Machine Learning – The next big game changer


cars


In the recent history there have been many great disruptions in the software development. If one should be named, I would select the usage of public clouds. The cloud has enormous computing power, superior capacity and great tools which are ready to use. The cloud can quickly be taken into use and it is cheap. The productivity, reliability and even security of the services have improved a lot, and the cloud has proved a great business value.

The next big game changer and the major enhancer of the business will be the efficient usage of machine learning, deep learning, computer vision and related areas. I call it as the era of the machines. Some companies are already using machine learning and even in a large extent, but my view is that the breaker is still coming. For a consumer this could mean, for example,

  1. Superior gaming experience based on machine learning and virtual and augmented reality
  2. Machine learning is applied to 3D point clouds, 3D modeling and 3D virtual maps.
  3. The value of the information is increased by machine learning algorithms.
  4. Machine learning is used in 5G near-field sensoring, IOT and applications.
  5. Autonomous vehicles use machine learning algorithms in their driving logic.

The key is the data. Machine learning finds correlations, patterns and structures from the existing data. If the new data is having similar structures, machine learning can make good predictions. There are two main attributes that affect on the accuracy of the predictions: 1) the number of similar correlations or frequencies in the data and 2) the amount of data. If there is less structure but more data, the predictability can still be good.

In future we are surrounded by the data. Data needs to be pre-processed and handled in real-time before it is collected to larger containers, and the operations are performed in grid like networks. Machine learning and security must be there where the processing occurs and where the data is located.

Machine learning has many applications; for example, it opens new possibilities to track improper usage and patterns. Machine learning has been used to detect credit card misuse and faulty money transactions for a long time. On the other hand, in wrong hands machine learning makes threats more complex and difficult to prevent. Machine learning can find patterns in weaknesses, learn from the errors and bypass the shields. It is possible that machine learning can be used to break up a certain type of encrypting (see https://mlconvex.ai). Question here is do prime numbers follow some structure or patterns. Generally the answer is no, but in some other geometry or space the answer might be yes. Machine learning is a kind of the preliminary stage of quantum computing.

The era of the machines will change the world dramatically. The reasons why “great” things have not been invented earlier or today, among other things, are

  1. Digitalization: Companies are struggling with their digital journey; digital transformation has been started but it is not on that state that the machine learning can be utilized.
  2. Demand: Business or a buyer does not know what machine learning can do or what is the competitive advantage it can offer.
  3. Skills: If there is not demand, there is also no supply.
  4. Group of talents: In order to solve new things and new problems, this requires a group of talents where all required skills are represented; like business, programming, cloud knowledge, machine learning algorithms, data, mathematics and other domain expertise. In many cases one or more key skill is missing, time and money are spent significantly and the results are low in value. However, it is not necessary a fault of the machine learning, but the problem setting has failed or the team is wrong.
  5. Leadership: In order to create a solid roadmap, to understand what is needed, to build the right team, to shape the problems, etc. this requires a good leadership. Start from problems that have value and are solvable. Set milestones where the investments are returned. Make something and improve.
  6. Other priorities: Machine learning should be seen as a next revolutionary opportunity and as an enabler to grow the business to the next level.

My claim is that in most businesses machine learning can be used to achieve significant business results. However, a specific focus may be needed in the problem setting. Data correlations, patterns or structures need to be recognizable by the machine. This may require some data manipulation, filtering, a new point of view to the problem, finding the essential or the core behavior of a complex problem, use of some unconventional approach like solving some other problem from which the answer is derived, redesign of the mathematical model or similar. Business and mathematical frameworks need to be well understood. Great results are achieved by creative thinking and with a group work of talents.

When the machine learning field is not well organized, the market is fragmented, the machine learning strategy is missing and there are lack of visions, the early adopters can achieve significant leaps in their businesses and gain competitive advantage. Are you the one?

More topics on Blog and articles.