Robot-making company Boston Dynamics has successfully incorporated ChatGPT technology into one of their robots, enabling real-time conversation. The project, conducted with ‘Spot’ the robo-dog, is based on the recent breakthrough in foundation models (FMs) emerging in the field of chatbots.
Seeking a real-world application for testing Spot, Boston Dynamics envisaged using the robot as a tour guide. Not only would this allow Spot to interact with its environment, but it would also necessitate conversation with the audience alongside decision-making about subsequent actions. The team designed Spot to operate like an improv actor – given a general framework, it fills in specific interactions in real-time using FM technology.
Technological Fusion: Robot and Speech Recognition
To equip Spot to converse and answer questions, Boston Dynamics installed a speaker and microphone and connected it with ChatGPT from OpenAI, as well as Whisper’s speech recognition API. In order to allow Spot to appear as though it was in conversation with the audience, the company developed a program to translate Spot’s speech into movements of Spot’s gripping tool, akin to the mouth of a puppet. This was further enhanced by incorporating light-hearted elements like a costume and googly eyes.
Surprising Results and Future Implications
While the experiment was largely successful, Boston Dynamics faced some unexpected results. When asked to identify its ‘parents,’ Spot was able to identify previous ‘Spot’ models as its ‘ancestors.’ It also exhibited a degree of improvization by suggesting to go to the IT help desk to find out about Marc Raibert, founder of Boston Dynamics.
Despite its minor flaws, such as intermittent slow response times and susceptibility to internet and system bugs, the BD team expressed excitement about Spot’s potential. They believe that a future where robots can understand and respond in real-time to human speech is shortly arriving.
Such a development can permit robots to work more efficiently with humans across various fields, from tools to guides, companions, and entertainers. The success of Spot’s experiment serves as a stepping stone towards achieving this vision.
Ongoing Synergy Between AI and Robotics
AI and robotics are two fields that have very close link. As generative AI has become mainstream, the development of robotics has also seen a spike in progression. Tech companies are exploring ways to integrate AI into robots in various ways.
Amazon is paving the way for technological efficiency in ecommerce operations by integrating artificial intelligence (AI) and robotics capabilities into its warehouse operations. The new AI and robotics-enabled technologies are aimed at reducing delivery times and streamlining the process of identifying inventory.
Google DeepMind and 33 other research institutions have recently unveiled an ambitious project, known as Open-X Embodiment. The goal of this initiative is to solve the conundrum of having to train machine learning models for each robot, task, and environment separately ― a process that often requires a substantial amount of time and effort.
Tesla’s Optimus robot, initially announced in 2021, has demonstrated significant advancements in its capabilities. A recent video shared by the company on its official Twitter account, @Tesla_Optimus, showcases the robot autonomously sorting objects using its end-to-end trained neural network.
Agility Robotics, known for its innovative bipedal robot, Digit, has announced the opening of RoboFab, a state-of-the-art robot manufacturing facility located in Salem, Oregon. This facility boasts the capability to produce over 10,000 robots annually. Meta is also exploring AI in robotics a has this year used AI to give robots vision similar to a visual cortex.
Last Updated on November 8, 2024 10:24 am CET
