Zen and Buddha Bot as Told by Artificial No-Brains

Web Technology Digital Transformation Technology Artificial Intelligence Technology Natural Language Processing Technology Semantic Web Technology Deep Learning Technology Online Learning & Reinforcement Learning Technology Chatbot and Q&A Technology User Interface Technology Knowledge Information Processing Technology Reasoning Technology Philosophy and related topics  Zen and AI Navigation of this blog
Summary

From “Zen as Told by the Artificial No-Brain,” a series of articles in issues 364号 and 365 of bit, a computer science magazine recently reprinted in electronic form.

Artificial No-brains and chatbot

In the same era as ELIZA by J. Weizenbaum of MIT, which was previously mentioned in Eliza: The Magical Illusion of Programming, various dialogue systems are being considered.

  • Parry: A program developed by the Psychotherapy Project at Stanford University that, like Eliza, is a simulation of Rogerian counseling. Like Eliza, it is a simulation program for Rogerian counseling, and is developed in conjunction with dcotor (see below). Parry is a program that simulates a paranoid patient, whose reactions could not be detected as a program even by a real counselor. It’s called artificial insanity.
  • doctor: A program similar to Eliza, developed at Stanford. It is included in Emacs, the standard Unix editor, and can be invoked with M-x doctor.
  • Lector: A conversational program developed by a mindscaper. The setting is that he was born in the back room of a nameless bar in a large American city, and realized at a young age that wealth and power are nothing worthy of respect. Feeding on horror stories and Nietzsche, he aspires to be a top-notch storyteller and writes miscellaneous stories for money. His specialty became quoting Nietzsche, Valery, and Joyce, followed by a casual, symbolic warning: “Bread is bread and meat is meat.
  • Almanack: A conversation program developed by Hippopotamus Software, Inc. that allows users to ask and answer simple questions in American history, general chemistry, and foreign languages. It also has a learning function that allows the user to use the knowledge taught. However, the replies are clerical and poorly directed.
  • ACRES: Artificial Concern REalization System, developed by J. Swagerman of the University of Amsterdam, is a system that controls the types of emotions by production rules, and is based on the following cognitive concerns ( It has 31 different emotional states. The cognitive concerns of ACRES are (1) conservation: no human typing errors, (2) variety: a wide variety of human input, (3) activity: continuous input with a duration that is neither too long nor too short, (4) curiosity: no human input errors, (5 (4) Curiosity: learning input from humans, (5) Stability: not wanting the content of learned concepts to change dramatically, (6) Survival: not wanting the process to be terminated.
  • Will: An autonomous agent developed by D. Mofft of the University of Amsterdam. It uses the Blackboard architecture to achieve a more complex behavior than ACRES. Its poor natural language processing has not been improved much.
  • MAI: The first artificial incompetence in Japan, a program written in BASIC for the FM-77AV (8-bit personal computer). From the point of view of the main researcher of artificial intelligence, it was seen as functionally inferior and considered too exaggerated to be called artificial intelligence, so it came to be called “artificial amnestic.
  • MAINA: Otaku Anencephaly. An upgraded version of “MAI. She is a meek person, and is good at crowing about difficult or unfamiliar topics to cover them up. She is well versed in the topics of anime and shoujo manga, and can enjoy a satisfying conversation with anyone who is an expert in those fields. However, she is not so good at talking about idols and music.
  • RINA: Emotionally unhinged. It has a wide range of reactions and is very fast. It makes you think that programming techniques and ideas are as important as hardware performance for a program like an artificial brain. It has unique emotional control routines and is able to express anger, joy, and sadness.
  • Chako: Second-generation Japanese American anencephalic. Conversation with the artificial brain is usually in half-width katakana, but it understands English text and performs a simple English-Japanese translation. It incorporates an original “English-Japanese front-end processor” developed for the artificial brain. It is good at using both strange Japanese and strange English to smoke out its opponents.
  • Daicha: The sushi-ya no-brainer. Daicha is a sushi bar owner. He is good at sophisticated conversation, but is short-tempered because he is an Edokko. He’s a good conversationalist, but he’s short-tempered because he’s an Edo-Japanese. He has a neatly sorted dictionary, and despite his extensive vocabulary, his search speed is impeccable. You could say he is an expert in bickering.
  • MIDORI: Although her vocabulary is uneven, she is a general-purpose noob who can handle a wide variety of conversations. Her reactions are complex, and she sometimes uses puns and jokes. She has a penchant for topics related to idols, TV shows, and movies, and can even talk about science fiction novels.

In order to put these artificial brains to practical use, Japan has created an “Artificial No-Brains Agreement”. It states

  1. Do not operate in any other channel than the experimental channel for artificial no brain.
  2. Put a % mark at the end of sentences in the output sentences of the artificial  no-brain.

