There is a singular attractor in the consciousness of the global tech community: chatbots. I was fortunate to attend the ASITE (AIM) “ChatGPT in Education” forum on how to empower teachers with Large Language Models. Professor Christopher Monterola gave a fascinating tour of the mind-chilling power of the chatbots available for free today. Most charming was his wit-laden conversation with ChatGPT “in excellent Filipino.” But I felt that my own personal franchise was ebbing rapidly. The following are the six impressions I came to as a rank outsider in the AI space:
1. Global Wisdom of the Crowd: The so called “wisdom of the crowd” rests on the idea that certain information is scattered among members of the crowd so that the reduced form (average) of this body will approximate some value of interest. The interest is heightened if the average approaches the true value. The likelihood improves with the “judgmental competence” of the sampled members of the crowd. The popular Delphi method directly samples the ideas of members of recognized competence on a particular issue in society so the reduced form is closer to the truth on that issue. Meta analysis in science also follows the same logic. All these descended partly from the Condorcet Jury Theorems (Marquis de Condorcet, 1785): “The judgmental competence of the community deciding by majority vote (50% + 1 one-man-one-vote) between two alternatives A (say, airport) and B (say, seaport) will exceed the judgmental competence of any monarch as the number of voters becomes very large.”
Chatbots train on all the digitized text available on the global net. Thus, its “crowd” is the living global population that use digital media to communicate ideas and the population long dead whose ideas still live in the global net. Instead of a hundred live subjects, it samples billions of respondents, whether dead or alive (from Plato/Aristotle up to 2021 in the case of GPT-4). Its answers to queries are the compressions of the wisdom of the ages but also of the shared “fictions” stealthily imbedded therein. Western biases will naturally be heavily weighted. Sure, they can hallucinate and make mistakes, but that is a weakness they share with humans (“To err is human”). That being said, with chatbots, you are literally in a Socratic dialogue with the occidentally slanted wisdom of the ages. No wonder that they sound so wise to us acolytes of the Western dominated textual global resources!
2. I was personally chastened by ChatGPT’s and BARD’s responses to my prompt for an alternative plotline to Noli Me Tangere and El Filibusterismo: Rather than as a sadist, paint Padre Damaso as an average struggling cleric with a tick for the “rule of law” and an aversion for extreme wealth inequality among the indios. The responses had Padre Damaso dig into the rumor that the wealthy Ybarra family had a hidden and perhaps ill-gotten wealth located in the family mausoleum (revealed in El Fili). In BARD, the confrontation ends up with Padre Damaso defeated and exiled from San Diego with the hidden wealth question answered in favor of the Ybarras; in ChatGPT, Padre Damaso discovers that indeed the Ybarra wealth had a guilty provenance but is conflicted on what was in the best interest of the townspeople. He opts to enter into an entente cordiale with the Ybarras. In both accounts, the controversial wealth ends up being used to uplift the townspeople. The fairytale ending is perhaps predictable, since fairytale endings dominate the statistical average wisdom in written text; tragic and the morbid endings (as in Hans C. Andersen’s The Little Match Girl or as El Fili’s) are a decided minority.
3. On Philippine education: One reason among many for poor outcomes in PISA (Program for International Student Assessment) et al., say, is poor teaching. Teachers impart the information and ideas that they themselves understand vaguely to students and often resort to authority for validity. Centralized monitoring of teachers is difficult besides being culturally and politically dissonant for principals. (1.) Teacher performance and accuracy of materials could be monitored de-centrally by students who can compare learning materials in real time with a simple click in the 3-D phones; these allows students to become “precocious,” ones who can readily correct teacher mistakes. The educational system’s response is to consider these kids as disruptive influences rather than as the best opportunity for learning. (2.) Teacher capabilities will also be upgraded by having chatbots as teaching assistants at their fingertips. It may even take costly textbooks out of the pedagogical equation. Thus, a great promise resides in universalizing access of both students and teachers to open access chatbots. (3.) Corollary to this is the necessity for affordable universal access to reliable internet signals.
4. A related issue: The Hippasitic oath. Hippasus was a member of the Pythagorean thought collective that preached that the world as experienced is an imperfect representation of the world of forms hovering above us in the knowledge sphere. This world of forms is written in the language of mathematics (then of integers, ratios, and geometry: 2πr is true for all circles). But there was a blemish in the mathematics then known: it cannot come to terms with irrational numbers, which is neither of these. This threatened the perfection of the Pythagorean dogma. Therefore, Pythagoras, by virtue of his authority, forbade the members to contemplate — let alone talk — about irrational numbers. Hippasus however obsessed over the idea and deduced from accepted logical reasoning the existence of such numbers. The story goes that Pythagoras ordered Hippasus be put to death by drowning. A case of Pythagoras hallucinating? For irrational numbers were found later to be extremely useful in the mathematics of complex numbers.
5. In this regard: (i) Every teacher should be required to take the Hippasitic oath: “I should never claim validity on the basis of authority; I will only claim it on the basis of sound logic and evidence!” Related is that: (ii) Every teacher should swear by the axiom of scientific provisionality: “Every scientific canon is provisional.” That means the attitude inculcated among students is that even Einstein can be wrong. Advanced chatbot replies or a combination of them can be considered as provisional benchmarks of truth or falsehood in the classroom setting. (iii) Teacher training must shift to inculcating in teachers the idea that doubting Thomases among students are to be nurtured, not suppressed. (iv) The Department of Education and Commission on Higher Education should formulate clear rules against using authority as proof of validity.
6. I confess I now am bordering on hallucination! But that, I claim provisionally, is a privilege of every sentient/human.
Raul V. Fabella is a retired professor of the UP School of Economics, a member of the National Academy of Science and Technology and an honorary professor of the Asian Institute of Management. He gets his dopamine fix from bicycling and tending flowers with wife Teena.