Van Gogh Starry Night
AI can ‘paint like’ Vincent van Gogh based on the information it holds, like his choice of subjects and his brush strokes. But taking in the spectator – that something is by van Gogh – is not the same as producing ‘art’. Wikimedia Commons image shows ‘The Starry Night’ by van Gogh.

New methods will have to be invented to test students in era of ChatGPT

AI can produce student essays, but that’s not the final product in education — what counts is training the human mind

The public space has been abuzz with the giant strides taken by artificial intelligence (AI) in the past year or so. The latest sensation is ChatGPT, which has the propensity to produce student essays on virtually any subject, compose music, write teleplays, fairy tales, and answer test questions at a level above that of the average intelligent human being. Where a human being could take days doing it, ChatGPT can achieve that in seconds. 

Since these are abilities that education is expected to inculcate in students, the question is what education should now try to achieve since students will simply take recourse to ChatGPT instead of producing work.

Before speculating on the matter, it is necessary to first acknowledge that AI is not like human intelligence in that it solves problems through statistical methods without ‘understanding’ what it is actually doing. For instance, AI could produce a painting that can pass for Vincent Van Gogh’s based on the information fed to it on what elements have been identified with the painter’s work – like his choice of subjects and his brush strokes. 

But taking in the spectator – that something is by Van Gogh – is not quite the same as producing ‘art’. 

Watch: Will ChatGPT make teachers redundant in the near future?

The pronouncement that something is ‘great art’ is made after much debate has taken place in the public space and these debates have raised philosophical issues of importance. The role of movements in art and their historical purposes would naturally be among them. 

AI mimicking Van Gogh does not go through these lines of ideation but simply achieves resemblances. An AI-created Van Gogh picture that fools the expert only calls into question the notion of expertise in art evaluation.

Obstacle or aid in education? 

Coming now to the production of student essays by ChatGPT as an illustration of what AI can do for education. It is not the student essay that is the final product in education but the training of the human mind. The student essay was only the evidence produced by the student that his or her mind had been trained. Producing essays of a high order – as judged by the teacher – does not mean that the art of essay-writing is redundant but that new methods will have to be invented to test students before they are deemed to be ‘trained’.

Unfortunately for us, learning by rote was encouraged in the education system and that will evidently be the first casualty since AI regurgitates lessons much better. In mathematics, for instance, a ‘good student’ is encouraged to work out every single problem listed in the exercises section in textbooks with the understanding that they will exhaust all the possibilities that the student will be confronted with in an exam; a sounder teaching method might have been to get at the underlying principles. 

The Federal Premium: Why a Nobel laureate feels India is not doing enough for science

Seen in this context, AI is actually an obstacle in the path of education rather than an aid. Education will need to produce human beings who think “as humans are capable of” and not simply “solve problems” which AI could well do. It will become the additional task of sifting out AI-aided material submitted by students.

High-tech plagiarism

Having taught undergraduate and graduate students off and on, my mind goes straight away to how students at that level will be dealt with in the age of AI. Since what is important is to test the level of ‘understanding’, it seemed opportune to mark the students on the basis of the intelligence of their questions with regard to each lesson. 

The first test criterion announced was that the same questions should not be repeated by different students – which would imply that they were copying from each other.  The questions should also not simply pertain to facts (e.g.: “How many planets are there in the solar system?”) but should go on from there to inquire into other phenomena (e.g.: “Are the orbits of all the planets on one plane?”).

Also read: Centre ropes in tech giants to use artificial intelligence in agriculture

I discovered that, by and large, the students at that educational level are unable to respond to these requirements satisfactorily and, most often, they simply copy from each other. The students are, however, not to be entirely blamed because the teachers they have been taught by have simply handed down facts to them as ‘knowledge’ and they have duly memorised them. Any deviation from this norm naturally perplexes them. 

AI may have a much larger range to copy from but copying from an enormous multitude of sources is also what it is doing. It is this aspect of AI that made Noam Chomsky describe it as “high-tech plagiarism”. If students actually ask difficult questions by using their ‘human intelligence’, one may be sure that they will be penalised for their arrogance. AI is better behaved here since it does not have the ego of the incompetent teacher.

Need for enquiring minds

This gives us a clue into what kind of minds should be created by education and the answer is ‘enquiring minds.’ If the thinking up of intelligent questions is gathered to be a criterion by which understanding is judged, those questions should not simply clear up doubts but should open up the field for more questions. The orbits of the planets being on one plane would, for instance, call into question how the solar system was formed, and the notion of centrifugal and centripetal forces attaining equilibrium.

One doubts that the kinds of minds created by even doctoral studies in India will be able, by and large, to respond ‘intelligently’ to knowledge inputs – in the sense of their leading to more inquiry and opening up the field to further questioning. If one studies the titles of doctoral dissertations produced by education in India, one is not struck by their clarity but usually by the awkwardness of the ideas. 

This suggests that there is a paucity of fundamental understanding and most students take refuge in some esoteric/obscure corner of the given field rather than where conceptual clarity will be called into question.

Science: ChatGPT is confronting, but humans have always adapted to new technology

The following suppositions may be contested but I would like to argue that AI will find fundamental reasoning elusive and will simply gorge on available facts. I implied that the test for human intelligence in future could be the way in which it differentiates itself from AI but our education system has hitherto focussed on producing automatons, ‘artificial intelligences’ of a fallible kind since they have learned things by rote, which AI does better. 

Since these ‘human robots’ now crowd the education system at various levels, a question that should be put before formulating a future education policy is how actual ‘human thinking’ can be inculcated through them when their kind of ‘thinking’ is not what human beings should pride themselves in. Would not the most gifted among the students then have to rely entirely on their own resources to produce original work since education itself is indistinguishable from ChatGPT?

(MK Raghavendra is a writer on culture, literature, cinema and politics)  

(The Federal seeks to present views and opinions from all sides of the spectrum. The information, ideas or opinions in the articles are of the author and do not necessarily reflect the views of The Federal)

Read More
Next Story