Super-sized neural networks with billions of parameters, capable of human-level conversation and complex task solving, have received much attention in recent years. This development has also reached robotics. Many robotics researchers are now working on robust fundamental models: large neural networks that can solve many different tasks and function on different robotic bodies.
In this project, Kevin Luck and his colleagues will explore new ways for such fundamental models to more reliably identify the robot's embodiment and current task, so that they can generate tailored robot control policies. They aim to achieve this by modularizing and internally structuring robust fundamental models. In doing so, they aim to not only make these models more explainable, more applicable to different robotic devices and more rapidly deployable, but also to potentially reduce the energy consumption for (re)training such large models as new data becomes available.
By the end of this research project, the computer scientists hope to have achieved initial results and launched a research program that can expand the application areas of large neural networks in robotics, address their current problematic energy inefficiency, and provide researchers and citizens with a better understanding of the internal workings of a robust fundamental model.
AI Ideas and Initiatives
Developments in artificial intelligence are being watched with great interest worldwide. For example, AI in healthcare can lead to faster and better diagnoses. But also, for example, improve the lives of people who have difficulty communicating with their loved ones due to paralysis. It is important for the Netherlands to take advantage of the opportunities AI has to offer. To boost the knowledge and innovation base in the field of AI in the Netherlands, and to strengthen the connection with European researchers, this call has been developed.These grants are used to investigate promising ideas and innovative or high-risk initiatives in the field of artificial intelligence (AI). These honored projects are carried out in collaboration with at least one foreign European collaborative partner organization.
About this call
This call was designed in 2024 in collaboration with the AiNed Foundation and attracted a total of 40 projects. Nearly €800,000 was distributed in this final round of evaluation. It explicitly involved curiosity-driven, adventurous research and the ability to quickly explore a promising idea in the field of AI. The proposed research is groundbreaking and it is not certain in advance whether the intended objective will be achieved. What matters is that every result, whether positive or negative, advances science.
Read more about the other grant winners on the NWO website.
