By: Arya Brijith, International Center for AI and Cyber Security Research and Innovations (CCRI), Asia University, Taiwan, arya.brijithk@gmail.com
Abstract
This article describes Akinator, examining how it captured the internet’s attention and produced considerable hype. The AI algorithm utilized in this game is also explained to debunk claims that arose in 2007. It is also understood that this fascinating game is constantly updating its information, which includes learning from its users.
Keywords
Akinator, AI, mind reader, game, algorithms, decision tree, binary search, learning, model, collaborative filtering, expert setting, technique.
Introduction
Effective methods for building logical inferences have been devised that allow artificial intelligence systems to recognize an unfamiliar subject from its description. In the digital age, where smartphones have become an extension of ourselves, apps have taken on a life of their own. Elokence, a French business, created the video game Akinator which is an AI-powered game. Among them, Akinator, a mind-reading operation, has captured the imagination of millions of people around the world.
During gaming, it asks a series of questions to try to figure out which fictitious or real-life character, item, object, or animal the player is thinking about. It examines the value of questions in verbal exchanges between people as well as the cognitive processes that support them.
The article highlights the evaluating question-asking process using online games. The enigma of how the genie reads your mind has been solved. In actual life, there is no such thing as a mind reader. Using Machine Learning methods and approaches, the game can forecast your character. The game constantly broadens its knowledge base and picks up new skills through its interactions, enabling it to keep up with contemporary fictional and nonfictional characters. [R5]
The game’s high success rate has piqued people’s curiosity, and it is a good case study for learning about the question-asking process. The spotlight then shifts towards the use of machine learning algorithms and techniques that promote question-asking research.
Figure: A still from the game site
About Akinator
In this game, a genie tries to guess your word or character within 20 yes or no questions. Think of a character, celebrity, or any famous personality, the genie can guess it. The technique considers the users’ response to dynamically alter the suggested candidates during the recommendation process. The most efficient tag information will be chosen for users to pick during the interaction.
The game constantly broadens its knowledge base and picks up new skills through its interactions, enabling it to keep up with current fictional and nonfictional characters. In a study researching empathy through face mirroring, other researchers studying human-robot interaction used the Akinator online service as a source for discourse.
At the end of the game, the system shows responses that are like the ones the player was considering and asks them to select the right answer. In this manner, the game learns from its past failures and learns through data gathering.
Algorithms/techniques that may be deployed
- Binary search algorithm- The binary search technique finds a target value’s location inside a sorted array. The center item of the array is compared to the target value. The loop continues until the object is found. The first query is about the root node. This goes on inner and inner. Using the insight that the array is sorted, binary search attempts to minimize the time complexity to O(log N).
Mathematically, if each question can eliminate half the objects, 20 questions are enough to allow the question asker to distinguish between 220=1,048,576220=1,048,576 objects. [R1].
- Decision tree algorithm- In supervised machine learning techniques, a decision tree is a powerful tool that may be applied to both classification and regression problems. It is made by continually splitting up the training data into subsets according to the values of the attribute up until an ending condition is reached, such as the maximum depth of the tree or the absolute minimum number of samples necessary to divide a node. The questions are employed as internal nodes in the Akinator problem, while the characters are used as leaf nodes. This way, the Akinator can conclude after receiving the yes or no answers from the player.
Figure: Decision tree
- Collaborative filtering- It is a machine learning approach that filters content based on previous interactions and data collected on the user’s end. Collaborative filtering applications often require very large data sets. It might be utilized in the Akinator platform for additional features or functionalities such as proposing comparable games or delivering tailored content suggestions.
- Expert Setting-The term “expert setting” often refers to a complex or specialized configuration setting within a piece of software, a computer system, or an application. Users who have a thorough grasp of the program and wish to modify it to suit their requirements or preferences should utilize this setting. Users have access to a broader variety of choices and parameters in an expert setting than they have in the default or standard settings. With the help of these sophisticated choices, the behaviour of the program may be adjusted, optimized, or even substantially changed. Accessing the expert setting calls for a higher level of technical competence and software or system understanding. It is designed for people who understand how things function and are confident in making alterations that might not be advised for normal users.
