Let's Open Data and feed a smarter AI.
"Sharing is good, and with digital technology, sharing is easy." — Richard Stallman
Scientists and engineers are playing with different approaches to imitate intelligence, that is to replicate things humans can naturally do effortlessly to make sense of the information in our world. One of the most promising approaches is trying to simulate some functions of the brain by building artificial neural networks.
Artificial Neural Networks
Deep Learning is a technique that uses neural networks organized in multiple layers to learn from a massive amount of data and draw conclusions about it.
Neural networks are built around a simple computational unit called the neuron connected to many others creating a network. Today, they are implemented using artificial neurons run by software that tries to imitate how the human brain works.
Biological Neural Networks. "The Brain"
The human brain is much more complex and really difficult to replicate as we are yet far from understanding it in full. It is estimated that there are 100 billion neurons and thousands of trillions of synapsis.
Thus far, artificial neural networks haven’t even come close to modeling the complexity of the brain, but they have shown to be good at problems which are easy for a human but difficult for a traditional computer, such as image recognition and predictions based on past knowledge.
The Challenge
There are many teams of researchers around the world testing different models of neural network architectures and engineers figuring out new solutions using them.
This technology has been in a hype during the last years as it’s being used increasingly in the industry to solve problems that were in the past reserved only to the human intellect. Although people started experimenting with neural networks in the 50’s, it was not until the last years that a combination of a sheer volume of data to learn from, the use of GPU and massively parallel computing, and new learning approaches, made them practical to be applied to bigger challenges. Their advocates once called lunatics, now have the attention of the tech giants.
Google, Facebook, IBM, Baidu and Microsoft among others, have been building internal teams for researching practical applications of this technology that could bring us a better understanding of the world we share every day in photos, videos, searches, audio, etc. They are even releasing many of the tools and frameworks their internal AI teams use: Tensorflow, Infer.NET, Facebook AI.
Why is Open Data more important than ever?
As of today, for this technology to work we need a lot of data. Most of the biggest tech companies leverage their gigantic databases to find new solutions and insights. Although their sharing of tools is useful to the community, we need to do more to open more data to the public to move the field faster.
Maybe the new Open AI non-profit could be used as a hub to share many more public datasets along with the tools.
I believe that this technology has the potential to increase the productivity of our civilization multiple times thus giving us more free time. We all will benefit from opening more data to the world so multiple teams and startups that might be working on new ideas may use it to make a smarter AI and find the answers much needed by our society.
