Hyper-Intelligence
for human beings

Development of Artificial General Intelligence
and Strong Artificial Intelligence Algorithms.

ABOUT US

‘HYNTEL’ is an abbreviation for HYPER-Intelligence.
Hyntel Inc. was established in January 2022 by Dr. Kwon Daesuk, the founder of Clunix, a Korean venture company, with the goal of developing algorithm of artificial general intelligence or strong artificial intelligence.
Hyntel Inc. seeks to realize the Universal, Artificial General Intelligence or Strong Artificial Intelligence that learns and judges like a human by identifying the essence of human intelligence at the algorithmic level.

MISSION

Hyper-Intelligence for
human beings"

Hyntel Inc.'s top priority is to contribute to the survival and co-prosperity of human beings
through universal artificial intelligence,
and through this, we aim to achieve economic benefits
for shareholders and employees at a reasonable level.
From everyday problems of each individual to national or global problems such as global warming, economic crises,
the poor and the rich,
and political and ideological confrontations, Hyntel will find realizable and sustainable solutions
through artificial intelligence having human-level or higher capabilities.
Hyntel Inc. provides and utilizes development results for the sake of mankind or
for personal moral and respectable purposes, but does NOT provide them for free in any case.

TEAM

CEO

Kwon Daeseok

CEO

Kwon Daesuk

Born in Seoul in 1969, influenced by the cosmopolitanism and philanthropy of astrophysicist Carl Sagan in 1980, started programming with Apple-II computers.
In 1982, the second year of middle school, made a lifelong goal ‘the development of universal artificial intelligence’, and since 1985 have studied Symbolic AI using Lisp, Prolog, Forth, etc., but in 1987, started studying artificial neural networks realizing human intelligence could not be implemented in that way. In 1988, after entering the Department of Computer Science and Statistics of Seoul National University, first studied backpropagation learning of artificial neural networks and gave up implementing artificial intelligence by artificial neural networks.
A new idea, 'Active Network Theory', was created in 1990 by researching algorithms that enhance the strengths of artificial neural networks and symbolic AI and overcome the shortcomings. He went on to a parallel processing lab of Seoul National University, and earned his PhD degere in 1999 developing a cluster supercomputer.
Founded Clunix, inc., a supercomputing company, in January 2000 and managed it for 22 years.
As the management of Clunix Inc. is on a stable track, founded Hyntel Inc. in January 2022 after seeing that the algorithm created in 1991 works well on the latest computer and produces good results.

CTO

Paeng Won-Gi

CTO

Paeng Won-Gi

Born in Masan in 1985, entered the Department of Physics at Hanyang University in 2003.
In 2013, received PhD in Particle Physics from the Department of Physics at Hanyang University.
Since March 2014 to February 2018, worked in Rare Isotope Science Project in Institute for Basic Science(IBS) in Korea for the topic related to nuclei and dense nuclear matter.
Related to elementary particles, nucleon interaction, nuclear matter, and neutron star, published 12 SCI papers.
In April 2018, joined Clunix Inc. and started study and research for artificial intelligence.
In 2022, founded Hyntel Inc. with Dr. Kwon.

ALGORITHM

Hypothesis of General Intelligence
& Self-organizing Active Nework of Concepts(SANC) Model

Hyntel's artificial intelligence research is based on the 'Hypothesis of General Intelligence' of founder,
Dr. Kwon, and is an attempt to prove that hypothesis. Some of its contents are as follows:

1. Intelligence operates with the same basic algorithms in any field and at any level.

2. Intelligence is a black box that ‘remembers previous events' and 'predicts the next event’ efficiently and accurately.

3. All intelligent activities can be reproduced with very few simple principles, including Hebbian learning.

CEO Dr. Kwon established a model called
‘Self-organizing Active Network of Concepts’ (SANC) that conforms to this hypothesis
and confirmed its working potential at the laboratory test in collaboration with Dr. Paeng.
Compared to the existing deep learning-based artificial neural network, the ANOC model has the following characteristics.

Universality and Generality

- The same algorithm is used not only for natural language processing, but for error detection, image recognition, chatbot implementation, and handwriting recognition.

- The same algorithm is used from low level such as morphological analysis or outline detection of an image to high level such as semantic analysis, Q&A, and picture creation.

- As a result, our algorithm can interpret the language of whales or aliens, predict economic crises or tomorrow's weather in theory.
Unsupervised Learning

- Learning only with raw data and unprocessed corpus without vocabulary dictionary, grammar definition, or tagged training set.
Only Small Training Set required without cleansing data

- Can be trained with training sets of hundreds or thousands of items, less than 1/100 of that of deep learning, and does not require large training sets of tens or hundreds of millions of items.

- Tagged training set is unnecessary, and it learns well with training set including noisy data or wrong data.
Fast learning time, low power consumption/computing power requirement

- Learn in minutes with a single CPU.
No need to train for months with hundreds or thousands of GPUs.
No seperation of learning phase and application phase

- Learning and utilization are not separated in Active Network of Concepts.
As a result, analysis, prediction, and evaluation occur during learning, and learning occurs during utilization.
Explicit and directive learning
and modification possible

- Specific knowledge can be added to the trained network,
or trained knowledge can be easily corrected.
Traceability

- It is possible to trace and visualize the process how the network derived a result.