Truth in the AI era

Can legislation enable this?

I still remember, in my school days, when my classmates and I fell into an argument about who invented “the zero” (the zero of 0, 1, 2, 3…), some of them said “god” and some others said, “Zero was always there and didn’t need any invention.” And when we finally went to our maths teacher, who said “Euclid”, we fell quiet: Our strict maths teacher, whom we respected and, I should confess, feared (he could fail you if you went against his instructions, and in that school, if you failed in maths, you risked not being promoted to the next class). What our school teachers asserted was the “truth”. It took me another 25 years to figure out that Euclid had nothing to do with it and the zero was a creation of some ancient Indian, possibly Aryabhatta or Brahmagupta.

As I write this today and a battle royale is raging between ChatGPT, Google’s Gemini, Perplexity and several others about which of these artificial intelligence tools provides the most “truthful” answers, I cannot help but think of all the other contenders for such a role in the past few decades. For me, other than our school teachers, there was my mother (amma), the young uncles and aunts in our home while we grew up, our cricket coach (“will a yorker work on this pitch”), Encyclopedia Britannica in our school library and many others.

It appears that I, like all other humans, have always needed an arbitrator whose judgement helped us decide what was “right” or “correct”. In recent years it has been internet search engines like Google claiming the role that my amma, my school and college teachers and the Encyclopedias played in my life. And it looks like the new claimant will be the winner of the Chat AI battles going on.

Maybe we will learn something about us human beings if we find out what type of questions we folks ask ChatGPT, for example. It appears that the topmost category of questions is fact-checking questions such as “who was the 16th President of the US?” or “How far is Calcutta from New Delhi?” The second-most popular type of question is about weather information: “How’s the weather in Bangalore today?” The third-most popular are language questions: “What does the word ‘epicentre’ mean?”). The fourth-most popular are technology-related (“How can I improve this Python code?”). And, finally, the fifth-most popular are about historical events and personalities (“Who was C V Raman?”). All these, on reflection, are not much different than what I (and, I am sure, you dear reader) asked my school teachers, my amma, my dad, my uncles, and aunts in our growing-up years.

It appears that in February 2024 alone, more than 180 million people in the world asked ChatGPT at least one such question with India (at 9 per cent of all users) coming close to the United States (10 per cent of all users) in the ranking of users by country. People in the 30-44 age group use it the most, followed by those in the 18-29 age group. Does this mean that a whole new generation is growing up learning their “truth” from ChatGPT and its like? And this, of course, brings up a sensitive and complex question: Do ChatGPT, Gemini and their like require regulatory supervision?

We can of course, as a country, save ourselves the bother by banning these products outright as China, Russia, Iran and some others have done, but that is not India’s style.

So, before answering that question, it is worth casting a glance on our present system of establishing the “truth”. For example, to establish whether a business’s financial statements are “truthful” we have a system whereby a group of people called “chartered accountants” must certify that. If one member in a married couple wants to exit from the marriage, they have to appear before a court and establish the “truth” of that claim. If you want to sell a medicinal “drug” in India, you must get that certified by the drug controller, a government authority. Before releasing a film in India, you have to get the ok of the Central Board of Film Certification. For the internet India has regulations under which government bodies can make a request to internet sites to remove certain types of content (“porn” and “defamatory” are two such types). The breakthrough regulation that prevented the internet being choked down by law suits and screams is the separation of action on the internet between the creators of content and “intermediaries”, where internet sites were classified as mere “intermediaries” who provided tools for publishing content and the actual responsibility for content was borne by people who created and posted the content. This was done by inserting Section 79 of the India Information Technology Act 2000, which, many observers note, was what made the internet possible (disclosure: I personally wrote it as a member of a government committee).

While there is little doubt that some types of content on ChatGPT-type entities will generate protests in India and lead to such content being directed officially to be removed, the larger question is: Can ChatGPT-type content creators lead to influencing public opinion in “wrong” or “prejudiced” directions, particularly among impressionable young people?

More importantly, should ChatGPT-type entities be granted the legal role of “intermediaries” under the Indian IT Act when they themselves are increasingly claiming to “generate” original content? What new legislation will protect the “truth” while at the same time not discouraging innovation in this area?

---

The writer (ajitb@rediffmail.com) is an internet entrepreneur