Software Development
AI Researchers, Startups And Labs Raise Funds To Build AI World Models – But Why?
By TechDogs Bureau

Updated on Mon, Dec 16, 2024
Share


It was a groundbreaking moment in generative AI’s short but eventful journey. Now we have various players such as Amazon, Anthropic AI, Microsoft and others building AI chatbots and virtual assistants. Yet, according to AI leaders and researchers, a new wave of AI applications is upon us – one that is witnessing major funding.
World models, also known as world simulators, are emerging as a groundbreaking advancement in artificial intelligence (AI).
Recently, World Labs, co-founded by AI researcher Fei-Fei Li, secured $230 million to develop “large world models (LWMs)” with a focus on "spatial intelligence." This funding marks a significant milestone in the domain of AI models, as it promises to enhance AI system’s ability to understand and interact with the physical world.
So, what are AI world models and are they the next big thing in AI? Let’s explore!
What Are AI World Models?
World models are AI systems inspired by human cognitive processes, replicating the mental frameworks we use to interpret and interact with our surroundings. Human brains create abstract representations and use the five senses to build a concrete understanding of the world, leveraging what can be called a “model” long before AI adopted the phrase. These predictive models help our brains perceive the world around us and respond accordingly.
Similarly, by processing sensory inputs into abstract representations, world models enable AI to predict and reason about the consequences of actions. For instance, just as a human baseball batter’s brain anticipates the trajectory of a fastball.
As AI researchers David Ha and Jürgen Schmidhuber out it, "Their muscles reflexively swing the bat at the right time and location in line with their internal models' predictions."
World models can do the same by going beyond traditional AI models by creating internal representations of how the world works. This can allow machines to plan, predict and execute tasks with greater precision.
Are AI World Models The Next Big Thing?
Take OpenAI’s Sora, considered an early world model, showing a realistic Minecraft world here:
It can simulate actions like painting brush strokes on a canvas or generating video game environments. While other text-to-video generators will show anomalies like limbs twisting and merging or missing fingers.
This is because AI models are trained on years of video data that might accurately predict how an arm moves. However, it doesn’t actually have any idea why it moves that way!
An AI world model with even a fundamental understanding of why a human arm moves like it does will excel at rendering its video. However, to give AI models this insight, world models must be trained on a range of data spanning images, audios, videos, text and more.
AI world models can create internal representations of how the world works, reason about the consequences of actions and predict what might happen in physical space. They might just be the key to unlocking broader reasoning capabilities in AI systems!
Here’s what AI experts have to say, signaling AI world models might be the new shiny thing in AI.
What’s The Vision For AI World Models?
Fei-Fei Li, the co-founder of World Labs and “godmother” of AI, announced that she had raised $230 million from investors such as Andreessen Horowitz, Radical Ventures and New Enterprise Associates to develop large world models or LWMs.
The San Francisco-based AI startup will train the "large world models" on a combination of synthetic and real-world data. While these world models would use the same transformer-based architecture as OpenAI's ChatGPT, Li said it would not be the "be-all and end-all" of their models, suggesting World Labs might incorporate other elements into its LWMs.
Although World Labs declined to share its valuation, this move puts it in the driver’s seat in the spatial computing arena. With World Labs aiming to create AI systems capable of perceiving, generating and interacting with 3D environments, a successful world model will be in high demand for applications like augmented reality (AR), virtual reality (VR) and robotics.
“We envision AI systems that understand the physical world as intuitively as humans do,” said Justin Johnson, a co-founder of World Labs and a computer vision researcher.
Current AI-generated videos, for instance, suffer from inconsistencies and unnatural portrayals. This is where world models can enhance realism by understanding physical laws and object interactions in the real world. Alex Mashrabov, CEO of personalized AI video platform Higgsfield believes, "With a strong world model, instead of a creator defining how each object is expected to move — which is tedious, cumbersome, and a poor use of time — the model will understand this.”
Further, equipping robots with world models would allow them to navigate and interact with their environment more effectively, enabling advanced automation. According to Meta’s chief AI scientist Yann LeCun, world models could someday enable machines to achieve complex objectives through reasoning, such as cleaning a room or optimizing supply chains. "We need machines that understand the world; machines that can remember things, have intuition, have common sense," LeCun stated.
By building AI models with spatial intelligence, World Lab and others aim to move beyond 2D data interpretation, paving the way for richer and more immersive AI interactions. Fei-Fei Li sums it up saying, "Our goal is to create models that not only see and hear but also think, reason and interact as humans would in a three-dimensional world."
Conclusion
World Labs, Meta, Higgsfield AI and many other pioneers are pivoting to build AI world models that can generate, understand and interact with dynamic 3D environments. This has a major impact on everything from gaming and robotics to virtual meetings and training simulations.
As AI technologies mature, they could revolutionize various industries and unlock new possibilities for AI applications in our real world. With Fei-Fei Li’s initiative backed by substantial funding and visionary leadership, World Labs is at the forefront of this transformative journey to build LWMs.
Do you think AI world models will redefine the future of artificial intelligence? Will they be a key breakthrough in creating artificial general intelligence (AGI)?
Share your thoughts in the comments below!
First published on Mon, Dec 16, 2024
Enjoyed what you read? Great news – there’s a lot more to explore!
Dive into our content repository of the latest tech news, a diverse range of articles spanning introductory guides, product reviews, trends and more, along with engaging interviews, up-to-date AI blogs and hilarious tech memes!
Also explore our collection of branded insights via informative white papers, enlightening case studies, in-depth reports, educational videos and exciting events and webinars from leading global brands.
Head to the TechDogs homepage to Know Your World of technology today!
Disclaimer - Reference to any specific product, software or entity does not constitute an endorsement or recommendation by TechDogs nor should any data or content published be relied upon. The views expressed by TechDogs' members and guests are their own and their appearance on our site does not imply an endorsement of them or any entity they represent. Views and opinions expressed by TechDogs' Authors are those of the Authors and do not necessarily reflect the view of TechDogs or any of its officials. While we aim to provide valuable and helpful information, some content on TechDogs' site may not have been thoroughly reviewed for every detail or aspect. We encourage users to verify any information independently where necessary.
Tags:
Related News on Software Development
CDK Hosts Fourth Annual CDK CONNECT On Oct. 22
Thu, Oct 10, 2024
By Business Wire
Csharpcorner Announces Action AI 2024 Conference
Wed, Sep 25, 2024
By PR Newswire
GitLab Achieves FedRAMP In Process Designation
Wed, Jun 26, 2024
By GlobeNewswire
Join The Discussion