Machine learning is a subfield of computer science and a specific application of data science that involves developing and deploying algorithms to provide a computer, a software program, or a process with the ability to learn without being explicitly programmed.
Central to machine learning is the use of algorithms that can process input data to make predictions and decisions using statistical analysis. Thus, instead of manually analyzing data or inputs to develop computing models needed to operate an automated computer, software program, or processes, machine learning systems can automate this entire procedure simply by learning from experience.
Benefits of machine learning: Advantages and opportunities
The collective benefit from using machine learning does not rest alone on analyzing big data to extract information automatically. Organizations are using machine learning systems primarily to automate the operation of a computer or software program, thus automating intended processes or tasks; or to complement other methodologies used in data analysis.
Benefits from using machine learning create several opportunities that further translate to variety in applications. Take note of the following specific benefits from and pros of machine learning:
1. Supplementing data mining: Data mining is the process of examining a database or several databases to process or analyze data and generate information. Take note that the pervasiveness of the digital information age has lead to the generation of big data at a faster rate, thus making manual data analysis and interpretation impossible.
When applied in the field of data mining, machine learning does not only automate the analysis of big data but also provides actual assumptions that can be used to support decisions. Remember that data mining is about discovering properties of data sets while machine learning is about learning from and making predictions on the data.
2. Continuous improvements: In their book “Data Mining: A Knowledge Discovery Approach,” computer science professors Krzysztof J. Cios et al. defined learning as the “ability of an agent to improve its performance based on past experience.” When applied in the context of machine learning, algorithms function as agents that can continuously improve the performance of a particular machine learning system using historical data.
Continuous improvement through machine learning has many practical applications. Consider a machine learning system for predicting consumption patterns for a particular market as an example. The system will not only reference historical data to make predictions for a particular period but will continue referencing newly acquired data to do improve how it analyzes consumption patterns.
3. Automation of tasks: The more practical benefit of using machine learning involves the development of autonomous computers, software programs, and processes that can lead to automation of tasks. By supplementing data mining and through continuous improvement, machine learning systems have been developed and deployed to perform tasks on their own.
Automation can complement human activities nonetheless. MIT professor Erik Brynjolfsson and MIT program director Andrew McAfee mentioned that automating one or two steps within a process through machine learning would mean freeing up humans to focus on more critical tasks.
There are notable examples of tasks being automated by machine learning. Google has used this technology to index and rank websites in its search engine. Both Google and Facebook also use proprietary algorithms to deliver online advertisements. Intelligent personal assistants such as Siri from Apple and Google Now from Google use machine learning to answer questions, make recommendations, and perform actions.
Autonomous driving technologies, face recognition, loan application processing and fraud detection, diagnosis of diseases in healthcare, and drug discovery or formulation are other examples of automated tasks. Nonetheless, an expansive range of industries and sectors are currently taking advantage of machine learning systems to introduce innovations, improve the delivery of their services, and promote efficiency in operations.
Limitations of machine learning: Disadvantages and challenges
The benefits of machine learning translate to innovative applications that can improve the way processes and tasks are accomplished. However, despite its numerous advantages, there are still risks and challenges. Take note of the following cons or limitations of machine learning:
1. Error diagnosis and correction: One notable limitation of machine learning is its susceptibility to errors. Brynjolfsson and McAfee said that the actual problem with this inevitable fact is that when they do make errors, diagnosing and correcting them can be difficult because it will require going through the underlying complexities of the algorithms and associated processes.
2. Time constraints in learning: It is impossible to make immediate accurate predictions with a machine learning system. Remember that it learns through historical data. The bigger the data and the longer it is exposed to these data, the better it will perform. For example, using a system to play games and beat human opponents would require feeding the system with historical data and continuously exposing it to newly acquired data to make better predictions or decisions.
3. Problems with verification: Another limitation of machine learning is the lack of variability. Brynjolfsson and McAfee said that machine learning deals with statistical truths rather than literal truths. In situations that are not included in the historical data, it will be difficult to prove with complete certainty that the predictions made by a machine learning system is suitable in all scenarios.
4. Limitations of predictions: Brynjolfsson and McAfee reminded that unlike humans, computers are not good storytellers. Machine learning systems cannot always provide rational reasons for a particular prediction or decision. They are also limited to answering questions rather than posing them. In addition, these systems does not understand context. Depending on the provided data used for training, machine learning is also prone to hidden and unintentional biases. Human input is still important to better evaluate the outputs of these systems.
A note on the pros and cons of machine learning
The aforementioned benefits and opportunities, as well as the limitations and challenges associated with machine learning should not be regarded as absolute truths. Remember that there are different ways to develop and deploy a machine learning system for more specific applications such as detection, classification, and characterization, among others. The advantages of a machine learning system are dependent on the way it is developed for a particular purpose. Furthermore, the limitations of machine learning are dependent on the type of application or problem it is trying to solve. It is also important to note that these limitations generally revolve around the quality of data and processing capabilities of involved computers.
FURTHER READINGS AND REFERENCES:
- (1) Brynjolfsson, E. & McAfee, A. 2017. “The business of artificial intelligence: What it can and cannot do for your organization.” Harvard Business Review
- (2) Cios, K. J., Pedrycz, W., Swiniarski, R., & Kurgan, L. A. 2007. “Supervised learning: Decision trees, rule algorithms, and their hybrids.” Data mining: A knowledge discovery approach. NY: Springer Science and Business Media