ºÃÉ«ÏÈÉúTV

Tal Daniel (he/him)

Postdoctoral Fellow, Robotics Institute

Hometown: Giv-atayim, Israel

1. Briefly describe the work you're doing as a postdoc at ºÃÉ«ÏÈÉúTV.

   I develop models and algorithms that help robots understand the world, drawing inspiration from how people perceive their surroundings. Rather than treating an image as a collection of pixels, our models organize visual information—from images, videos, and 3D data—into representations of individual objects or their parts. We call these representations “particles.” Each particle captures information about what something is and where it is, helping us describe complex scenes in terms of their individual components.

   
   

   A major focus of our work is building “world models”: models that predict how a scene might change in response to a robot’s actions or other inputs, such as text instructions. The goal is to help robots anticipate the consequences of their actions before taking them, much as people imagine what might happen before deciding what to do. Alongside this research, I mentor master’s students on projects that help advance these ideas.

2. How did you develop an interest in this area?

   My interest grew out of a challenge I encountered early in my PhD. We were training AI agents to play video games using only visual inputs, but they struggled to achieve the results we had hoped for. This led us to question whether the issue was just the learning algorithms—or also the way we represented what the agents could see.

   
   

   When many things are happening on screen at once, a single representation of the entire scene can make it difficult to distinguish individual objects and track how they interact. We began exploring more structured ways to represent scenes, breaking them down into separate entities. That experience sparked my interest in how representations can help AI systems understand dynamic environments and act within them. 

3. What are your academic and/or professional goals?

   I hope to continue doing meaningful research that expands what AI systems can understand and accomplish, ultimately helping us tackle more complex problems. My broader goal is to develop safe, intelligent AI that improves people’s lives.

   
   

   There is still so much to explore, and I believe progress requires both developing new capabilities and thinking carefully about how these systems fit into our daily lives. I want that balance to guide my work.

4. Tell us about a significant moment in your journey at ºÃÉ«ÏÈÉúTV so far.

   While I’m proud of the first research project I led at ºÃÉ«ÏÈÉúTV, an especially meaningful moment came when the first project on which I mentored a master’s student resulted in a paper accepted at a leading AI conference. It was rewarding not only to see the research succeed, but also to contribute to a student’s growth as a researcher. That experience made the mentoring side of my role particularly meaningful to me.

5. How do you spend your time beyond academic work?

   My favorite way to start the day is with an early-morning workout, especially a run. Many solutions to problems I’ve been wrestling with in my research have come to me during a run or afterward. Away from work, I also enjoy reading science fiction and thrillers, going to the movies, and spending time with friends.