started · updated
AI advancements in robotics and language models
Engineers from the Massachusetts Institute of Technology (MIT) and New York University (NYU) have developed FLEX, a new AI-powered platform designed to give robots intuitive understanding of natural language commands. By utilizing the GPT-4o language model, the system allows robots to infer intent and adapt their movement to environmental context. During testing, robots equipped with this software could autonomously avoid objects similar to those specified by a user, such as avoiding a smartphone when instructed to avoid a laptop.
In a separate development, the SpeakLeash Foundation and the AGH University Academic Computer Center Cyfronet have released Bielik 3.1, an upgraded AI language model. The new version has been trained on 500 billion tokens, expanding its capabilities from 32 to 56 supported languages, including all official EU languages and several non-European languages like Arabic, Chinese, Japanese, and Korean.
Bielik 3.1 also introduces agentic competencies, enabling the model to plan steps and use external tools to complete complex tasks through function calling, structured output, and reasoning modes. In simulations for customer service sectors such as telecommunications and aviation, the model showed significant performance improvements over its predecessor.
Entities
AGH University · Bielik · Massachusetts Institute of Technology · New York University · SpeakLeash Foundation