Computing for the definition of intelligence

For a good time, I thought would use distant reading computing techniques to define the word "intelligence", specifically, "artificial intelligence".

I began by searching arxiv for pre-prints contgaining the phrase "artificial intelligence". This resulted in approximately 15,000 records. Using the records' abstracts as proxies for the articles themselves, I created a Distant Reader study carrel ("data set") from the citations. To give you an idea of the carrel's size and scope, the data set is 2.6 million words long (where the Moby Dick has about 0.25 million words long), and a word cloud illustrating the data set's keywords follows:

Wordcloud illustrating the study carrel's scope

I then extracted all of the sentences from all of the abstracts. I vectorized ("indexed") the sentences, and stored the result in a database. I then searched the index for the word "intelligence" and specified the index to return the single closest sentence from the result. Apparently the closest sentence comes from an article called "The relationship between Biological and Artificial Intelligence" by George Cevora (2019 - https://arxiv.org/abs/1905.00547):

Intelligence can be defined as a predominantly human ability to accomplish tasks that are generally hard for computers and animals.

Not a great definition because it defines things in the negative, but a good start, if you ask me.

In order to garner a more well-rounded definition, I expanded my query to return 20 sentences, and I got the following result:

This paper is a survey of a large number of informal definitions of ``intelligence'' that the authors have collected over the years. Finally, we survey the many other tests and definitions of intelligence that have been proposed for machines. Independent from the still ongoing research in measuring individual intelligence, we anticipate and provide a framework for measuring collective intelligence. Our understanding of intelligence is directed primarily at the human level. In this regard, intelligence becomes a product of information processing. In such a case, intelligence arises as a result of information processing. The discussion at hand is about the elements that constitute the canonical idea of intelligence. We define intelligence as the ability to acquire skills that narrow this search, diversify it and help steer it to more promising areas. To accomplish this, we focus on three basic aspects of intelligence. Intelligence today is considered as a fundamental property of each and every living being. Intelligence can be defined as a predominantly human ability to accomplish tasks that are generally hard for computers and animals. Artificial Intelligence An important aspect of intelligence is the ability to adapt to a novel task without any direct experience (zero-shot), based on its relationship to previous tasks. Intelligence is not a single weighable quantity but a subject that spans Biology, Physics, Philosophy, Cognitive Science, Neuroscience, Psychology, and Computer Science. (Artificial Intelligence) Is understanding a necessary and sufficient condition for intelligence? The disadvantages of this type of intelligence are discussed. Artificial Intelligence Nonetheless, most of these systems lack many capabilities that we would naturally consider to be included in a notion of "intelligence". Human intelligence has the remarkable ability to assemble basic skills into complex ones so as to solve complex tasks.

Such is more difficult to consume, but if I measure the cosine distances between each sentence and its sucessor, I can tranform the result into pseudo-paragraphs, which is (usually) easier to consume:

This paper is a survey of a large number of informal definitions of ``intelligence'' that the authors have collected over the years.

Finally, we survey the many other tests and definitions of intelligence that have been proposed for machines. Independent from the still ongoing research in measuring individual intelligence, we anticipate and provide a framework for measuring collective intelligence. Our understanding of intelligence is directed primarily at the human level. In this regard, intelligence becomes a product of information processing. In such a case, intelligence arises as a result of information processing. The discussion at hand is about the elements that constitute the canonical idea of intelligence. We define intelligence as the ability to acquire skills that narrow this search, diversify it and help steer it to more promising areas. To accomplish this, we focus on three basic aspects of intelligence. Intelligence today is considered as a fundamental property of each and every living being. Intelligence can be defined as a predominantly human ability to accomplish tasks that are generally hard for computers and animals. Artificial Intelligence An important aspect of intelligence is the ability to adapt to a novel task without any direct experience (zero-shot), based on its relationship to previous tasks. Intelligence is not a single weighable quantity but a subject that spans Biology, Physics, Philosophy, Cognitive Science, Neuroscience, Psychology, and Computer Science. (Artificial Intelligence) Is understanding a necessary and sufficient condition for intelligence? The disadvantages of this type of intelligence are discussed. Artificial Intelligence Nonetheless, most of these systems lack many capabilities that we would naturally consider to be included in a notion of "intelligence". Human intelligence has the remarkable ability to assemble basic skills into complex ones so as to solve complex tasks.

