When Robots Met Language (Models): An Exploration of Science and Strategy in the vision-language-action Models Space

We study the emerging technology of vision-language-action models (VLAMs), a potential breakthrough for the robotics industry. VLAMs connect artificial intelligence’s large language models with robots: they are a software innovation that increases flexibility and expand capabilities of robotic hardware. As the technology is yet in its infancy, we focus our analysis on three directions. First, we study the technological trajectory of VLAMs in scientific publications using scientometric techniques. Second, in order to assess market transformations associated with VLAMs and potential migration in the locus of value generation in robotics, we collect information and offer insights on the actors developing the technology ‘in the wild’. Third, we discuss the strategies put in place by a leading actor in the field to gatekeep the competitive landscape. We then generalise and rationalise our findings through established theoretical frameworks. The paper is the first to single out the forces at work in the development of an innovation poised to rejuvenate the robotics industry, one of the backbones of contemporary economies.