These dogs sure are barking.
A machine learning expert successfully integrated ChatGPT into the famous robodogs made by Boston Dynamics, a high-tech product that both the NYPD and the FDNY have recently released.
Techie Santiago Valdarrama merged the program into the yellow and black four-legged bots. It was done as a way to improve the “automated missions” that the robots ran to assess factors in their near environment.
Users can now speak directly to the robodogs and they will reply “in plain English”. posted on Twitter along with a clip of the GPT hounds in action.
“Every mission uses configuration files that are miles long and hard to understand. Only technicians can handle them. At the end of each mission, the robots capture a ton of data. There is no simple way to query everything on demand,” Valdarrama wrote.
“That’s where ChatGPT comes in. We show you the configuration files and the results of the mission. Then we ask questions using that context. Put that together with a voice-enabled interface, and we have an amazing way of querying our data!
Now, the robots can answer questions “in real time” for both their past and future targets, according to Valdarrama, who had pre-programmed commands related to specific contexts.
Technology expert colleague Raul Junco described the integration as a “game changer”.
As for what this pairing is capable of moving forward, “your imagination is pretty much the limit,” boasts the AI expert. Its current uses are for detecting thermal anomalies, reading both analog and digital gauges, detecting acoustic anomalies as well as rust, as well as identifying changes in its environment.
In practical use during April, the FDNY used its non-ChatGPT robot dog to conduct search and rescue inside the collapsed parking garage in lower Manhattan.
However, one user raised concerns that the technological breakthrough could create a “wild west” scenario with the power it wields. Others also expressed fear about what this could mean moving forward.
“We are limiting ChatGPT to answer questions from a context we generated, so there is no way to get any other text. ChatGPT cannot control the robot at all. It’s just a conversation,” Valdarrama wrote.
“We think about ethics and safety. I promise you, the pesticides in your food are 100 times more dangerous than this.”