When people hear the words ” artificial intelligence ” and “AI” for the first time, they probably imagine intelligent robots like “Doraemon” and “Astro Boy”.
The robot, which behaves like a human being, is supported by the anime culture that Japan is proud of, and is famous all over the world.
However, robots with intelligent behavior such as “Doraemon” and “Astro Boy” are unfortunately not currently being developed.
The reason why such a robot has not been developed is because there is an important problem called ” frame problem “.
So this time, I will introduce the “frame problem”, which is an important problem in artificial intelligence. If you read this, you may understand in detail why “Doraemon” was not born.
Table of Contents
- What is the frame problem?
- What are bombs and robots?
- Frame problem in tacit knowledge
- How to solve the frame problem?
- Computational complexity problem
- cognitive wheel
- tacit knowledge calculator
- at the end
What is the frame problem?
The definition of “ artificial intelligence (AI)” varies, but broadly speaking,
Technology that allows computers to perform intellectual actions such as language comprehension, reasoning, and problem solving instead of humans [1]
can be said.
Autonomous driving and chatbots are easy to imagine.
This AI is mainly classified into ” weak AI ” and ” strong AI “.
|
Almost all AI currently in widespread use falls under the category of “ weak AI ”. The argument that led to this classification is the “frame problem”.
The frame problem was proposed in 1969, during the first AI boom, by computer scientist John McCarthy, who coined the term “AI”, and cognitive scientist Patrick Hayes.
What is the “frame problem”?
Robots with limited information processing capacity cannot deal with all problems that can occur in reality [2]
That’s the problem.
To understand the meaning of this problem, it is necessary to use words such as “What is finite information processing capacity? “, ” How does a robot behave ? ” It will be easier to understand if you break it down.
“Bombs and robots” are easy to understand by focusing on these three disassembled elements .
From here, I will introduce what “bombs and robots” are all about.
What are bombs and robots?
The “bombs and robots” thought experiment developed by Daniel Dennett in his paper “Cognitive Wheel: Artificial Intelligence and the Frame Problem” is a very famous example of the frame problem in an easy-to-understand manner.
The characters are “Dr.” and “Robot invented by Dr. (R1)”. The purpose of this thought experiment is until the robot takes its own battery out of the warehouse.
R1 robot
The doctor gives instructions to R1 Robo.
Doctor: “The battery is going to run out soon, so bring it from the warehouse. However, let’s assume that there is a time bomb on top of the battery.”
R1 Robo took out the battery from the warehouse, but brought a time bomb along with the battery. And then an explosion.
The doctor challenges the development of “R1-D1” so as not to repeat the same mistake again.
R1-D1 robot
Dr. added a new function to the developed “R1-D1”. It is to “compute all that is possible with the intended implications”.
To explain in detail, in addition to the original purpose of “removing the battery from the warehouse”, for example, “What if the time bomb is on top?”, “What if the battery and the bomb are on the same wagon?” Etc., calculate everything that can happen in the process of achieving the purpose, and deal with each.
That way, even if the bomb is on top of the battery, the idea is to be able to dispose of it.
The doctor gives the robot the same instructions as last time. The robot goes to the warehouse and tries to take the battery out of the warehouse by itself, but there are too many problems to deal with and it explodes. Since the problems that could occur as secondary problems were infinite, the amount of calculation became enormous, and as a result, nothing could be done, and the time bomb reached the end of time.
The doctor does not give up and challenges the development of “R2-D1”.
R2-D1 robot
Dr. Judging that the previous problem was due to trying to deal with “problems that didn’t need to be dealt with”, he added more new features.
It is to “distinguish between relevant and irrelevant connotations”.
In other words, distinguish between the original purpose of “removing the battery from the warehouse” and its irrelevant implications (“What is the color of the walls?”, “What is within a 2m radius of the battery?”, etc.). I tried to ignore the irrelevant and accomplish my purpose. Let’s call this robot “R2-D1”.
The doctor gives the same instructions.
Doctor: “The battery is going to run out soon, so bring it from the warehouse. However, let’s assume that there is a time bomb on top of the battery.”
Then R2-D1 Robo stands still in front of the warehouse and explodes.
It’s possible that we kept calculating what we should ignore, conditions that had irrelevant implications, and ended up doing nothing.
In this way, this thought experiment expresses that no matter how flexible a robot can be made, it cannot deal with all problems that can occur in reality.
Now that we’ve read the “bombs and robots” example, let’s look at the frame problem from a different perspective.
