Spot, Boston Dynamics’ famous robotic dog, has just received an update that allows it to communicate with humans via ChatGPT, an artificial intelligence system capable of generating natural language. This innovation could make Spot easier to control and query, especially in the area of industrial inspection. The team that achieved this feat is that of Levatas, a company specializing in industrial artificial intelligence that works with Boston Dynamics.
In September 2019, Boston Dynamics began shipping its quadruple Spot robot, capable of climbing stairs or traversing rough terrain. The robot, oddly reminiscent of those from the War of the Worlds series on Canal+, is used in particular in industry to carry out inspection missions, such as detecting and reading analog gauges, identifying thermal anomalies, checking for the presence of unauthorized persons, detect signs of corrosion or damage to equipment or even scan workplace safety conditions.
These transfers generate a large amount of data that is stored in a database. Until now, only people with technical skills and specific training could access and use this information.
But thanks to ChatGPT, an artificial intelligence system developed by OpenAI, Spot can now interact with humans in natural language.
We have integrated ChatGPT into our robots.
We had a lot of fun building!
Read on for the details: pic.twitter.com/DRC2AOF0eU
—Santiago (@svpino) April 25, 2023
A smarter, more talkative robot dog
Boston Dynamics’ robotic dogs can’t bark, but they can speak in complete sentences thanks to generative AI.
Levatas, an AI software company, has partnered with Boston Dynamics to integrate OpenAI’s ChatGPT with Spot. By combining ChatGPT with Google’s text-to-speech capability, Spot can answer users’ questions with their voice. It can also provide you with information about its condition, its past or future emissions, or even the results of its analyses.
An advantage for customers and users
The integration of ChatGPT Spot has several advantages for customers and users of the robot dog.
First, it simplifies the robot control interface, which no longer requires complex control panels or detailed reports. Just talk to Spot to give him directions or ask for information.
Second, it provides access to information that was previously difficult or impossible to obtain. For example, we can compare data from different missions to gain new insights. You can also ask Spot to compare his observations to those of other robots or to established standards.
Finally, it makes Spot more user-friendly and interactive. So you can talk to the robot like you would talk to a colleague or a pet. You can even ask him to leave if you’re too offended by his presence.
A step towards better cooperation between humans and robots
According to Chris Nielsen, the founder and CEO of Levatas, the goal is to create better collaboration between humans and robots.
We want robots to be seen as partners, not tools. We want them to be able to understand the context they’re working in and adapt to users’ needs, he explains. For this purpose, Levatas has created an industrial artificial intelligence laboratory that tests and develops the cognitive abilities of spot robots.
In an interview, Nielsen claims that these capabilities have allowed large industrial sites to use these robots in automated reconnaissance missions, rather than installing thousands of expensive sensors everywhere. Each of these spot missions have lengthy to-do lists detailing what to check on each ride. These tasks result in large sets of observational data being entered into a database. *Only technicians can manage them. At the end of each mission, the robots capture a lot of deals. There’s no easy way to query all of this on demand, he says.
This is where ChatGPT comes in. According to Nielsen, his team created a much simpler way for Spot and its human controllers to communicate using natural language. People can speak Spot using regular voice commands, but perhaps its most useful feature is Spot’s new ability to instantly analyze tons of information and use that information to answer previously unanswered questions. For example, we can cross-check information from different missions [de reconnaissance] for the first time without pre-programming this capability,” says Nielsen. We give ChatGPT the raw data and instructions for interpretation and respond to customer requests.
Routing, querying and obtaining previously hidden information about the status of an industrial site is a major problem for these companies. ChatGPT integration now means anyone can talk Spot, not just people with technical expertise and robotics training. Customers can query anything they want and get the information they need, all in plain language, explains Nielsen.
The addition of a ChatGPT brain to these machines means they can now offer an extra layer of security too. Rather than relying on a single remote operator, any human in the facility will be able to safely and quickly interrupt or modify the robot’s autonomous mission without the need for technical training. It will be enough to talk about it.
According to Nielsen, his team is currently investigating how to use GPT-4 to solve complex tasks, not just follow orders, he says. The big leap will happen when we get big language models with tools like GPT-4 that work locally on devices instead of online,” he says. This feature of GPT-4 which is online limits where you can use these bots as many of these large installations do not have site wide internet connectivity.
Once they get these complex new brains to work independently inside the robots, they can leave some of them in each facility, give them a primary mission, and ask them to figure out how to better carry them out at home without tech tools can training. These spot units will be able to scout facilities, discover a variety of equipment and conditions, and then report back with safety, protection and maintenance recommendations for human teams, he says.
A potential standard for the future of robotics
The addition of ChatGPT Spot is just one example of the advancements in the field of robotics. Thanks to artificial intelligence, robots are becoming increasingly intelligent, autonomous and able to communicate with humans.
This opens up immense prospects for the future of robotics, especially in areas such as health, education, agriculture, security and even entertainment. So robots could help humans to accomplish difficult, dangerous, or repetitive tasks, but they could also learn in other areas.
However, these advances also raise other types of questions. For example, how can we ensure the safety and respect of humans towards robots? How to avoid the risks of hijacking or hacking robots?
These questions require joint reflection and appropriate regulation. We must also ensure that robots do not replace humans, but complement and enrich them. As Santiago Valdarrama, the artificial intelligence expert involved in the project, says: We don’t want robots to be smarter than us. We want them to be more useful than us.
News of Spot’s new abilities comes just weeks after New York Mayor Eric Adams announced that the city’s police department will once again use Boston Dynamic’s robotic dogs in life-threatening situations such as bombings and hostage situations. They are even used to inspect hazardous waste landfills.
Adams said dogs can save lives in situations where you can’t get the police in because it’s dangerous, although there are many privacy issues.
Though Levatas doesn’t work with the NYPD, Nielsen said he could see a valuable use case with ChatGPT, where anyone could communicate with the bot to ask for directions or find out where the facility is located.
Source: videos in text
And you ?
how do you read it