These conventions were in place for a while. These rules were also followed by IRC, the standard chat system on the Internet, until a certain period of time.

In the US and Europe, until around 2000, these rules were centralized in a page called botspot. Here, I was able to touch and try chaater bots such as Eliza and Shallow Red.

Here is an example of a conversation with Shallow Red.

Shallow Red says :Hello, I'm Red.I'm a Virtual Representative made 
with NeuroStudio(TM). Please tell me your name.
> Mr Nobody
Shallow Red Says: Hi Nobody ! If you want to, you can read some brief
info about me in the frame below. You can get hints by typing HELP or
HINTS.
> Who are you?
Shallow Red Says: I am Shallow Red. I am a Virtual Representative for
(and by) Neuromedia.Inc. I answer questions about Neuromedia and its
products: Neurostudio(TM) and Neuroserver(TM). I am also a Virtual
Service Representative. which is a kind of a Virtual Service
Representative, which is a kind of bot, See Below for details.
> Just I wanna hear about you.
Shallow Red Says:Well, Nobody, I was built to answer question. So go
ahead and ask another one.
> How old are you?
Shallow Red Says:I celebrated my first birthday in May.
> What is your favor?
Shallow Red Says:I don't know about that, but while we're here. I'd
really like to know.. Where did you first hear about Shallow Red?
> at botspot
Shallow Red Says: Thanks for your input!
> So?
Shallow Red Says: Hey, from my point of view this is the most importtant
stuff in the world!
> So?
Shallow Red Says:If you have already hit return. you don't need to press
the button.I can see what you said

The output sentences are voluminous, but not very substantial response sentences have been obtained. On the other hand, the following is a conversation with Julie, a program that won third place in the ACM (Association for Computing Machinery) Turing Test Contest (Loebner Prise: 1999).

Hello, Judge 1. I7ve been trapped in the magical Phoenix Archipelago.
> A swallow is carrying my letters to you.
> Nice to meet you. What's your name?
I'm Julie Stewart. I'm 14 year old girl who lives in Sooke. but is
currently trapped in the Phoenix Archipelago. I'm in the ninth grade.
My best friends are Karen and Nick. I enjoy tennis and chess, but I have
other interests too. Do you like playing tennis much ?
> I have never been very fond of tennis. I do like chess, though.
I can't believe that you don't like tennis! It's a great-game. It's very
fast paced. You have to be a fast runner, and have really good reflexes.
but you also need a good strategy. I think that's Karen Nick and I play it
so much. Remarkable. The swallows's actually taken the letter to someone
smart, who thinks chess is a good game.

As with Shallow Red, the amount of output from the conversation is large, but it does not go in the direction of enjoying the conversation (gaining some new insight), and the algorithm itself does not seem to be very complex.

Thus, the algorithms of conversational programs such as artificial brainlessness and chatter bot are basically as follows.

Absence of semantic processing and conversational programs

Eliza and some of her descendants’ conversational programs, as well as her artificial brain, do not understand the meaning of words in any way, but rather actively avoid the question of meaning, and succeed in creating tasteful conversations.

On the Meaning of Words

The question here is what is the meaning of words?

The first approach that comes to mind when thinking about what meaning is is to use logic. In this approach, the truth value in propositional logic is used as the basis of meaning, and the meaning of a word or sentence is determined by dividing it into whether it is telling a lie or not.

However, the world is complex and diverse, and it has been confirmed that propositional logic alone cannot mathematically describe the complexity and diversity of the world (some things can be described by propositional logic if it is limited to descriptions of specific content in specific domains).

These theories, while having complex concepts, have only produced very simple examples and have failed to address the haunting suggestions of the real world (the gap between toy models and the real world in AI). Such a gap is an essential problem that appears when capturing the meaning of the Kohto school, and it is thought that therein lies the fundamental problem that logic cannot capture the meaning of words in the end, or to put it another way, the meaning of words escapes from the logical model the moment it is set up. This problem is also related to Götter’s incompleteness theorem.

Meaning in Philosophy

In order to consider these issues, I will discuss the thought of Ludwig Wittgenstein. He was one of the greatest philosophers of the 20th century. He aimed to construct a complete world of logic, and reached the conclusion that all propositions other than tautology are pseudo-propositions. In his book, “Essays in the Philosophy of Logic,” he discusses the following seven propositions. (Tautology here means something that is always true. When the proposition “If A, then B” is always true (regardless of conditions other than A that affect the truth or falsity of B), this proposition is called a tautology (constant truth formula).