Frame problem in tacit knowledge
Japanese computer engineer Hitoshi Matsubara takes an approach that combines tacit knowledge with the frame problem [3]. Before that, let’s talk about “implicit knowledge”.
In fact, the robots in “Bombs and Robots” have their names.
The “R” in “R1-D1” stands for “Robot”, but the “D” stands for “Deduce”.
It means that the robot in this thought experiment dealt with the problem a priori, that is, from the beginning, it was a completely blank slate with no empirical rules.
The opposite of “deduction” is “induction”. In other words, it means thinking based on various rules of thumb. Based on these, let’s think about connecting “tacit knowledge” and the frame problem.
First of all, why do people act when faced with unknown problems?
For example, let’s replace the robot part in “Bombs and Robots” with a human. Humans would probably rush to dodge the time bomb and take the battery out of the warehouse instead of dealing with the extra stuff.
However, it is possible to do such a thing because the empirical knowledge of life is naturally gathered in the mind, and it is possible to deal with various problems by using that knowledge. It’s from
This knowledge is called “tacit knowledge”. The presence or absence of “tacit knowledge” is a very important concept in distinguishing between humans and computers.
In addition, it is believed that there is a vast amount of “tacit knowledge” that is a collection of knowledge inscribed by genes and empirical rules that are naturally accumulated from everyday experiences.
Based on the affirmation of the existence of tacit knowledge inherent in humans, this is the idea that humans solve the frame problem in a pseudo manner because humans deal with all kinds of problems inductively.
I personally find it very interesting to think about the frame problem in terms of the relative view of humans being inductive and robots being deductive.
Now that we’ve taken a closer look at the frame problem, let’s see what this problem represents in the real world. Let’s consider the frame problem more closely.
How to solve the frame problem?
From here, I will introduce three elements that are considered necessary to solve the frame problem.
- Computational complexity problem
- cognitive wheel
- tacit knowledge calculator
I will explain each in detail.
Computational complexity problem
John McCarthy and Patrick Hayes, who first proposed the “frame problem”, said that the essence of the problem is the amount of computation.
According to the original definition, “When trying to describe (program) a certain action logically on a computer (that is, a computer), it is troublesome to describe events that do not change one by one.”
Therefore, the solution approach needs to consider and reduce “description + processing”.
Taking the blackboard as an example, it is good to write as little as possible on the blackboard as small as possible. In other words, it is desirable to solve both the versatility and computational complexity problems cited in computer science.
As mentioned above, although the processing has been solved to some extent, it is still said that there are many descriptions. GPT-3, which is probably the closest thing to a tacit knowledge calculator, is also famous for having many parameters.
cognitive wheel
The “cognitive wheel” in the title of the paper refers to one of the historical inventions, the “invention of the wheel.”
The invention of the wheel is considered to be the oldest and most important invention in human history, and it is a well-known story that the Ubaid dynasty of the Mesopotamian civilization was used as a pulley potter’s wheel in the 5th millennium BC.
It is believed that mankind was able to invent the wheel not because it was modeled after other creatures.
In the same way, Daniel Dennett alludes to a direction in which artificial intelligence is not modeled after humans, but is viewed as an independent computer. From there, the discussion became lively about ” weak AI ” and ” strong AI ” introduced at the beginning .
tacit knowledge calculator
In order to realize artificial intelligence that can deal with any problem, that is, solve the “frame problem”, the focus is on whether or not tacit knowledge can be expressed on a computer, that is, artificial intelligence that can perform processing based on empirical rules on a computer. will be
So far, the process of how humans select appropriate tacit knowledge and solve the frame problem in a pseudo manner is still unclear.
Therefore, the current situation is that tacit knowledge cannot be expressed by computers, and as long as that is the case, it will be difficult to develop robots that can break through the frame problem.
However, since deep learning research is currently being actively conducted, it can be said that the narrowing down to achieve the purpose has been improved to some extent.
The example of AI defeating the world Go champion is a concrete example of this.
at the end
It can be said that all living things face the frame problem in their daily lives.
For example, “What should I do for dinner today?”, “How should I write a report for college?”, “What do I want to do in the future?”
At such times, most people solve their problems by gathering information that already exists, such as looking at recipes posted on the internet, gathering the content they want to write from reference materials, and choosing the course taken by the adults around them. You know.
That doesn’t solve the problem in terms of the “frame problem”. Because you are just solving the problem within the frame.
However, I think that everyone has solved the problem by making an empirical judgment on the spur of the moment. Therefore, it can be said that all living things solve the frame problem or not.
When thinking about how to solve the problem in front of us, it might be good to measure the presence or absence of frames.
