Artificial intelligence has been pursued for many years. The algorithms
developed and tested failed to make any progress until 2007, when
mathematicians used recent learnings about how the brain worked
to rewrite their approach. Since this time, machine learning using
support vector machines, and then neural networks has opened the
door to a more hopeful future of robotics.
Machine Learning Video - Stanford University
In this video, Professor Andrew Ng shares his journey into understanding
the maths behind machine learning, and how it may be applied in
Machine learning is still in its infancy
Humans are not capable of advancing the mathematics of complex
activity, it requires the processing of machines to compute this
in a self-learning environment
The potential for humans to be replaced by machines [robots]
to carry out menial tasks is getting closer
Machine learning comes about through a combination of offline [manual]
and adaptive [automatic] training and learning. We first have to
train a machine as to what is correct and incorrect decisions. Once
a workable benchmark is developed, the algorithm continues to learn
online. With some manual input to further tune the algorithm to
detect new responses to particular patterns of data, we can reduce
the error rate.
This type of computation requires direct processing of real-time
data. Combining such processing with machine learning can provide
a reasoning flow and enable runtime updates of the machine-learning
Machine learning can reduce the time to insight from weeks to minutes.
Its applications are numerous, however we can expect the primary
focus to be on scenarios in manufacturing and IT services.