Unique features of FRONTEO's proprietary AI engine "KIBIT"
FRONTEO's in-house developed AI, "KIBIT," which specializes in deriving discoveries from massive amounts of data, has unique features that set it apart from generative AI such as LLMs (large-scale language models) and LLM-based ChatGPT.
Four unique features of AI "KIBIT"
Vectorization of unique technology
Unique algorithms based on numerous patents
Intuitive data visualization
Energy-saving Green Micro AI
Point 1
Vectorization of unique technology
At KIBIT, we use distribution hypotheses to convert words into numerical values, a process called "vectorization," which is essential for analyzing natural language (human language) using AI.*1One of our major features is that we develop our own unique technology in strict adherence to the principles of our
Technically speaking, this is a method to "capture the relevance of words and sentences based on the co-occurrence of words." It has been shown that vectorization approaches based on distributional hypotheses produce better results than Transformer, which is widely used in generative AI.*2.
*1 An important concept in natural language processing, which states that meaning is formed by surrounding words and context rather than by the word itself.
*2〈Yamada et al.(2020)>
Point 2
Unique algorithms based on numerous patents
Transformer is widely used in generative AI to perform advanced analysis of documents (natural language) and networks.* We have developed our own proprietary algorithms that are different from those used in other software. By making full use of our more than 70 in-house patents related to algorithms, data analysis and visualization methods, such as analyzing word patterns to capture context, appropriately weighting words, and optimizing features, we can output optimal analysis results with high accuracy.
We even provide our own in-house developed software equipped with AI, which can be implemented while customizing and developing additional software to suit a company's data and challenges.
* Transformer: A deep learning model and a natural language processing method announced in 2017.
Point 3
Intuitively visualize and map your data
The results of data analysis with KIBIT are not simply output; we also place importance on "visualizing" them, in other words, making them visible in order to further enhance their value.
Properly processed data* By plotting this data on a flat surface to create a "2D map" that people can visually recognize, it is possible to see the interactions and relationships between pieces of information through expressions such as color and scattering, which can lead to "serendipity" in the discovery of unexpected information that would be overlooked through simple analysis.
* Dimension compression or reduction: Replacing the original huge amount of text information with feature vectors, which are sets of many numerical values, and compressing the data down to two-dimensional latitude and longitude information while preserving the relationships between the vectors.
Point 4
Energy-saving Green Micro AI
Generative AI, which is gaining attention from all quarters, actually consumes a large amount of electricity and the associated CO2Issues such as emissions and water consumption have been pointed out.
On the other hand, KIBIT is a power-saving and environmentally friendly (=Green) AI with a simple algorithm with minimal parameters (=Micro). Its strength is that it can analyze and learn data at high speed with extremely low power consumption, at the CPU level of a normal PC, rather than a GPU at the level of a data center.
