Nvidia Isaac Is A Virtual Simulator For Training AI And Robots Speedily And Safely
Isaac aims to make robot testing a more scalable process than real-world trials
Nvidia wants to help train robots in a virtual environment before such autonomous machines are released into the real-world, something its aims to achieve with its Isaac simulator.
Built on top of Nvidia’s Jetson machine learning and artificial intelligence (AI) platform, Issac uses Nvidia’s graphics processing technology to render a realistic virtual reality environment in which AIs can self-learn through carrying out tasks and interacting with the simulation.
The idea behind Isaac is to ensure AIs and smart systems are trained well enough to be safe when they are sent out into the real-world for more testing and fine tuning.
Robot simulator
The use of virtual environments should also help AI and robot developers to set up and run tests far faster than they could if they had to conduct them solely in a physical environment.
“We need to create an alternate universe,” said Nvidia chief executive Jensen Huang, noting the need to create a very realistic virtual environments to accurately train AI and robots; Isaac achieves this by using the Unreal Engine 4.
“We would simulate in this environment and and run it on top of any Nvidia GPUs, and inside that computer is a virtual brain and when we’re done [with training ] it we literally taken that virtual brain and put it into a real robot. And this robot wakes up almost as if it was born to know this world and the last bit of domain adaptation it does is done in the physical world,” he explained.
At Computex 2017, Nvidia demonstrated how an AI powered robot can be trained to play hockey from scratch through using Isaac.
The system has the advantage of being able to train multiple robots at a time, a task in the physical world would be difficult to do at scale. By taking the virtual brain of the smartest robot in the simulation and put it into other robots, thereby enabling the evolution of that brain’s learning to be further scaled across multiple robots.
Repetition of this process essentially enables the speedier iteration of AIs and their deep learning systems in a environment where testing is more fluid, safer and scalable than real-world testing in the early stages of development.
With Nvidia continuing to develop more AI based technology, it is likely that Issac will be put to use by companies such as those working on driverless cars.
Furthermore, with partnerships with the likes Hewlett Packard Enterprise, Nvidia is bringing more of its AI software and hardware to the cloud and data centre, further opening up access to deep and machine learning technology.
