Posted on

General AI – the Holy Grail?

General AI is a theoretical form of artificial intelligence (AI) that can solve any task using human-like cognitive abilities. It has been the goal of many AI researchers since the earliest days. It has yet to be realised, and the truth is that all the useful applications have come from a more pragmatic view based upon narrow AI

Unlike Narrow AI, which is designed for specific tasks, General AI seeks to emulate the breadth and depth of human intelligence across many domains. Some of the clearest distinctions are:

  • General AI should be capable of performing any intellectual task that a human being can; Narrow AI is specialised in a specific task or set of tasks.
  • General AI should possess the capacity to learn from experience and apply knowledge to unfamiliar scenarios; Narrow AI is limited by learning from predefined tasks and cannot generalise knowledge.
  • General AI should exhibit adaptability across diverse domains akin to human intelligence; Narrow AI is highly specialized and lacks versatility.

Proponents of General AI point to how far the technology has come and argue that General AI is just around the corner, but they have been saying that since the beginning of the development of AI. In order to create General AI, which more closely resembles human intelligence, developers will have to overcome some fundamental issues:

  • Currently, AI systems have no concept of whether their answers are correct
  • They have no idea when they reach the limits of their expertise
  • They have no sense of the context in which their knowledge exists.

In order to overcome these hurdles, it is unlikely that implementing General AI will be achieved by evolutionary development, but by a complete paradigm shift in the technology.

This image was created by AI

Do you think General AI is likely to be achieved anytime soon?
What benefits do you think General AI would offer?
What threats do you think General AI would pose?
Record your thoughts in your learning diary before moving on…