Hey, Fans!!!
Are you waiting for the latest information based on Google’s robot, which goes to school and takes orders? Well! You are at the right place. As you know, LATE Scientist Fei Xia of Google sat in the middle of an open kitchen last week. Where he entered an order into a laptop attached to a one-armed, wheeled robot that resembled a vast floor lamp.
What Can The Robot Do?
He penned, “I’m hungry.” Immediately, the robot wheeled up to Xia and cautiously took up a bag of multigrain chips with a giant plastic pincer. The most impressive part of the demonstration in Mountain View, California, at Google’s robotics lab, was that not a single human coder had instructed the robot on how to respond to Xia’s demand.
Its commanding software had been trained on millions of pages of text scraped from the web to convert spoken phrases into a series of physical movements.
The History Behind the Robot’s Making
During the demonstration, Google senior research scientist Karol Hausman stated, “To deal with the diversity of the real world, robots need to be able to adapt and learn from their experiences.”
Machines need to understand that there are numerous possible combinations of words that might result in various meanings to communicate with humans. In his own words, “it’s up to the robot to understand all the minor subtleties and intricacies of language,” Hausman argued.
The Initial Stage
Google’s demonstration represented progress toward the ultimate objective of developing robots with the ability to communicate with humans in various settings. OpenAI’s text generator GPT-3 is only one example of the outstanding linguistic abilities of programs developed in recent years by feeding massive volumes of text collected from books or the web into enormous machine learning models.
Google and other significant technology companies utilize these language models extensively for advertising and search. New services have emerged that apply AI’s language capabilities to mundane but essential jobs like code generation and ad copywriting offered by several companies via cloud APIs.
Advancements And Failures
Despite these advancements, AI programs often fail miserably or spout incomprehensible nonsense. It’s possible that careful programming is needed to guide a robot without going berserk properly. Language models taught using web content also lack a grasp of truth and often reflect prejudices or harsh language in their training data.
Demo Robot
Hausman’s demo robot was powered by PaLM, the most advanced version of Google’s language model. It can do many cool things, including using plain English to describe how it arrived at an answer to a query. The same method is used to devise the series of actions the robot will carry out to complete the task.
The robot butler was developed by Google researchers using technology from Everyday Robots, a spinoff of Google’s parent firm Alphabet’s X division, which focuses on “moonshot” research projects. They developed a new program that combines PaLM’s text processing abilities to convert human speech into a set of instructions the robot can follow, such as “open drawer” or “pick up chips.”
The Training Phase: Physical Activity
A different training phase taught the robot how to perform physical movements like picking items up with the use of remote control from humans. Due to the constraints of its surroundings, the robot is less likely to act erratically if its language model makes a mistake.
To Conclude With!
With PaLM’s linguistic abilities, a robot can interpret vague instructions. Join us in a shared experience. The robot’s response was to place the blue block next to the green one.
Follow us for more updates!
