The advancement of technology has reached a critical point. Artificial Intelligence (AI) is steadily taking over. Machine Learning (ML) is the main branch of Artificial Intelligent technology and it plays a crucial role in the world of economics. Admittedly, it has the potential to revolutionize whole industries and organizations.
Creating algorithms is a really good way to automate a lot of tasks, but it is not always possible to do so. Some processes are really difficult to imitate, thus we have to find new ways to overcome such issues. Thus, a solution to this problem is the application of machine learning technology. ML is based on algorithms targeting to create systems that are able to learn and improve automatically. Initially, these systems are built on a model with imported data, which is called “training data”.
They are leveraged by the system to make predictions and actions without even being programmed to do so. A great example is the checkers game experiment. Robert Nealey was considered a really good checkers player, a game similar to chess with the difference that all pieces have equal value. He was chosen to play against a computer that was built by Arthur Samuel based on ML. Surprisingly, the results were shocking… a machine produced by humans was able to defeat them. This experiment gave scientists food for thought. Moreover, the evolution of machine learning though is strongly dependent on the evolution of its related fields, for example, data science, which plays a key role in the data that are “fed” to the machines.
How it works
There are 3 main phases.
- Firstly, the machine goes through a decision process, based on the imported data. It tries to find a pattern model that these data are “hiding”
- The second step is the evaluation of the already made decision of the discovered pattern. It is seeking relevant examples to compare the accuracy of the pattern.
- The last part is called the optimization process. If there is a chance that the model can fit better with the “training data”, relative adjustments are automatically made until the variance of the model and the data set are minimized to the established error limit.
Machine learning applications are numerous:
- Data Mining
- Energy Production
- Automotive – Aerospace engineering
- Medicine – Biology
Of course, the above are just some examples of the numerous applications of machine learning.
Types of Machine Learning
- Supervised learning
- Unsupervised learning
- Semi-supervised learning
- Reinforcement learning
In addition, questions about the ethics and Bias of Machine Learning usage and its application to Artificial Intelligence have been raised over the years. The collection of “bad” data could lead to unwanted or dangerous results like digitizing human and cultural prejudices. An indicative example is a system used for criminal risk evaluation which was found to be discriminative against black people. Moreover, in such systems, the “AI Black Box Problem” of knowing the inputs and outputs but not the rationale behind the results has produced a debate on the Accuracy and Fairness of Machine Learning Algorithms.
To conclude, another serious issue that scientists may face is widely using ML systems in the healthcare industry. It is believed that it would be a difficult task to “guide” these systems to work favorably to the patient’s needs and not to get money-making machines. Thus, we should create protocols that prevent any kind of such malicious ML usage. Technology exists to help humans and we have to be very careful not to turn this balance against us…
Want to learn more? Check out the video of simplilearn