Deep Learning, Magic, and the Internet of Things

November 2, 2015
Enlighted Inc.

Enlighted attended the recent Greenbiz Verge conference in San Jose, California. The conference agenda was rich in expert dialogue focused on leveraging technology and social innovation to solve the world’s most urgent problems.

Greenbiz founder Joel Makower hosted a Q&A with legendary tech investor and sustainability champion Steve Jurvetson. Jurvetson is a partner in Draper Fisher Jurvetson and sits on the boards of Tesla, Space X, and Enlighted.

Jurvetson described his investment philosophy as the pursuit of sustainable value in themes that span industries and will garner more than just a financial return on investment. He described profit as a beneficial side effect of good investment rather than the primary goal and described how a banker or financier mindset leads investors to miss opportunities due to a focus on dollars rather than real value.

Jurvetson believes we have reached the point where neural technologies in the form of learning computers will help us to solve many of the world’s most complex challenges. He calls this trend “deep learning.” The goal of deep learning is to mimic the brain. We do this by connecting a network of virtual nodes and creating a generic learning machine. We then train this generic brain to learn things – to recognize patterns within a specific problem domain. So instead of programming our computers to do something, we program them to learn. It is about building systems that adapt.

The Internet of Things is built on this philosophy of deep learning. In smart transportation, for example, we connect a network of nodes — vehicles and the sensors within them. Once these nodes are connected and gathering/transmitting data we have a neural network – a learning system.

Human software developers can then build application layers that harvest the learning from this network and provide additional services and functions that improve the health of the system over time.

This type of learning can optimize any system that is too complex for the human mind alone. Why build new roads, Jurvetson asked, when we can program our network of cars to use the ones we have more effectively. Or agriculture – we need to produce as much food in the next fifty years as we have in all of history to date, he explained, and it’s unlikely we can do this without learning machines. Or telecommunications. We have hundreds of new satellites launching that observe the Earth and gather data. There will be too much data for humans to parse, we need computers that work with humans to make sense of it.

The current engineering paradigm is brittle, he explained. It is top-down, command and control decision-making, and is not adaptive enough for today’s complex, uncertain world; it does not promote resilience.

The Internet of Things, Robotics, Deep Learning, these engineering models will make us more agile and resilient, he argued. They will represent and encapsulate our collective learning and allow us to adapt faster to new challenges as they arise.

Share This Page