The Way Forward with AI-Complete Problems
Sven Groppe, Sarika Jain · New Generation Computing · 2024
In the AI world, we come across the types of intelligence: Artificial Narrow Intelligence (ANI) and Artificial General Intelligence (AGI). ANI is designed and trained to perform specific tasks as programmed and cannot generalize its knowledge beyond what it is programmed for. Examples include self-driving cars, search engines, chatbots, and Virtual Personal Assistants. AGI, on the other hand, aims to perform any intellectual task that a human can. It has the ability to understand, learn, and make decisions. Intelligent agents are “intelligent” software that can work autonomously, seek necessary/present/relevant/authentic information, coordinate with each other, understand the contents, take necessary actions to make life simple for human beings and improve their performance by acquiring knowledge. Consider the simple and straightforward problem of machine translation from language A to language B; it includes tasks such as Natural Language Processing (NLP) in both languages, reasoning, knowledge engineering, contextual understanding, and social intelligence. Task-specific algorithms will not be able to solve the said problem as it involves simultaneous solutions to multiple subproblems. Such problems whose solution is beyond the capabilities of Artificial Narrow Intelligence (ANI) and which require Artificial General Intelligence (AGI) to reach human-level machine performance are termed as “AI-complete.”