7 November 2018

Intelligent patents for artificial intelligence

Thankfully, not all leaders in the field hold such a bleak point of view of the future of AI. Google’s AI chief John Giannandrea believes comments such as those from Musk to be overblown and simply fear mongering2, and AI is already transforming many of the less dramatic aspects of our lives. Speech recognition and natural language processing have provided us with intelligent personal assistants, such as Amazon’s Alexa and Apple’s Siri, and it seems only a matter of time before driverless cars become a common sight on our streets. Given the technological advances that are involved in these, and other, applications of AI and its potential applicability to just about every industry, it is inevitable that there will be an increasing interest in patenting AI related inventions.

The definition of AI in its broadest sense can include any intelligent activities performed by a computer; from speech recognition to autonomous agents. AI is itself not a new technology; the ideas behind neural networks, for example, can be traced back to the 1940s and 50s, when neural networks were proposed as a method for computing logical functions and for incremental learning using supervised feedback3. It is not until recently though that computing power has made the implementation of AI on a widespread basis feasible.

As mentioned above, one method of implementing an AI is a neural network (NN), which is loosely based on the neural system of the human brain. Neural networks are formed from non-linear units called neurons, which are interconnected to produce a network which is capable of computing a wide range of logical functions. Each neuron accepts a number of inputs, performs a weighted sum of them, and outputs a value in dependence on that sum. NNs are typically arranged into layers of neurons, with each layer accepting the outputs of a previous layer as its input. So far this does not sound particularly spectacular; there are far more straightforward and transparent ways of calculating functions on a computer.

To read the full article, please click below.