This particular example is not extraordinarily helpful because only two "paragraphs" are returned. This tells me each sentence is very similar to its predecessor.

To distill the original results another way, I can submit it to an LLM and tell the model to summarize:

The paper surveys various informal definitions of intelligence and proposes a framework for measuring collective intelligence independent of individual intelligence research. The authors define intelligence as the ability to acquire skills that narrow, diversify, and steer search towards more promising areas. They focus on three aspects of intelligence: (1) adapting to novel tasks without direct experience (zero-shot), (2) assembling basic skills into complex ones to solve complex tasks, and (3) understanding. The paper also discusses the limitations of artificial intelligence systems in terms of their lack of certain capabilities that are inherent in human intelligence. Finally, the authors argue that intelligence is a multidisciplinary subject that spans biology, physics, philosophy, cognitive science, neuroscience, psychology, and computer science.

Again, not bad, especially if you take the time to close read the pseudo-paragraphs.

Finally, I submitted the original results to another LLM, but this time I posed a question, "What is intelligence?", and what follows is the result:

Intelligence is a multifaceted concept that has been studied and debated by various disciplines, including biology, psychology, neuroscience, philosophy, and computer science. At its core, intelligence can be defined as the ability to acquire and apply knowledge, skills, and adaptability in various tasks and situations. This includes the ability to process information, learn from experience, and solve complex problems.

The canonical idea of intelligence suggests that it is a fundamental property of living beings, including humans, animals, and potentially even artificial intelligence. However, there are also different types of intelligence, such as zero-shot learning, which allows AI systems to adapt to novel tasks without any direct experience.

The discussion around intelligence often focuses on the elements that constitute it, such as the ability to learn, reason, and adapt. While understanding is considered a necessary condition for intelligence, it may not be sufficient on its own. Other factors, such as creativity, problem-solving skills, and emotional intelligence, are also important aspects of human intelligence.

Despite the progress made in artificial intelligence, most systems lack many capabilities that are considered essential for true intelligence, such as common sense, creativity, and emotional intelligence. Therefore, while AI can perform certain tasks with great efficiency, it may not be considered truly intelligent in the way that humans are.

In summary, intelligence is a complex and multifaceted concept that encompasses various aspects of information processing, learning, and adaptability. While AI systems have made significant progress in mimicking human-like intelligence, there are still many capabilities that are unique to humans and may not be easily replicable by machines.

So? What's the point? We are all suffering from information overload, and we use various tools to help us overcome this problem. Some of those tools include: tables of contents, back-of-the-book indexes, page numbers, running chapter headings, eye glasses, pencils and pens, bibliographic management applications, etc. Why not use the computing features of a computer?

Important! The results of this investigation do not return truth. Instead measurements and observations are returned which are expected to be followed up with close reading. To this end, all of the sentences returned from the original search results can be pointed to in their original documents.

Similarly, the results are intended to be jumping off points for dicussion. For example, if I were to ask you "What is intelligence?", then you would say a few things, and somebody else would say a few more things, but this process affords you a larger vocabulary for discussion. You can use the results to "riff" off the original question.

So, what is intelligence? Read the results above and discuss.

For extra credit I could apply the same process to works of epistemology, and then compare and contrast. For example, what do Plato, Aquinas, Locke, Hegel, Kant, etc. have to say on topic of intelligence. Then ask yourself, "If I were to use traditional reading technques to accomplish this task, then how long would it take me?"