Hello everyone, My name is Kenshin Fujiwara and I am the CEO and founder of HACARUS Inc.
Through 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.
Today I will explore how to implement AI into the existing system and discuss some barriers that prevent the smooth progress of AI applications in the business field. But, before jumping into the body, I want to take some time to summarize the relationship between AI, machine learning, and deep learning. AI is a general concept that is very broad. Within this concept, machine learning is one of the computational processing methods used. Zooming in a little further, deep learning can be thought of as an advanced form of machine learning.
Problems with AI Implementation
In the world of AI, there are several problems with the implementation process and AI project management that hinder its performance. So what will it take to ensure a smooth transition from a traditional system to an AI-based one? First of all, it is important to clarify the goals of your organization and what problem you are trying to solve using AI.
For companies that try implementing AI simply because it is trendy, I think that it is unlikely to create any value within their business. Even in cases where a mid-term management plan is put in place that calls for the creation of “high value-added business through the introduction of AI,” it often leads to failure. A common mistake is that the use of AI becomes the main objective and thus ends up being the wrong way of approaching the problem.
To help prevent these mistakes it is important to decide what you want to accomplish with AI and what your objectives are before embarking on a project. As long as the objectives are clear, the hurdles to implementing AI can be highlighted and addressed. It is also helpful to know about some of the common hurdles that many companies face as they attempt to implement AI within their business.
Barriers Preventing AI Application
Therefore, I would like to discuss the five common barriers that prevent the smooth progress of AI projects today. Earlier to me, Kaoru Kawamoto, Professor of Data Science at Shiga University, pointed out four barriers in his book, “The Power of Analysis to Change a Company” (Kodansha’s new library of knowledge), which must be overcome in order to utilize data and apply it successfully to business.
These four barriers are:
- The Data Barrier
- The Analysis Barrier
- The KKD Barrier
- The Cost-Effectiveness Barrier
Applying these barriers to the world of AI, I found that there is an additional fifth barrier, the human resource barrier. Therefore, I would like to further explain these five barriers based on my experience and findings. At first, I will explain about the first two barriers, the data barrier and the analysis barrier today, then I will continue to discuss the others in the following posts.
The Data Barrier
While a lack of data is always a problem, even when there is a significant amount of data, it is often in insufficient quantity or poor quality. In this situation, the AI will not be able to produce the expected output. In other words, in addition to quantity, the quality of the data is also important for ensuring a certain degree of accuracy. Similar to a human, AI also cannot make an accurate decision without a certain amount of information. Even though the amount and quality of data will depend on the application, there is still a set amount of data that is required for each case.
Another problem with data is the issue of bias. Looking back once again at the cat and dog photo example, the balance of training photos is extremely important. If, for example, there is a large number of dog photos, but only a few cat photos, the accuracy of the model will be low.
Furthermore, the training data could also be missing critical samples, or it could contain outliers that render it unusable. For this reason, the quantity and quality of the data required need to be guaranteed when building the model. Even when obtaining a data set that is adequate in both quantity and quality, there are still several factors to be aware of. There are many cases where the accuracy of the analysis cannot be ensured due to errors in the analysis method itself.
The Analysis Barrier
The second barrier is the inability to gather data with sufficient accuracy for use in the field. For example, let us look at product inspections. When scanning for defective products, it is important to make zero mistakes. Putting it into numbers, this task requires 100% accuracy. Therefore, it is necessary to make efforts to improve the accuracy that can be achieved with AI technology. At the same time, we also need to consider how to overcome areas that cannot be addressed by AI. This can be done by analyzing and improving the systems design and various business processes.
It is also important to mention that advances in AI technology are constantly evolving. Each year, AI technology is improved to the point that analytical accuracy that was deemed unachievable the previous year is now possible. To overcome the analysis barrier, it is important to stay up to date with the current AI trends. By staying informed, you can have a better idea of what AI projects to focus on now, and which ones might be more feasible in the future.
In the next blog, I will continue to discuss other barriers that prevent the smooth progress of AI applications in the existing system. For your updated knowledge and insight about AI technology, subscribe to our newsletter or visit HACARUS website https://hacarus.com.
