Kiara Nirgin, co-founder of Chima, a research lab in applied artificial intelligence, reflects on new concepts such as "world models." These are AI systems capable of creating interactive and dynamic environments based on simple prompts. An example is Google DeepMind's Genie 3, which generates three-dimensional worlds from a single sentence, as well as Fei-Fei Li's work from World Labs, focused on creating large world models (Large World Models, LWM) that endow AI with "spatial intelligence" and allow it to interact with its environment.
This technology opens up endless possibilities for creating and exploring various worlds—from historical to fantastical. It is particularly relevant for Generation Z, who grew up in the smartphone era and are accustomed to rapid information consumption. How will these technologies affect their ability to maintain prolonged concentration?
What are world models?
Large language models, which underpin chatbots like ChatGPT, learn to generate text by processing vast amounts of data. In contrast, world models are trained on physical laws and properties of objects, enabling them to simulate real and imaginary worlds.
For instance, with Genie 3, one can create a scenario where "a helicopter pilot carefully maneuvers over cliffs," and then control the helicopter in this created world, similar to a video game. You can explore the Palace of Knossos or observe processes occurring during a hurricane in Florida. This means that AI not only generates content but also creates context, providing the user with a unique experience.
World Labs employs a different approach, focusing on spatial intelligence. Their LWMs are designed to be robust and manageable, allowing for the export of created worlds into games and virtual reality. This could lead to more people starting to create content using AI rather than just consuming it.
By the end of 2025, we witnessed significant announcements in the field of AI related to world modeling. Yann LeCun is launching the AMI Labs startup with a valuation of over $3.5 billion, while the company Runway introduced a new line of models for gaming and robotics. Every day, more startups emerge, such as Iconic, which recently raised $13 million to develop an AI game engine capable of creating "living adaptive worlds."
The pace of technological development is impressive. For example, Ryo Lu from Cursor, a startup valued at nearly $30 billion, shared how their CTO created a city simulator called IsoCity, including pedestrians and cars, in just three days. This was made possible by modern AI tools.
Creating world models will require users to be more engaged than current forms of entertainment. Users will not be able to just observe—they will have to actively explore, make decisions, and interact with their surroundings, which contrasts with the habitual content consumption on TikTok. This resembles deeper video games like Minecraft, which require curiosity and active participation.
The Attention Paradox
An interesting paradox is that modern young people, particularly those from Generation Z, are often perceived as unable to concentrate. Studies show that constant consumption of short videos can lead to reduced attention spans. However, according to a 2022 McKinsey report, the claim that their attention is limited to eight seconds is not true. Generation Z is simply more selective about where they direct their attention.
When they find interesting content, they are willing to invest time in it. The same McKinsey report emphasizes that this generation is contributing to a revival of print media, especially among young adult literature. They also actively watch documentaries and long streams.
The launch of Sora, a video creation app from OpenAI, confirms this complex dynamic of content consumption. The app quickly gained popularity, accumulating over a million downloads in five days, despite limited access. However, Sora is a platform similar to TikTok, where users view videos created by AI, indicating a preference for passive content consumption.
This raises interesting questions about Generation Z's interaction with world models. If developers can integrate complex video generation into familiar formats, how will the younger generation perceive the slow process of exploring new worlds? The rapid growth of Sora shows that AI tools that align with current habits are in demand, but it is important to understand whether world models will be useful for successful content interaction.
Three Possible Future Scenarios
I identify three scenarios that could influence the development of this field and the cognitive development of youth.
Cognitive Renaissance. Generation Z perceives world modeling as a means of creative expression and learning that was previously unavailable. They become not only consumers but also creators of content. The growing creator economy and the popularity of games like Minecraft confirm this trend.
The Great Rejection. Young people may feel that creating world models requires too much effort and time. The instant gratification offered by short content platforms may distract them, and new modeling could become a niche interest for enthusiasts.
Hybrid Future. The most likely scenario is that the younger generation will use both short content and world models. Rapid consumption will provide entertainment, while world models will serve for learning and professional growth. People will learn to switch between different modes of information perception.
An important aspect is how the benefits of world modeling will be distributed. Those who are willing to invest time and effort will be able to develop spatial thinking and creativity skills, but this could also lead to new digital inequality.
Questions of Freedom and Ownership
As artificial intelligence develops and increasingly complex worlds are created, new questions arise about control and ownership in the digital environment. Companies like Iconic are creating living worlds with AI agents that evolve over time. This raises questions about how interaction with these worlds may affect users' autonomy and their real relationships. When interesting experiences occur in digital spaces owned by companies, it raises concerns about freedom of action.
There are no clear answers to these questions, but they should be discussed, especially considering experts' forecasts that "world models will become dominant for AI architectures" in the next three to five years. We are witnessing not only a new technology but also a change in the way people interact with the digital environment and reality.
Ultimately, the impact of virtual worlds on youth will depend on how we choose to use them. They can be tools for learning and creativity that help develop new skills, or they can lead to the displacement of real experiences. It is important to remember that different modes of interaction lead to different cognitive outcomes and potential inequalities.
Source
