The reason why AI research, which is an extension of the algorithmic framework, cannot achieve true artificial intelligence is due to the lack of 'affordance'

Research into artificial intelligence (AI) is progressing rapidly, with models now appearing that can draw pictures and write text with human-like accuracy. However, while these models are called 'AI,' in reality, they are simply algorithms based on data sets provided by humans in advance and then processed accordingly. They are not truly AI with independent intelligence. A paper has been published stating that 'affordances' are necessary to create this true AI.

Frontiers | How Organisms Come to Know the World: Fundamental Limits on Artificial General Intelligence
https://doi.org/10.3389/fevo.2021.806283

Research into AI has made incredible progress since Alan Turing posed the question 'Can machines think?' in his 1950 paper ' Computing Machinery and Intelligence .' Today, machines can draw pictures automatically and converse with the same level of accuracy as humans, but these are merely the results of algorithms that are limited to a narrow range of tasks, and do not represent the birth of true intelligence.

To address a variety of issues with a single AI model, we need to explore 'artificial general intelligence (AGI),' a computational system that can connect, integrate, and coordinate multiple functions. AGI can analyze, create, and act on its own, demonstrating the distinctive characteristics of intelligence.

AGI must autonomously set goals and improve situations without human intervention, rationally choose paths based on the context and task, and select valuable tasks and relevant contexts from a wide range of options. However, unlike the digital world, even when trying to solve tasks at a higher level, the real world is filled with incomplete, ambiguous, or contradictory information, and even humans find it confusing on a daily basis.

In this paper, the research team states that 'affordance' is important in AGI research. Affordance comes from the word meaning 'to give or provide,' and refers to the meaning that the environment gives to an animal. For example, when we see an empty chair in a plaza, we recognize that 'that is something to sit on.' This recognition is affordance.

If you are in a dark room where you can't see anything, your first reaction will be to 'bump into furniture, accidentally find a light switch, light up the room, and see where you were.' This act of 'using the switch you accidentally found to light up the room and see where you were' is not deduction or induction, but 'insight.' When a mathematician solves a difficult problem through a kind of inspiration, this is also considered insight.

This insight is the ability of humans to deal with ambiguity and contradictions, and is also the source of creativity. What is required of AGI is the ability to derive problem-solving from interactions with the surrounding environment, and this insight requires affordances, the research team states.

For example, when an AI-equipped robot picks up an object, it is virtually impossible to list all possible uses for it in advance and have it learn how to use it; it cannot be handled algorithmically. In other words, only living organisms (humans) can interpret and apply the meaning of a given object. The research team argues that the current framework for algorithmic research in AI and robotics is incapable of identifying and utilizing affordances, making it impossible to develop AGI.

The research team states that biological behavior that can apply affordances is necessary for AGI, and that this discussion will have a diverse impact not only on future AI research and evolutionary theory, but also on the philosophy of science.