Email
Password
Remember meForgot password?
    Log in with Twitter

article imageReview: Would a master algorithm solve all our problems? Special

By Tim Sandle     Jul 6, 2019 in Technology
Algorithms are with us, from predicting the music we may like to helping with medical diagnosis. While technology improves how far can it progress and is there a master algorithm that can unlock all mysteries?
This is the question posed by Pedro Domingos in his book The Master Algorithm:How the Quest for the Ultimate Learning Machine Will Remake Our World. The book provides an in-depth overview of machine learning, before testing out the hypothesis that a master algorithm may exist, a code that is able to unlock the secrets to a more powerful form of artificial intelligence.
By unlocking the master algorithm, the author argues, we can make accurate predictions across all fields of knowledge. This would enable humanity to either follow and be cognizant to what lies ahead of seek to change events that are get to come. This could also come down to the level of each person, providing each of us with an individual digital future.
This concept works of we accept the variation of time and understanding that we are not on a trajectory that is a singularity but rather a series if phased transitions, each of which we could potentially alter, if we knew what lay ahead.
The book begins by charting how society is changing, "one learning algorithm at a time", portraying how machine learning is altering science, business, technology and many other fields. Beyond this, the next waves are with technologies capable of building themselves. This is a paradigm shift in computing; there was a time when computers simply computed, giving an answer to a data input. Now computers are creative. The book tests out the limitations - if any - of this creativity.
Computing creativity is central to the learning algorithm, which, at today's rudimentary stage provides the basis of machine learning. At the core of the learning algorithm is the nature of prediction, with machines predicting what we want, learning from when this does not meet our expectations, and using data to improve the nature if the next round if predictions.
The central focus with the book, which leads to the ultimate state of machine learning, is with the idea of a universal master algorithm. This would become the final stage model for all machine learning, capable of processing data from the past, present and future.
One interesting concept that could follow on from this is the eventual fusing of artificial intelligence and the human brain, something possible as technology evolves if humanity wishes to embrace this. This could take human evolution to a new level, although the concept is, naturally, scary to many. This could be especially so if this form of AI is not democratic, an opportunity not available to all. The worst outcome could be a new form of inequality, a new social class, with an elite connected to AI, able to plot the path of the general populace who do not have this advantage.
These are all hypothetical scenarios, of course, and no master algorithm has been created even supposing it is ever possible to create one at all. This aside, the book does deal with the history of machine learning and casts an eye over its short term and long-term future. This is through dividing the book into five main areas: inductive reasoning, connectionism, evolutionary computation, bayes theorem and analogical modelling, before considering the next phase.
The author is well qualified to work on the subject, having been working with machine learning for over twenty years, latterly through his role as a professor at University of Washington. Domingos says society in general needs to understand what is happening with technology and he sees unlocking machine learning as the basis to improving all forms of artificial intelligence.
More about Algorithm, master algorithm, machine learning, Artificial intelligence
 
Latest News
Top News