In a development some consider disturbing, a computer engineer merged one of Boston Dynamic’s Spot robot dogs with OpenAI’s ChatGPT, allowing the robot to verbally answer questions and receive verbal instructions.
In a Twitter thread, Santiago Valdarrama, who according to his LinkedIn profile has “designed and implemented software and machine learning systems for some of the largest companies in the world,” details his experiments with the robots.
“These robots run automated missions everyday,” Valdarrama wrote, adding that each mission requires “miles-long, hard-to-understand configuration files” that laymen wouldn’t understand.
Valdarrama wrote, “That’s where ChatGPT comes in. We show it the configuration files and the mission results. We then ask questions using that context. Put that together with a voice-enabled interface, and we have an awesome way to query our data!”
In the video, Valdarrama says, “Funny thing is, you ask Spot, what’s his name and the answer is, ‘I’m OpenAI.’”
You are now signed up for our newsletter
Check your email to complete sign up
“That is just messed up,” Valdarrama said.
With ChatGPT installed, the robot even displays gestures, shaking its head to say “no” or bowing to say, “yes.”
When a Twitter user asked if “there is any kind of person who thinks about ethics when you do projects like these?” Valdarrama responded, “We do think about ethics and safety. I promise you, pesticides on your food are about 100x more dangerous than this.”
Another Twitter user, Jana, commented, “I appreciate your research and work, but I just don’t think we should be connecting bots to GPT at this juncture…even in this limited context,” and when asked by Valdarrama why that shouldn’t happen, Jana simply responded, “Respectfully, with every piece of information that you feed into the LLM, I believe you are inching closer to this,” and posted a video entitled, “How to prevent AI from killing everyone.”
Valdarrama did say he implemented restrictions on the AI, writing in a subsequent tweet, “We are limiting ChatGPT to answer questions from a context we generate, so there’s no way to get any other text,” adding that, “ChatGPT can’t control the robot at all. It’s just conversational.”
- Meet the Overemployed Who Moonlight With ChatGPT On the Sly for a Second Paycheck
- Google Is Trying to Create a ‘Digital God,’ Elon Musk Says
- Layoffs Rolling as Lyft Dumps 26% of Employees In Struggle To Get Profitable
Integrating AI, like ChatGPT, with robots has been the long-term goal for tech giants, including Microsoft.
In a paper, published in February this year by Microsoft entitled “ChatGPT for Robotics: Design Principles and Model Abilities,” the company outlined design principles for merging ChatGPT with robots.
Microsoft recently invested billions into OpenAI, the company responsible for the development of ChatGPT.
The paper asks, “Have you ever wanted to tell a robot what to do using your own words, like you would to a human?” and links to a technical paper that describes a “series of design principles that can be used to guide language models towards solving robotics tasks.”
“We believe that our work is just the start of a shift in how we develop robotics systems,” the company wrote, adding that, “and we hope to inspire other researchers to jump into this exciting field.”