AGI technology

Artificial Intelligence (AI) has grown rapidly since 2010. AI systems have been developed for a variety of applications, such as image recognition, speech recognition, big data analysis, natural language processing, gaming, etc., and a lot of AI applications work well.

Basically, AI is categorized into three forms: narrow (or weak) AI, artificial general intelligence (AGI), and artificial superintelligence (ASI).

An individual narrow AI system performs a specific task or solve a specific problem that requires some intelligences, but it does not understand things beyond the scope of the specific task or problem.
An individual AGI system can perform a wide range of intellectual tasks, moreover, it can deal with unfamiliar situations or solve unfamiliar problems by employing learnt knowledge or skills. AGI is capable of thinking abstractly and reasoning, and can achieve human-level intelligence.
ASI is a form of artificial intelligence that significantly exceeds human intelligence in most areas. Now we are still very far away from the realization of ASI.

So far, all developed AI applications, such as face recognition, chatbot, autonomous vehicle, etc., are designed with narrow AI technology. Although narrow AI technology advances very fast in recent years, there are some limitations on it:

Firstly, a huge number of training samples are needed to train a narrow AI system for performing a specific task, furthermore, the number of required training samples increases exponentially with the complexity of narrow AI system. As the complexity of narrow AI system increases, training samples will become severely insufficient, and the training time will become incredibly long. For example, some experts estimate that autonomous vehicle needs 11 billion miles of driving to perform just 20 percent better than human, and this translates to more than 500 years of nonstop driving in the real world with a fleet of 100 cars.
Secondly, narrow AI models are quite different from structures of human brain, and consequently it is most unlikely for narrow AI system to analyse problem and reason like human. Owing to the lack of the ability of abstract thinking, a narrow AI system hardly can handle unfamiliar situations. For example, an autonomous vehicle developed in US may not work in other countries because of different road conditions and people’s driving styles.

However, human does not have such problems as narrow AI. Human needs much fewer training samples to learn a skill than AI. Additionally, human can perform abstract thinking and reasoning after a few years of education. Normally, it takes an adult less than a year to master skills of driving car, and an experienced driver can drive in different countries without much difficulty.

AGI is very different from narrow AI though they have similar names. The basic idea of AGI is to model structures of human brain and thereby achieve human-level intelligence. AGI can overcome limitations on narrow AI and has great potentials. AGI has many advantages over narrow AI:

The training of AGI system requires much less training samples than narrow AI system.
An individual AGI system can perform multiple tasks, but an individual narrow AI system usually performs single specific task.
AGI system can perform abstract thinking, thereby it can apply learnt knowledge and skills to solve unfamiliar problems or handle unfamiliar situations.
AGI has greater potentials than narrow AI, and it will reach a much higher intelligence level than narrow AI.

Until now, the development of AGI is still in very early stage because only little is known about how human brain works. However, many research achievements on neuroscience provide us insights into human brain’s neural structures, in addition, we can analyse our own thinking processes by recalling memories in our own brains. These two ways can lead us to discover principles of human brain.

ByALGO has been working on the research of AGI since 2015. By relentless in-depth study of neuroscience and detailed analyses on our own brain activities, we have made remarkable breakthroughs that enable us to model the cognitive system of human, involving brain, emotion, consciousness, etc. Our long-term goal is to implement AGI systems with human-level intelligence to support applications of robotics, autonomous driving, education, skilled jobs, etc. Certainly, we will carefully control the safety of AGI technology and make sure that it will not bring about risks to human being.

Interested investors are welcome to contact us via email: tech@byalgo.com.