Machine learning is the current frontier of computer science research. Tech corporations like Google, Amazon, and IBM use machine learning algorithms in their software to emulate the way humans learn useful truths about the world by observing the information all around them.
True and False Generalizations
When a jogger notices that dogs bark at her every time she jogs through a neighborhood at night, she might not learn anything about dogs in general except that they often bark at strangers. When that jogger encounters several stray dogs that don’t bark at her in the same neighborhood at night, even as all the dogs behind the gates in their yards bark like crazy, she may begin to generalize that dogs bark at strangers only when they are situated within their territory. When she jogs through the same neighborhood during the day without getting barked at by any dogs, she might refine her generalization to say that dogs automatically bark at joggers only at night and only when they are situated within their territory. Now this jogger has a useful bit of wisdom to guide her on her path through the neighborhood.
Arguably, general rules extracted from observations about the world are the basis of all knowledge. But generalizations made by people are frequently incorrect and based on ugly impulses such as resentment or bigotry. These false generalizations are structurally identical to the wisdom that has undergirded human civilization since the beginning of recorded history. They are general rules extracted from observations about the world, but they are false. They can come from flawed people using motivated reasoning or from careless assumptions based on too little information.
Computers, on the other hand, are wisdom-extracting machines with no such shortcomings. Machine learning emulates the process of extracting general rules about the world from a sample set of observed facts. ML algorithms do make mistakes, but they have much smaller error rates than humans. Instead of interpreting observations to conform to preconceived beliefs like humans do, computers simply report what they see without bias. For these reasons, computers are much more efficient and accurate in their accumulation of wisdom than humans. They are still much slower at learning than people, however, and they cannot easily apply wisdom acquired in one domain to problems found in another. When advances in technology enable computers to overcome these limitations, general artificial intelligence will be smarter than human intelligence.
How Machine Learning Is Used
According to Forbes, ML is used to perform common computing tasks such as image processing and language translation. These tasks require the ML algorithm to begin with a limited number of constraints that a programmer can describe relatively easily in a Python script used as program input. When the computer has been given this initial set of rules, it can then be fed a teaching set of images, videos, texts or audio clips to use for extracting general rules about the subject defined in the program input. As the teaching set grows, the generalizations the computer is able to make become more accurate, and its wisdom in the specialized domain defined by its program input becomes more useful and true.
Researching new algorithms for machine learning and AI is an exciting career choice for many computer scientists. If you find artificial intelligence fascinating, you may want to learn more about how machine learning can benefit humanity.
- Does Data Science Include Machine Learning?
- Is it Risky to Have an Alexa or Google Home?
- Top 20 Most Affordable Bachelor’s in Data Science