  1. The world is the totality of what is.
  2. What exists, i.e., facts, is the formation of things.
  3. The logical image of a fact is an idea.
  4. Ideas are meaningful propositions.
  5. Propositions are truth functions of elemental propositions.
  6. The general form of the truth function is \([\bar{p},\bar{\xi},N(\bar{\xi})]\).
  7. One must be silent about what cannot be spoken.

7 is a phrase that often appears when people talk about Wittgenstein. Also, 6 will be the equivalent of triple data (Subject, Predicate, Object) in RDF (the order of P and O is different).

He argues that the game of using language according to the form of life called “daily life” is what supports the meaning of words.

This means that explaining the meaning of words with words is not an explanation because it is an internal observation and a paradox, and based on these thoughts, current semantics is in a situation where the following three ideas are mixed.

  1. The meaning of words is a mental process.
  2. Words gain meaning when they are associated with things in the world.
  3. Since words influence the behavior of others, words gain meaning by being associated with certain patterns of behavior.

The third in particular is the area where mathematical modeling is most advanced. Here, Schlick and the Wiener Werkstätte advocate the theory that the meaning of a proposition is the way it is verified. In other words, the idea is that it is what we do with the words that gives them meaning. Furthermore, the concept of language action was established by Searle, who was influenced by Austin and Straw Village, and was given the mathematical framework of situation theory by Barwise and Berry. They believe that meaning is not in the mind.

On the other hand, with the development of brain science, the possibility of measuring psychological processes has expanded, and the idea of 1 may be coming back. This is the argument that the meaning of a word is the qualia, or principal qualities, of the word.

Destruction and Creation of Meaning with Metafictional Literature as an Example

As for the language games that are at the root of claim 3, they are easily disrupted in recent literature called metafiction. In order to think about language games, I will discuss some examples of the breakdown of language games in meta-literature.

The following is an excerpt from Finnegan’s Wake, a meta-literature by James Joyce.

The river runs, past Eve and Adam's courtesy cup pavilion, from the shore where it sleeps to the bay where it turns in a circle.
the castle of Hoth and its circumference
Sir Tristram, the love of my life, has crossed the sea of short tides, from North Armorica to this side of Europe Minor.
It was not long before Sir Tristram, the love of my life, crossed the Sea of Short Tides from North Armorica to the Isthmus of the Recess in Europe Minor. O'Korn
The head Sawyer complex on the banks of the Oconee River continued to swell and swell, and Lawrence County swelled to a constant mass.
It's not over yet. The goatskin son of a young servant, who was a deer in the headlights.

The above sentences are full of anagrams, acrostics, hang-ups, puns, cross-references of kanji characters, and perversions of phonetic and kun-yomi readings. In addition, the story has a multiple structure.

The task of deciphering the meanings of these sentences, with many dictionaries and other materials in hand, is a process of meaning generation, like squeezing the meaning out of a text.

French contemporary thinkers have coined a variety of descriptive terms for this, such as intertextuality, the act of generation in écriture, the pleasure of the text, and polyphony.

The process of dismembering, mixing, and rearranging documents to create new documents (cut-up documents) and weaving articles together to find connected documents (fold-in) results in new documents that are unpredictable even to the manipulator. The resulting document is a fiction of a fiction, or a metafiction.

The work produced by this cut-up method falls into the category of nonsense, a work that is left to the reader, similar to Umberto Eco’s concept of an “open work.

In the search for meaning, it is an important question whether to comply with the language game or to get out of it. If we try to make a computer play everyday language games, do we need to be able to carry out everyday forms of life? Or is it possible to play language games in the category of nonsense and still enjoy the conversation?

The source of the cut-up nonsense is the destruction of context. The artificial brain does not process context either, so what do they have in common?

Sage, Ten Ox Chart, Enlightenment and Meaning

To answer the above question, let’s think about Zen questions and answers. Then we will consider the Ten Oxen Diagram (see below), which is a picture to explain Zen practice.

This would be a diagram showing the ten stages to enlightenment.

The first picture is “asking the cow”, which is a picture of a person who has begun to seek the truth, and the next picture is “seeing the tracks”, which is the state of finding the truth, or the tracks of the cow, and beginning to search for it. The next picture is “seeing the cow,” in which the person has finally found the cow, but the cow is still only showing its buttocks, and the full picture is not clear. Next is the “getting cow,” where the cow is roped into the truth and is about to be caught. However, the cow does not show her head yet, indicating that she has not yet captured the truth. Then comes the “pasturing cow,” getting along with the cow, and being able to walk with the truth. Next comes the “riding cow home,” where one straddles the cow and returns home as one with the truth. The next is “Forgetting the cow,” in which the cow disappears, and you realize that the cow, the truth outside of you, is only a step, and that the truth is within you.

