Skip to main content

Monitor and controler relationship between humans and machine

by content 12.12.2022

Human-in-the-loop concept

Human-in-the-loop is a term in the field of engineering and computer science, widely used in the interactive simulation of human and robot activities. In recent years, the term has been widely used in Artificial Intelligence (AI), where machines and humans are in an interrelationship, contributing to improving overall results and speeding up machine training. In other words, this is the process of combining machine and human intelligence to get the best results, improving performance in the long run.

The purpose of Human-in-the-loop (HITL) is to optimize models and algorithms through human intervention and evaluation, in order to create a better and more accurate Machine Learning model.

Human-machine collaboration: Increasing AI accuracy

Previous industrial revolutions liberated people from hard labor, but today's most advanced technologies such as Artificial Intelligence (AI), RPA, etc., are developed to replace humans in repetitive, time-consuming tasks that need many human resources.  However, it is a fact that AI lacks the concept of absolute certainty like humans, so it cannot make predictions and decisions with 100% reliability.

HITL is a useful tool that helps humans directly intervene in the decision-making steps in the automatic operation of machines, in order to improve the most reliable accuracy rate. The machine can continue to adjust the model based on the results calibrated by the human.

In the process of building and developing FPT.AI's Artificial Intelligence-powered solutions, HITL is integrated into many different stages of the product/solution life cycle, from the initial machine training to the implementation stage for customers.


Training and testing phase: HITL is commonly used in the training and testing phase of AI models to train models more accurately, quickly, and efficiently.

Implementation phase: HITL is also an indispensable tool in the implementation phase. Because in the deployment phase, there are many situations arising that the machine is not prepared to handle. The model will therefore require human intervention so that humans can verify the AI model's predictions and send that feedback to replace the AI-generated prediction. These scenarios are used to retrain and refine the model in the future.

Even when the model is highly accurate, human supervision and frequent thorough checks are required to ensure optimal system operation.

Practical applications of HITL

HITL is capable of independent deployment or integration into FPT.AI's Artificial Intelligence-powered solutions, in order to provide customers with the most optimal solutions for the operation process.

Some HITL models are integrated and deployed by FPT.AI for customers such as:

  • Humans rate 100% of AI results
  • People evaluate according to the criteria of businesses
  • Humans rate AI results by warning threshold
  • Humans randomly evaluate AI results

In fact, FPT.AI has integrated HITL into many steps in the training process and trained machine learning models for FPT.AI solutions. A typical case is the data labeling process when teaching the machine to read and extract information fields on identification documents. When the machine cannot process rare and special data such as foreign names, names of ethnic minorities, transliterated places, etc., humans will intervene to censor the information and teach the machine to understand the information that particular data. Thanks to HITL, FPT.AI's solution for extracting and digitizing data on all kinds of papers can achieve an accuracy rate of up to 98% - the highest in the market today.

In addition, FPT.AI also developed HITL as a separate solution to integrate it into the operating process of businesses, in order to help businesses solve many practical problems such as processing and reviewing documents of drivers on ride-hailing apps; handling records and papers of the banking, insurance industry, etc. Thanks to the smooth coordination between people and machines, the operating system and operating process are more optimized, helping businesses save operating costs and human resources.

At least until now, technology has not been able to completely replace humans, but humans can use machines to improve the quality of their work and optimize them. Therefore, HITL is a useful tool to help people focus on improving the accuracy of machines so that in the near future machines would be smart enough to do that job instead of human.  

  • Learn more about FPT.AI's solutions and book a consultation:
  •  Hotline: 1900 638399
  •  Email: [email protected]