Similar games
There are more such character-guessing games available such as:
The online psychic- In this game, the player will be given a series of numbers to choose from. The system then generates a page displaying a few of those numbers and asks if the number we are thinking of is displayed on that page and the process continues. Usually, it can display the exact result with 3-4 loops. The game website- https://www.transum.org/Software/Fun_Maths/Online_Psychic.asp
Nao robot – It is not exactly an online game. Nao robot is a 58 cm tall robot created by Aldebaran Robotics, a French robotics firm with its headquarters in Paris. In 2015, SoftBank Group bought the business, and it is now known as SoftBank Robotics. Many British schools have utilized Nao robots to educate students about robots and the robotics industry. Children were questioned about their opinions of the Nao Robot’s skills after playing a five to seven-minute character identification game with it. The replies from conversations with 30 kids, ranging in age from eight to twelve, revealed that when the robot tried to guess the character of the participant, the robot was more likely to answer correctly. This is associated with the guess-a-character game that children play online.
Mind Reader- An online game called Mind Reader challenges players against a machine that tries to predict their next action. The algorithm is built on an online learning architecture that continuously learns from user activity. The method employed in this case is based on the Expert Setting. Several strategies (or predictors) are used to forecast the user’s next action, and a meta-algorithm then combines all of the predictions to produce a single conclusion. In this game, depending on your turn, you can use the keyboard or the buttons below to select left or right. The machine tries to anticipate your decision. If the computer correctly guesses your decision, it wins the round; if not, you win. The first to achieve 25 victories, wins. Each round lasts for three seconds, the additional time goes up to ten seconds and will be carried over to the following round. The game website- https://web.media.mit.edu/~guysatat/MindReader/index.html
Conclusion
Akinator is a fascinating illustration of how multiple machine-learning algorithms might be utilized for interactive entertainment, to sum up. The game’s ability to correctly anticipate personalities, objects, and animals within 20 questions is amazing to observe. It makes use of a range of complex methods, including collaborative filtering, binary search, and decision trees.
It is acknowledged that the accuracy is dependent on the enormous training data set and that the effectiveness depends on the quantity and caliber of the questions accessible. The genie may anticipate the character in the user’s head by utilizing the training data and the specified methods.
But the puzzle of how this internet genie works is still a mystery. Although we may have a general understanding of how some algorithms operate, the techniques used in this game remain unknown. The datasets used to train the model are by far the most valuable component of AI technology, since the richer the datasets, the greater the model’s accuracy, even though we may not completely grasp the techniques employed by the genie in Akinator.
References
- Sasson, G., & Kenett, Y. N. (2023). A mirror to human question asking: Analyzing the akinator online question game. Big Data and Cognitive Computing, 7(1), 26.
- Xie, Q., Xiong, F., Han, T. et al. Interactive resource recommendation algorithm based on tag information. World Wide Web 21, 1655–1673 (2018). https://doi.org/10.1007/s11280-018-0532-y
- Suresh, S. R. (2017). A bayesian strategy to the 20 question game with applications to recommender systems (Doctoral dissertation, Duke University).
- Yung, I., D’ambrosio, F., Zaccaria, A., & Floreani, F. (2022). A Bayesian based Intelligent Troubleshooting System. Procedia Computer Science, 200, 602-610.
- Henkel, Z., Bethel, C. L., Kelly, J., Jones, A., Stives, K., Buchanan, Z., … & Pilkinton, M. (2017, August). He can read your mind: perceptions of a character-guessing robot. In 2017 26th IEEE International Symposium on Robot and Human Interactive Communication (RO-MAN) (pp. 242-247). IEEE.
Cite As
Brijith A. (2023) Akinator:A Mind Reader Taking Over the Internet, Insights2Techinfo, pp.1