing Project Manager
🎥 Captions - Technical Program Manager
QUICK HITS
Runway released Gen-3 Alpha Video to Video, allowing users to transform input videos with AI-generated styles and prompts across all paid plans.
Meta (Facebook) admitted to scraping public data from all Australian adult users to train AI models, without offering an opt-out option available to EU users.
Google AI Studio launched a new model comparison feature, allowing users to easily compare outputs from different AI models and parameter settings.
Researchers developed “g1”, an AI system using Llama-3.1 on Groq hardware that implements reasoning chains to solve complex problems similar to OpenAI’s new o1 model.
A new AI chatbot using GPT-4 Turbo successfully reduced belief in conspiracy theories among users, with effects lasting for months after brief interactions.
Montana State University is developing AI methods using neural symbolic regression to help farmers maximize crop yields through precision agriculture.
Researchers are developing AI-piloted drone swarms to detect and extinguish wildfires, with teams of up to 30 autonomous aircraft working collectively to prevent fire spread.
THAT’S A WRAP
SPONSOR US
Get your product in front of over 650k+ AI enthusiasts
Our newsletter is read by thousands of tech executives, investors, engineers, managers, and business owners around the world. Get in touch today.
FEEDBACK
How would you rate today's newsletter?Vote below to help us improve the newsletter for you. |
If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.
Reply