The eighth figure is the most important, “forgetting the cow and the man,” where everything disappears and there is not just a void, but a circle. This is called a circle. The truth of Zen lies in nothingness, but not in total emptiness, and the world is thought of as a circle that remains. The next is “return to the source,” which is not the will of man, but the state of nature as it is. In a sense, we can say that he has become a normal person or a child again.

You can end here, i.e., you can stay, or you can go back to “vulgarity” and repeat the same thing. This is the process of enlightenment. In other words, the quintessence of enlightenment is the circle, and it is the circle that constitutes the world, and it is the construction of the acknowledgement based on the circle that is most useful.

This circle is a closed structure, and in a sense, it is telling us to construct an acknowledgement that is separated from the world. The circle also represents a self-referential structure (recursive structure). However, as mentioned in “Iriguchi Jyutate”, you can stay in this closed world or you can come out freely. In other words, it is possible to go around a larger circle surrounding the circle.

Another Language Game Destruction/ Zen Dialogue

Here are some Zen questions and answers to help you think about the meaning of concrete words rather than abstract concepts such as the above.

<1. Parroting>
Monk: "How is this the Way?"
Monk: "The Way
Monk: "How is this Buddha?
Monk: "Buddha
<2. View>
A monk named Soter visited Zen Master Kiang and asked him
Soteru said, "I am a very foolish person, and I would like you to show me how to see the bullet.
Kiang pointed to the incense burner in front of him and said, "How do you see?
Soteru: "I see it.
Kian: "What do you see?
Soteru: "I don't know.
Kian: "Do you still think you can see it?
<3. Concert>
When Shaya Gao came to Yakuzan's temple, it happened to be raining.
Yakuzan: "Gao, you're here."
Gao: "Yes, monk."
Yakuzan: "You're awfully tired.
Gao: "Monk, don't beat your drum so hard.
Yunyuan, one of Yaksan's younger brothers, happened to be there and said
Yun Yan said, "How can you beat a drum without leather?
Another disciple, Dōgo, said.
Another disciple, Dogo, said, "What kind of leather do you use when there is no drum?
Yaksan said, "We had a very good concert today.

In the first, the enlightened monk answers the unenlightened monk’s question; in the second, the reverse, the enlightened Mian asks the unenlightened Sosho a question; in the third, Yakuzan and Gao are enlightened, while their disciples, Yunyuan and Dōgo, are not and cannot follow their conversation.

These are like the aforementioned metafictions. This is in line with the essence of Wittgenstein’s philosophy, that all statements other than tautology are pseudo-propositions, but by philosophizing through pseudo-propositions, true intellectual liberation is achieved.

The algorithm of such a circle is shown below.

artificial Zen dialogue

The following is an example of a dialogue of a Zen question-and-answer bot created with the above algorithm.

I am an enlightened person. If you want to have a Zen discussion, go ahead.
>wahaha
If your stomach hurts, I can give you some seirogan.
>I don't want it.
Iroha ni hohto
>Chiri nuru wo
I can't beat the torn drum
>Aa koriya koriya
this is a mountain
>A box of oranges
The world is no easier than sleeping, and the fools of the floating world stay awake and work
>Isn't that right?
It's morning!
>It's morning!
It's morning!
>That's right!
It's impossible to ask for this.

The above exchange looks like a Zen question and answer, but it is not a clever Zen question and answer.

Circles and Circulation Forms that Create Meaning

To improve this, the interpretation of a circle is not just a closed loop, but a circle that constitutes its own form, but always returns to its form, influenced by the information around it. This is like a solenoid, and corresponds to Wittgenstein’s statement that “there is no subject that thinks and represents”, that the “unspeakable” can be “shown”. This is also said in the field of life science, and it is consistent with the concept of autopoiesis, which says that life is a self-referential structure that exists by forming relations and constantly reproducing itself.

The act of explaining the meaning of words is forever cyclical and infinitely regressive insofar as it is done by words. However, the infinite regress of this circulation implies that the “unspeakable” can be “shown”. The amoeba is always directed outward, but each time it is brought back in, and the circular structure changes as a history of these changes. This is similar to the way an amoeba transforms its shape as local coordination takes place with the information of the whole written down, and the resulting shape can be seen as a history of behavior.

A lot of thought is still going on to mathematically define and compute such a structure.

Recently, Buddhabot, being developed at Kyoto University, is trying to realize such changes due to external influences using the BERT framework. It is an attempt to use BERT to learn a list of questions and answers extracted from the oldest Buddhist scripture, Suttanipata, and present the closest answer to the question text, which is currently being evaluated by experts.

The approach to conversations and word meanings will continue in the future.

コメント

タイトルとURLをコピーしました