Hello everyone, My name is Kenshin Fujiwara and I am the CEO and founder of HACARUS Inc.
In this series of blogs, I will discuss a wide range of topics of AI, from the history of AI to practical tips for successful application of AI projects. I hope my blog posts will help you gain a better understanding of AI and solve your business issues.
Following the last blog post, today I will continue to discuss the advance of AI technologies and the importance of machine learning in the AI field. Before jumping into the topics, I will briefly introduce the history and current status of AI first. This way, we can have a strong foundation built on basic AI knowledge.
History of AI
To start, the term artificial intelligence (AI) was first introduced in the 1950s. In this sense, AI has been around for more than half a century. Back then, AI was very rudimentary, only involving simple human interactions, maze solving, and so on. Moving into the 1990s, the “expert system” really started to see some significant improvements. The basis for AI back then was a system that broke down the findings of medical, legal, and other human experts into “yes” and “no” conditional branches.
Along with the computer boom, AI mechanisms were becoming more sophisticated. The idea was that as AI advanced, it would be able to make judgments about any event in the same way that humans can make judgments based on our accumulated knowledge and experience.
While it seemed promising, this attempt at utilizing AI had its limits. In fact, trying to obtain sophisticated answers by combining “yes” and “no” conditional branches requires a large number of resources to organize the division rules for the system. It also takes a lot of effort to keep the system up-to-date in regards to any legal, business, or scientific developments.
Thinking about it, the “decisions” we make when it comes to business and life choices are incredibly complex. On top of that, we often have to make decisions based on ambiguous information. It is hard to even imagine how difficult it would be to code this level of complex “knowledge” to a level where a computer can make a ‘“yes” or “no” decision.
In addition, the landscape of society is constantly changing and evolving. It’s common for new knowledge to emerge and previous knowledge to become obsolete. All of this is a tricky area to keep updating AI as AI itself continues to be under development. In the end, this form of AI proved to be applicable to only a specific range of problems. From this point on, AI development has entered a drought with very little development.
Machine Learning, Major Breakthroughs in AI
Although AI development was in a bit of a slump, academic research was still making progress in a variety of fields. Among these, the study of “machine learning” would eventually lead to major breakthroughs in AI.
Machine learning, as the name suggests, is a mechanism where the AI learns autonomously as a machine. When we get down to it, humans want AI to perform its processing independently without or minimal human intervention. That is to say, the possibility of an AI that can make decisions similar to those of humans. Machine learning can achieve this by reading the data and then successfully analyzing it with the power of a program.
The essence of machine learning is that AI learns how to divide the data well by reading a massive amount of data. In the process of learning, AI finds its own way of dividing the data and is able to make complex decisions and predictions, similar to that of a human.
Previously, I mentioned the “expert system” and how it was created by an individual who defined the division method for the “yes” and “no” conditional branches. However, the major difference with machine learning is that the system itself tries to discover the appropriate division method based on the data.
Weaknesses of Machine Learning
While machine learning has been a breakthrough in AI technology, it still has its weaknesses. One major weakness is that it requires human assistance to learn. Without human inputs, AI cannot improve the accuracy of its predictions and judgments. For example, if you want to build an AI that distinguishes between a photo of a dog and a cat. In this case, machine learning is required to set up the desired features, such as, “To distinguish between a dog and a cat, focus on features such as eye shape, nose shape, outline, and color.” In this way, a human touch was still required to select key features and elements. This means that machine learning still isn’t strong enough to serve in place of the human brain.
With the rise of deep learning development, a new AI boom began in the 2010s. During this time, structures called “neural networks” were created. These networks mimic human cranial nerves, which no longer Al need human assistance to learn.
Circling back to the previous photo example, this means that AI now can discover the features to differentiate cats and dogs in photos by itself. Along with the rapid evolution of AI development, deep learning has also benefited from dramatic improvements in computer processing power. Since AI is running programs that aim to mimic the human brain, it needs powerful computers to handle the processing loads, eventually the development of hardware has played an important role in the evolution of AI.
In the next blog, I will introduce several difficulties to use deep learning technology in business practice despite its breakthrough technology. For your updated knowledge and insight about AI technology, subscribe to our newsletter or visit HACARUS website https://hacarus.com.