AI’s quality depends on our understanding of its strengths and weaknesses, and whether data is available to it or not.

Mio Analytics has reported that the mainstream of developing an artificial intelligence (AI) has recently been around machine learning, which refers to the process in which AI can learn and develop itself from the data fed. Without a preceding trove of data, AI cannot be utilized, for example, at the beginning of transforming raw data into digital data, AI cannot identify which data gathered is true and which one is false, thus easily resulting in a faulty work. 

Nowadays, there is quite a bit news reporting on AI’s victory over humans such as in chess competitions, Go tournaments, quizzes, and image recognitions. In these cases, AI’s victory is the result of the countless calculations of a particular scenario in searching for the winning solution and that AI is optimized with data. As such it is not surprising that AI can beat humans.

However, AI’s work process is to learn from data, in which it can make a mistake like a human does. Its mistakes are different type from those of human, though, because it has a different thinking approach. Should a mistake happen, it is hard to identify the cause. Unlike human thinking approach, AI is not capable of knowing what is “normal” in data being fed. 

Therefore, when AI learns from a trove of data containing faulty or improbable data, it will treat such input as “normal” as we have seen in many cases of this type of AI mistake.  

“Intelligent” Utilization

Using AI appropriately to its strengths and weaknesses has become an “intelligent” utilization. In the case of Mio Analytics, for example, an AI system is used in automated document referencing to accept or reject a research by comparing a newly presented research with the ones that have been previously certified. If it shows that the new research does not cohere with the previous ones, it will reject the research. The whole process takes only a few seconds and it can work around the clock under the same criteria. 

However, the system works well only to screen a new research resembling with the previously certified researches. Unlike human screening, it does not recognize a new research when its results come from the outside-the-box framework. Therefore the system is used merely as a part of research screening at Mio Analytics. 

The time saved by AI screening is well spent on comparing and interviewing the applicants of the researches that AI could not do the job as well as of those that have passed the AI screening to make sure that they should really be certified. The whole process is achieved based on the human judgement that is different from that of AI after all.

Human Substitution

The strongest point of AI is its capability of instantly gathering, analyzing, differentiating the data and correcting a mistake if any. Human can be replaced by AI for those functions. However, it does not mean that AI surpasses humans but rather that the artificial intelligence can only be optimized with human intelligence.