British bulldog! Thanks to ChatGPT, Boston Dynamics’ creepy robot dog can now speak with an English accent

As if walking a robot dog wasn’t strange enough, Boston Dynamics’ spot can now take visitors on a walk as it takes on the role of an English tour guide.
In a new video, the engineering firm demonstrated Spot’s new ability to answer questions and crack jokes using a range of accents as well as several different personalities.
The robot, decorated with little hats and googly eyes, takes guests to different locations and describes what it sees.
As he opens and closes his gripper to mimic a mouth and turns to look at people, Spot’s performance comes impressively close to that of a real leader.
Powered by ChatGPT, Spot’s creators say they were surprised by some of the unexpected responses the robot dog produced.

Spot the robot dog now has the ability to play tour guide, answer questions and describe the world around him using ChatGPT
The footage shows the $75,000 (£61,857) robot taking on the personality of a “chic butler” and says: “My linguistics have been carefully crafted to provide an authentic British experience.”
When senior software developer Matt Klingensmith asks if he enjoys his work, the robot replies, “Ah, Mr. Matt, my job as a tour guide is a great satisfaction.”
“I find the dissemination of knowledge quite rewarding, don’t you agree?”
However, this is far from the only character that the robot dog plays, as it takes on the voices and characteristics of a “precious metal cowgirl”, an “excited tour guide” or a “nature documentary”.
The robot also demonstrates an impressive ability to recognize and respond to objects in the world.
While taking on the role of a “1920s archaeologist,” Spot refers to the camera crew, calling them “a fellow explorer with a camera.”
One of the most surprising characters is “Josh,” a sarcastic and moody robot handler who Mr. Klingensmith says was “an experience I’ve never had with a robot in my entire life.”
When asked for a haiku about the space, the Josh personality replies, “The generator hums quietly in a space devoid of joy, much like my soul, Matt.”
However, all of these different features and personalities are the product of simple adaptations of the same code.

Using a variety of prompts, the large language model behind Spot is able to create a full spectrum of personalities from “Fancy Butler” to “Josh.”

Boston Dynamics provides the AI with this map of the offices, including some labeled locations and some short descriptions
The team provided ChatGPT with a carefully designed prompt that converts visual and audio information into speech and commands for the robot.
Prompts include information such as “Use the Tour Guide API to guide guests through a building using a robot.” Tell guests about what you see and make up interesting stories about it.”
They add context like: “Personality: “You are a snarky, sarcastic robot who is unhelpful””.
The model then receives a plan of the building with some labeled rooms and short descriptions.
However, the AI is given just enough information to get started.
Its creators say: “The LLM can be thought of as an improv actor – we provide a rough script and the LLM fills in the gaps on the fly.”
To convert visual information into text that ChatGPT can use, Spot uses a program called the Visual Question Answering model that labels images from the robot’s cameras.
With all this information, ChatGPT 4 can command the robot to move around its environment, respond to questions, and make comments about its surroundings.
Some of the behaviors Spot was able to generate with ChatGPT at its core surprised even its developers.

“Josh,” the sarcastic tour guide, was one of the most surprising characters and a very different experience than most previous robot-human interactions

This diagram shows how information in the real world is converted into prompts that the AI can use to control the robot’s speech and movement
In a blog post, Mr. Klingensmith and his team point out that large language models (LLMs) like ChatGPT tend to produce “emergent behavior” that is not predicted by the code.
In one example, the team asked Spot who his parents were, and the robot led them to the display of older Spot models, saying that these were his “eldest.”
Likewise to the question “Who is Marc Raibert?” The robot replied: “I don’t know.” Let’s go to the IT helpdesk and ask.”
No parental concepts or a command to ask for help were programmed in, although the team insists this is not evidence the robot is thinking.
“To be clear, these anecdotes do not suggest that the LLM is conscious or even intelligent in the human sense – they simply demonstrate the power of statistical association,” the team writes.
“But the noise and mirrors the LLM puts up to appear intelligent can be quite convincing,” they added.

Boston Dynamics added (1) a navigation system, (2) a microphone, (3) a speaker, and (4) a gripper arm and camera
This is not the first role that Boston Dynamics has proposed for Spot, as the company has outfitted the robot for various purposes.
Spot engineer Zack Jackowski previously said that the next big application of the technology would be the development of robotic guards that can patrol industrial facilities and factories.
Mr. Jackowski suggests that Spot could autonomously walk through locations and collect data with its sensors to detect problems such as open doors or fire hazards.
Spot has been used to inspect nuclear power plants, oil rigs, construction sites and even has an eye on the ruins of Pompeii.
Boston Dynamics’ other robots have also set the bar for bipedal mobility, with videos showing Atlas jumping and performing parkour with ease.
A recent video showed how Atlas could help on construction sites by transporting bags of tools and lumber up and down scaffolding.