3 Sources
3 Sources
[1]
General Intuition lands $134M seed to teach agents spatial reasoning using video game clips | TechCrunch
Medal, a platform for uploading and sharing video game clips, has spun out a new frontier AI research lab that's using its trove of gaming videos to train and build foundation models and AI agents that can understand how objects and entities move through space and time - a concept known as spatial-temporal reasoning. Called General Intuition, the startup is betting that Medal's dataset - which consists of 2 billion videos per year from 10 million monthly active users across tens of thousands of games - surpasses alternatives like Twitch or YouTube for training agents. "When you play video games, you essentially transfer your perception, usually through a first-person view of the camera, to different environments," Pim de Witte, CEO of Medal and General Intuition, told TechCrunch. He noted that gamers who upload clips tend to post very negative or positive examples, which serve as really useful edge cases for training. "You get this selection bias towards precisely the kind of data you actually want to use for training work." This data moat is what reportedly attracted the attention of OpenAI, which late last year attempted to acquire Medal for $500 million, per The Information. (Neither OpenAI nor General Intuition would comment on the report.) It's also what has led to General Intuition's raising a whopping $133.7 million in seed funding, led by Khosla Ventures and General Catalyst with participation from Raine. The startup intends to use the funds to grow its team of researchers and engineers focused on training a general agent that can interact with the world around it, aiming for initial applications in gaming, and search and rescue drones. De Witte says the founding team has already made strides: General Intuition's model can understand environments it wasn't trained on and correctly predict actions within them. It's able to do this purely through visual input; agents only see what a human player would see, and they move through space by following controller inputs. This approach, the company says, can transfer naturally to physical systems like robotic arms, drones, and autonomous vehicles, which are often manipulated by humans using video game controllers. General Intuition's next milestone is two-fold: generating new simulated worlds for training other agents, and autonomously navigating entirely unfamiliar physical environments. That technical approach is shaping how the company plans to commercialize its technology, and sets it apart from competitors building world models. While General Intuition is also building world models on which to train its agents, such models aren't the product. Unlike other world model makers like DeepMind and World Labs, which are selling their world models Genie and Marble for training agents and content creation, General Intuition is focusing on other use cases to avoid copyright issues. "Our goal is not to produce models that compete with game developers," de Witte said. Instead, the startup's gaming applications center around creating bots and non-player characters that can surpass traditional "deterministic bots," or preprogrammed characters that produce the same output every time. "[The bots] can scale to any level of difficulty," Moritz Baier-Lentz, a founding member of General Intuition and partner at Lightspeed Ventures, told TechCrunch. "It's not compelling to create a god bot that beats everyone, but if you can scale gradually and fill in liquidity for any player situation so that their win rate is always around 50%, that will maximize their engagement and retention." De Witte also has a background in humanitarian work, which informs the startup's focus on powering search and rescue drones, that sometimes have to navigate unfamiliar environments and extract information without GPS. Ultimately, de Witte and Baier-Lentz see General Intuition's core functionality - spatial-temporal reasoning -- as a crucial piece in the race toward artificial general intelligence (AGI). While major AI labs focus on building ever more powerful large language models, General Intuition believes true AGI requires something LLMs fundamentally lack. "As humans, we create text to describe what's going on in our world, but in doing so, you lose a lot of information," de Witte said. "You lose general intuition around spatial-temporal reasoning."
[2]
Why world models are the next big thing in AI
He had read the Google DeepMind research paper showing that gaming data can be used to teach AI how to navigate a 3D environment. However, the interest from AI labs made him realize that his data from Medal, which receives roughly 2 billion video uploads per year from tens of thousands of video games, could be used to develop a unique foundational model for extending AI to the real world. Today, Pim de Witte announced that Medal is spinning out a new AI lab called General Intuition that has raised a $133.7 million seed round. The money for the round is primarily from Vinod Khosla, founder of Khosla Ventures and one of the first investors in OpenAI. Other investors include General Catalyst and the Raine Group. Moritz Baier-Lentz, who oversees Lightspeed's gaming investments, is also joining the startup part-time as a founding team member.
