The Economics of AI - Neural Networks and Machine Learning- the New Prime Mover

There are forces at work within the technology sector that will fundamentally change the nature of the economics of production and efficiency  that will disrupt the economy like that which has not been witnessed. At the center of this change are a plethora of developments in various fields of the hardware and software of computation that have been accumulating and maturing. It has been an uneven process but currently the center of the technology world is focusing on the aspects of neural networks and a rapidly advancing machine learning called deep learning. The technology when combined with appropriate data - can and will surpass the level of productivity and efficiency than that of many  human productivity- for such mundane tasks. Added to this layer- are an expansive array of informating that go beyond that of the human capacity in terms of speed of processing, precision in measurement,  vision enhancements in colour and levels, sounds and many other aspects of machine based data analysis.  As one leading developer recently stated- any task that a human requires a few moment to think about or perform- the new deep learning systems can be taught at a cost to automate such work  that over a time horizon produce outputs that far surpass that of living labour of the human. However it is not just these mundane tasks that will be automated. When the new smarter machines are designed put alongside to  complement the human within most production process - the outcomes of such a work effort will far surpass that of what we see today of either the machine or the human acting alone.

Industry insiders have compare these new deep learning AI systems as  the as the new electricity- that will transform industry and society  in a similar manner and form the core of a new massive level of change and innovation.

It is interesting from an economics perspective not only from how it challenges and transform the nature of production within the economy and society but it also has the potential to change the field of economics. 

LivingWork has been actively been involved in some of the early stages of the applications of such technology and has developing tools and experimenting with new measurement  vehicles within the realm of deep learning in neural network design on 3 projects that apply some of the latest technology.

To help develop more interest, understanding and evaluation into the applied research of the economic field LivingWork Analytics will provide and outline and update to the projects it is currently working on within these fields. The projects are mainly experimental - but the end goal is to develop a useful set of new statistical and monitoring tools into some critical areas.  More soon.

Current Projects: (updates coming soon)

1. Developing a Multilayer Perceptron (MLP) to Predict Low Wage Workers From Labour Force Micro Data

This given here is no wage data collected for the self-employed in fact there is very little data collected on the self-employed by Statistics Canada. It is a large data gap indeed. However LivingWork Analytics has constructed a Neural Network binary based prediction system to determine whether a worker is low wage or not. The neural network is a multilayered perceptron with 37 input variables from the LFS micro file. The network was designed with two hidden layers one with 120 nodes and the second with 54 nodes. It was trained on over 250,000 employees from the LFS micro file- to identify low wage workers. Using a relu and sigmoid function on the hidden layers- the neural network achieved a 91% accuracy rating on the training file- and a 88% accuracy rating on the test file. That is it successfully predicted the workers as being low waged or high waged on the known labelled training set. Such predictive powers are quite efficient. The stated neural network will be employed to estimate the low wage status of the self-employed. The NN is still being trained to achieve even higher levels of efficiency. More information on this model is available at LivingWork.ca.

2. Kaggle Contest Participation: Use satellite data to track the human footprint in the Amazon rainforest

3. Recurrent Neural Networks and Labour Market TIme Series Analysis- detecting signals from the noise