Artificial General Intelligence (AGI): The next stage of artificial intelligence

In addition to improvements and new applications in artificial intelligence (AI), most people agree that the next leap in artificial intelligence will occur when artificial general intelligence (AGI) emerges . We define AGI broadly as the hypothetical ability of a machine or computer program to understand or learn any intellectual task that a human can perform. However, there is little consensus on when and how this will be achieved.

One view is that if enough different AI applications can be built, each solving a specific problem, then these applications will eventually grow together into a form of AGI. The problem with this approach is that this so-called “narrow” AI application cannot store information in a general-purpose form. Therefore, other narrow AI applications cannot use this information to expand their breadth. So while it is possible to stitch together applications for language processing and image processing, these applications cannot be integrated in the same way that the human brain integrates hearing and vision.

Other artificial intelligence researchers believe that if a large enough machine learning (ML) system can be built, and given enough computer power, it will spontaneously demonstrate AGI. When we dig deeper into how ML actually works, this will mean having a training set that contains all the situations our hypothetical ML system might encounter. Expert systems attempt to capture domain-specific knowledge, but it has been clearly demonstrated decades ago that it is impossible to create enough cases and example data to overcome the system's underlying lack of understanding.

The problem with both of these approaches is that, at best, they can only create an artificial intelligence that looks intelligent. They still rely on predetermined scripts and millions of training examples. Such AI would still not be able to understand that words and images represent physical things that exist in the physical universe. They still cannot combine information from multiple senses. So while it's possible to combine language and image processing applications, there's still no way to integrate vision, hearing, and direct interaction with the environment as seamlessly as the human brain does.

What is required for AGI to be successful?

To achieve true AGI, researchers must shift their focus from ever-expanding data sets to a more biologically meaningful structure that Contains three basic components of consciousness: an entity-centered internal mental model of the environment; perception of time, which allows for perceiving future outcomes based on current actions; and imagination, which allows for considering multiple potential actions and evaluating and Select its result. In short, AGI must begin to exhibit the same situational and common-sense understanding as humans to experience the world around them.

To achieve this goal, artificial intelligence's computing system must be closer to the biological processes in the human brain, and its algorithms must allow it to build abstract "things" with infinite connections than today's artificial intelligence Intelligence requires huge arrays, training sets, and computer power. Such a unified knowledge base might be integrated with a mobile perception pod containing visual, auditory, motor, and speech modules. Such a pod would allow the entire system to experience rapid sensory feedback with every action it takes, which over time will lead to an end-to-end system that can begin to work as it gets closer to true AGI. Learn, understand and ultimately work better with people.

Even with such a system, the actual emergence of AGI is likely to be gradual rather than overnight, for two main reasons. First, and perhaps most importantly, developing AGI is obviously a very complex and difficult task, requiring significant advances in several different fields, including computer science, neuroscience, and psychology. While this means several years of research and development, involving the contributions of numerous scientists and engineers, the good news is that a lot of research is currently underway. The various components of AGI will emerge as they are studied in numerous fields.

Then, instant gratification may slow the emergence of AGI because many of its capabilities have market value in their own right. Features developed can improve the way Alexa understands, or new vision capabilities can improve self-driving cars, and individual developments are quickly brought to market because they are commercially viable. However, if these more specialized, individually marketable AI systems can be built on a common underlying data structure, they can begin to interact with each other, building a broader context that can truly understand and learn. As these systems become more advanced, they will be able to work together to create more pervasive intelligence.

As these aspects increase, artificial intelligence systems will exhibit more human-like performance in individual areas and develop to superhuman performance as the system is enhanced. But performance cannot be the same in all areas at the same time. This suggests that, at some point, we will approach the threshold of AGI, then equal the threshold, and then exceed the threshold. At some point after this, we will have machines with significantly better intelligence than humans, and people will start to agree that maybe AGI does exist. Ultimately, AGI must be implemented because the market demands it.

The above is the detailed content of Artificial General Intelligence (AGI): The next stage of artificial intelligence. For more information, please follow other related articles on the PHP Chinese website!