Automation is inescapable in the news, but we need to take the time to put business and technical developments into wider perspective. Will reusable and disposable aerospace technology democratize automation? How is China’s self-driving research a global market play? How important is ethics to the development of artificial intelligence?
Robotics Business Review has partnered with me to bring you a weekly roundup of the top robotics developments. This week, we look at AI and self-driving research developments from rivals Google and Baidu, as well as the latest questions around responsible training of AI. Are you ready to be updated?
Google works on limitless AI
Currently, artificial intelligence learns from input. Then it is deployed to complete a specific task or solve a problem. Google Inc.’s Brain AI Lab has turned this on its head with its Generative Adversarial Network (GAN). The goal of GAN is to have AI create content such as images and videos from its own understanding of what images and videos are. Google will test GAN by deploying two algorithms to compete with each other. One will create the content, and the other that will “critique” it.
The implications of what Google is working on should not be understated. Google is essentially looking to develop AI that is capable of learning, behaving, and communicating without any human input.
Automated spacecraft to dock
China this week launched its first cargo spacecraft, named Tianzhou 1. The goal is to have the cargo spacecraft dock with Tiangong 2, China’s space lab, to deliver supplies and refuel. There are no humans present — the entire process will be automated.
This demonstrates the growing importance of robotics in near-Earth space. Specifically, it isn’t just about sending robotic rovers to other planets or developing robotic arms, innovations that have come from established space powers like the U.S. and Russia.
Robotics could help China to compete in a leaner, faster, and more frugal way. Technologies such as automated spacecraft will likely play a significant role in future missions, such as China’s planned visit to the dark side of the moon in 2018 or its Mars mission in 2020.
The U.S. Marine Corps Warfighting Laboratory is testing the Tactical Air Delivery, or TACAD, unmanned aerial vehicle (UAV). It can be launched from a plane, deliver supplies weighing up to 700 pounds, and then discarded. To build the drone, the military is looking to “hobby-grade” technology and working with more affordable materials, like those found at hardware stores.
TACAD reflects a major shift in the drone industry. Previously, UAVs were a major investment, so they were used until they broke or were upgraded. Now, they could reach a cost point where they are used once and then thrown away or recycled.
From industrial automation to consumer robots, this approach could change not only how people build, sell, and use robots, but also how they perceive them. Frugal technology would allow more people access to robotics, specifically in emerging markets such as China, India, and Brazil.
Baidu makes self-driving research open source
Baidu Inc., the Chinese equivalent of Google, has announced that its Apollo autonomous vehicle project will be open source. Baidu said its platform will be accessible to automakers and other companies working on self-driving cars.
The company borrowed the project name from NASA’s lunar missions, which involved companies in multiple industries working together for a single objective. Baidu claimed it views self-driving research with the same importance.
At first glance, Apollo appears to be a way for Baidu to expand beyond the search vertical and become a leader in autonomous vehicle innovation. However, Apollo is more than that. It is Baidu’s way, and therefore China’s way, of dominating a market that is fast becoming global.
Instead of trying to push physical vehicles, which consumers may think twice about purchasing, China is focusing on the underlying software. Apollo may be a self-driving research strategy, but it has geopolitical undertones.
More About Global Robotics and AI:
- Smart Manufacturing Can Help China Become a Global Automation Leader
- The Impact of Automation Will Be Indirect as Robotics Develops
- Fears of Robots Taking Jobs Require a Response, Says LivePerson
- International Robotics Rivalries Intensify Amid Calls for Jobs Policies
- AI Rules Are Necessary, Say European Regulators
- Five Geopolitical Flashpoints That Could Affect Military Automation
- Connected Cars Yield Useful Data Analysis Through Xevo
- Asian AI Could Dominate CES 2017
AI can fall acquire bias
Do you trust artificial intelligence to be unbiased? If so, pay attention. A new study, involving researchers from Princeton University, the University of Bath, and more, has found that AI can have “racial and gender biases.” The main reason is because many AIs are studying human language to become smarter. In the process, they are absorbing the stereotypes and prejudices of human speakers.
There are already concerns that AI could lead to dangerous behavior without the proper safeguards built in. Like human children, machines could learn the wrong things from their “parents.” Some observers have raised questions about whose ethical principles we embed in robots and AI. Each developer is naturally influenced by his or her own culture, employer, and more.
A single solution to this problem is unlikely. What will probably emerge in the short term is that machine learning, service robots, and self-driving cars will have different “personalities” based on their creators and expected uses. It remains to be seen, though, what this will mean for business, government, and society as a whole.