General Learning (GL): Breaking the Paradigms of Artificial Neural Networks – Part 1

Often a determined discovery, or innovation, carries within itself an even greater, and more relevant, discovery that sooner or later will overcome the previous one, but which at first we do not see.

Technological myopia, if we may say so.

Perhaps the most current example of this is the innovative cryptocoin Bitcoin, which opened the door to disseminating the concept of decentralization of information and Blockchain technology, which was not perceived at first, but is now becoming more relevant and definitive, opening doors to new discoveries and inventions in the most varied areas and independent of the financial environment.

Within this context, is Machine Learning, especially Deep Learning, not experiencing a similar scenario?

I think so, and I call it deep myopia.

But before you ask where is the Blockchain of Deep Learning, which I will name here General Learning (GL), and which probably must have already come to mind, let’s talk a bit about Artificial Neural Networks, or ANNs, which in theory, by my vision and metaphor, are equivalent to the discovery of Bitcoin.

ANNs are a key finding for Artificial Intelligence, especially because of models that seek to approximate the real way our brain seeks to solve problems of information representation. This is a fact, but, for a good part of science, this advance would have the potential to attain the very functionality of our mind, and other more complex systems, such as reasoning itself and perhaps even consciousness, which is undoubtedly a challenge to prove.

But in my understanding, the current growth of Deep Learning (AI/ML/DL) using ANNs has created technological myopia for the fact that we can create generic learning platforms with any kind of algorithm, and without using up even any ANN architecture, ie General Learning (GL) platforms.

In other words, and back to our metaphor, General Learning (GL) are our Blockchain.

Note also that, similarly, just after Bitcoin, the most diverse types of other digital coins were proposed, and after the first ANNs, such as the Perceptron, invented in 1957 by Frank Rosenblatt, and a number of other similar pioneering architectures created by visionaries such as Von Neumann and Marvin Minsky, dozens of new artificial networks are presented to the market.

However, behind this evolution of ANNs, we need to break the paradigm to evolve to create training platforms and generic learning for any algorithm.

That is Deep Learning, without ANNs.

Rogerio Figurelli – @ 2018-12-08


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s