Boston Dynamics reveals talking robot dog that can see, hear and respond to people

If the sight of Boston Dynamics’ robot dog wasn’t eerie enough, the four-legged droid with a mechanical gait now has an uncanny voice.

The artificial canine named Spot previously popped up at Heathrow Airport and even helped to scout an unsafe historical site in Suffolk. While it normally goes about its job in silence, pulling objects and opening doors with its mouth-like gripper, that could be about to change.

In a new video, Spot can be seen talking to its creators in a human-sounding voice. This modified version of the droid can switch between over half a dozen personas, each more eccentric than the last.

One minute it sounds like a butler with a posh English accent, then it transmutes into a teenager that uses filler words such as “like” in every sentence. In all, the robot has over half a dozen personalities, each with its own quirks and intonations.

To demonstrate its way with words, Boston Dynamics has turned this modified model into a tour guide at one of its facilities. The robotics company transformed Spot into a talking mutt by using a series of plug-in artificial intelligence models, including viral chatbot ChatGPT.

To help it describe its surroundings, the bot was fed a 3D map of the building it was operating in with brief descriptions and labels for important locations. It then used the most advanced version of ChatGPT to fill in the blanks during conversations.

Spot the robot dog can switch between several personalities, from an excitable teen to a sarcastic man named Josh (Boston Dynamics)
Spot the robot dog can switch between several personalities, from an excitable teen to a sarcastic man named Josh (Boston Dynamics)

The AI chatbot can be prompted to write poems, emulate the style of famous writers, and even recognise and edit images. Spot is no different. With ChatGPT under the hood, the robot began captioning what it could see using the cameras dotted around its body. It then runs this info through a visual-question-answering AI model to get more context before generating a response.

By the looks of things, you'll have to lean in pretty close for Spot to hear you. That may not appeal to those put off by the robot's rigid movements, or that vice-like grip protruding out of its hinged neck.

The company says its goal was to make the bot more entertaining and nuanced, rather than factually accurate. Using a heavy dose of sarcasm, Spot sounds like Marvin the Paranoid Android from The Hitchhiker's Guide to the Galaxy. Asked to describe what it sees, the droid sullenly says: “I see the unfathomable void of my existence reflected in this QR-code filled board. Oh and also a large window.”

When it does talk, Spot’s gripper starts flapping to make it seem like it’s mouthing the words. The end result looks a bit like a glove puppet, albeit one with a hard bite.

The AI upgrade could make robots better problem solvers at work (Boston Dynamics)
The AI upgrade could make robots better problem solvers at work (Boston Dynamics)

So, could AI usher in new roles for robots? Boston Dynamics reckons the upgrade opens up the possibility of lateral thinking which, it insists, doesn’t mean that the machines will become sentient.

By associating between English words and culture, the bots can now answer more convoluted and indirect questions, the company explains. If they can't keep their traps shut, you may even be able to eavesdrop on them when they conspire to take over your workplace.

Jokes aside, AI could make robots better problem solvers, according to Boston Dynamics. In the clip, when Spot was asked to show its parents, the bot walked over to an exhibit of old robot dogs and picked out the earliest model.  This, Boston Dynamics implies, showcases its ability to inventively navigate vague or tricky questions that don't necessarily have a correct answer.

“A world in which robots can generally understand what you say and turn that into useful action is probably not that far off,” writes Boston Dynamics in a blog post. “That kind of skill would enable robots to perform better when working with and around people — whether as a tool, a guide, a companion, or an entertainer."