[3]
Spatial-temporal reasoning startup General Intuition closes $133.7M investment - SiliconANGLE
Spatial-temporal reasoning startup General Intuition closes $133.7M investment General Intuition PBC, a startup developing artificial intelligence models that can navigate three-dimensional environments, has raised $133.7 million in funding. TechCrunch reported today that Khosla Ventures and General Catalyst led the seed round. Raine Ventures participated as well. General Intuition was recently spun off from Medal B.V, a venture-backed startup with a popular video sharing app. The software enables users to record and share video game footage. General Intuition will use that footage to train so-called spatial-temporal reasoning models capable of interacting with 3D environments. The company initially plans to focus on two use cases. The first is building AI features for video games, while the other is powering autonomous search and rescue drones. According to TechCrunch, General Intuition has already developed a model that can understand 3D environments on which it wasn't specifically trained. The company's development roadmap also places an emphasis on building world models. Those are neural networks optimized to generate virtual environments. According to General Intuition, its engineers will use those virtual environments to train spatial-temporal reasoning models. AI-generated virtual environments currently have certain technical limitations. General Intuition may have to address those limitations as part of its commercialization efforts. One challenge is that today's world models can simulate only a limited range of interactions in the environments they generate. Google LLC's cutting-edge Genie 3 world model, for example, struggles to simulate interactions that involve multiple AI agents. Simulation consistency is another challenge. During training, an AI model may travel through a certain section of a virtual environment and return to that section a few minutes later. Such interactions can lead to rendering inconsistencies in the simulation. Spatial-temporal reasoning systems are particularly difficult to implement when they require the ability to follow user instructions. A logistics company, for example, might wish to equip its warehouse robots with the ability to retrieve items requested by workers. Implementing such capabilities requires specialized AI training data and neural network designs. General Intuition will use the proceeds from its seed round to hire more researchers and engineers. According to TechCrunch, the new hires will speed up the company's AI development efforts. A job posting indicates that General Intuition is also building new training data processing software. According to the listing, the software will ingest thousands of hours of gameplay footage per day from Medal, the company from which General Intuition spun off. One of the project's goals is to filter erroneous and duplicate training data. OpenAI reportedly offered to acquire Medal for $500 million last year. Sources told The Information that the ChatGPT developer sought to obtain Medal's video training datasets, which could potentially help it enhance Sora. In a February 2024 blog post, OpenAI researchers suggested Sora can be used as a world model like those General Intuition is using to train its spatial-temporal algorithms.
Share
Share
Copy Link
General Intuition, a spin-off from Medal, raises $133.7 million to develop AI models with spatial-temporal reasoning using video game footage. The startup aims to create advanced AI agents for gaming and search and rescue applications.
General Intuition, a new frontier AI research lab, has spun out from Medal, a popular platform for sharing video game clips. The startup has secured a substantial $133.7 million in seed funding, led by Khosla Ventures and General Catalyst, with participation from Raine
1
3
.Source: TechCrunch
The company's unique approach involves using Medal's vast dataset of video game clips to train and build foundation models and AI agents capable of spatial-temporal reasoning. This dataset, comprising 2 billion videos per year from 10 million monthly active users across tens of thousands of games, is believed to surpass alternatives like Twitch or YouTube for training purposes
1
.Source: SiliconANGLE
General Intuition has already made significant progress in developing AI models that can understand and predict actions in unfamiliar environments purely through visual input. The company's next milestones include generating new simulated worlds for training other agents and enabling autonomous navigation in unfamiliar physical environments
1
.The startup is focusing on two primary applications:
Creating advanced bots and non-player characters for video games that can adapt to different skill levels and enhance player engagement
1
.Powering search and rescue drones capable of navigating unfamiliar environments without GPS
1
.Source: The Verge
Related Stories
As General Intuition advances its technology, it may face challenges in simulating complex interactions and maintaining consistency in virtual environments. The company is also developing new training data processing software to filter erroneous and duplicate data from the vast amount of gameplay footage it receives
3
.The startup's unique approach has attracted significant attention from major AI labs. OpenAI reportedly attempted to acquire Medal for $500 million last year, highlighting the value of the company's dataset
2
3
.General Intuition's focus on spatial-temporal reasoning is seen as a crucial step towards achieving artificial general intelligence (AGI), potentially addressing limitations in current large language models
1
.Summarized by
Navi
[1]
[2]
1
Technology
2
Business and Economy
3
Business and Economy