Tesla CEO Elon Musk announced that Tesla is in the process of developing a Neural Network training supercomputer.
The computer, codenamed “Dojo”, will be able to “process truly vast amounts of video data”, according to Musk.
Musk stated via Twitter that the computer will be available in a “limited public release in 6 to 10 weeks”.
What is a Neural Network?
In essence, a neural network is a type of machine learning which models itself after the human brain. The “machine” that is “learning”, so-to-speak, is a computer that learns to perform some tasks by analyzing algorithms.
Those algorithms are sets of data, and the computer processes them in a way that mimics the way the human brain operates.
Neural Networks are applicable in a wide array of industries: Finance, Trading, Medicine, Gaming, Driving Automation etc.
How Neural Networks learn?
Neural Networks cannot be programmed directly by developers in order to perform a certain set of tasks. Just like a child’s developing brain, they need to slowly process and learn a certain dataset.
How do Neural Networks Apply to Tesla?
The goal of Dojo is to enable Tesla cars to operate using little to no human assistance. The computer’s main task is to create images of the car’s surroundings and navigate traffic accordingly.
As a result, self-driving cars will be able to make moral decisions alongside technical driving decisions. For example, if an inattentive pedestrian suddenly walks in front of a Tesla car, the latter should make a quick judgment call and avoid running over the pedestrian.
What is the purpose of developing Dojo?
Tesla is vying to increase the capabilities of its Autopilot functionality.
Consequently, Dojo is Tesla’s campaign to develop an improved version of the aforementioned self-driving system. The main goal lies in feeding larger datasets to its training system more rapidly.
In general, vehicles require vast sums of data to learn to drive on their own.
At the moment, Tesla’s full self-driving package includes limited autonomy, such as Navigation on Autopilot. The Tesla car drives along the highway and turns off at the correct, predesignated exit based on the user’s input.
During Tesla’s Q2 2020 Earnings Call, Musk stated that Dojo will allow the package to include much more features.
The Tesla CEO remarked that cars will be able to do “traffic lights, stops, turns…everything pretty much”.
Dojo = A Roadmap for Fully Autonomous Driving?
Obtaining answers to that question is a bit premature. During this year’s World AI Conference in Shanghai, Musk asserted that Tesla cars will have “basic functionality” for fully-autonomous driving in 2020.
Even if Dojo manages to accomplish the tasks at hand, rolling out the technology will still be subject to regulatory approval. It remains to be seen if politicians will give “green light” that allows cars to operate without human supervision.
All in all, time will tell!