Subtle problems with AI

Published: 14 Oct 2025 at 15:08 ADT

AI

I recently did a Google search for the word ashuddha vikalpa. This term originates from the Shaiva tantra branch of tantra, referring to “impure” vikalpas. Vikalpas is a broader term used in both Buddhism and Hinduism that refers to mental constructs that are shaped by thoughts or emotions. In other words, mental constructs that are shaped and colored by karma. ’Ashuddha vikalpa’ is actually a negative construction. ’A’ in Sanskrit is a prefix that implies negation, similar to English ’un’ in ’untoward’ or ’im’ in ’impure’. The classic usage of this that many are probably familiar with is the Sanskrit term ’ahimsa’, which means ’non-violence’. The ’a’ in ’ahimsa’ takes the place of ’non’, and ’himsa’ is analogous to the world violence. ’Ahimsa’, therefore, is actually quite accurately translated directly to ’non-violence’. ’Shuddha’ means ’pure’, so the ’a’ in ’ashudda’ takes the place of ’im’ in ’impure’.

The first result on Google was, as expected, the infamous “AI overview” in which Google attempted, valiantly, to define the term.

“Ashudya vikalpa” likely refers to the Sanskrit concept of vikalpa, but with an emphasis on its negative or “unclean” (ashudya) aspects, such as false imagination, fantasy, or delusion, as opposed to its more neutral meaning of “alternative” or “options.” In Yoga philosophy, it is considered a “disturbance of the mind” (chitta vritti nirodha) that creates mental activity without a basis in reality.

A valiant attempt indeed! Especially since there were actually only a small handful of results for ’ashuddha vikapla’; an admittedly very obscure term from a largely abandoned religious tradition that is over 1000 years old ¹. But there are some problems. For one thing, it references the passage ’chitta vritti nirohda’. This is a widely known passage from the famous ’Yoga Sutra’ by the Patanjali. Why did it do that? There are very few results for this search, and I believe the answer lies in the fact that one of the few search results is on the broader term ’vikalpa’, which refers to ’citta vritti nirohda’. ’Citta vritti nirohda’ means cessation of the disturbances of mind (chitta: mind, vritti: disturbance/fluctuation, nirhoda: cessation), but this overview makes it seem like ’chitta vritti nirohda’ as a whole means the disturbance of mind. This is a mistake or accidental connotation that an expert in the field would never make. It is, indeed, so amateurish that even I, with limited knowledge of Sanskrit, noticed it immediately, but which someone unfamiliar with the topic would most likely not realize. Later in the overview, it defines ’vritti’:

Definition: Vikalpa is a Sanskrit term for one of the five types of mental fluctuations (vritti) described by Patanjali in the Yoga Sutras. It refers to the mind creating or entertaining ideas, thoughts, or images that are not based on actual, present reality.

It never gave a full and accurate definition of ’chitta vritti nirohda’ as a whole; indeed, only the misleading and confusing definition. Furthermore, it defines ’ashuddha vikalpa’ as a disturbance of mind, ’chitta vritti’, which implies that the negation or polar opposite of ’ashuddha vikalpa’, ’shudda vikalpa’ is not ’chitta vritti’, which is entirely inaccurate. It is, indeed, defining ’vikalpa’ and not ’ashudda vikalpa’ because that is the only term for which there is a “reliable source”, a Wikipedia article (let us give thanks that at least the only reliable source is not Reddit). The distinction between ’ashudda’ and ’shuddha’ is that, while they are both disturbances of mind, ’ashudda vikalpa’ are those that maintain and deepen delusion or ignorance, while ’shuddha vikalpa’ are more pure and “reality aligned” thus leading to “divine grace”, “revelation” etc. These are versions of vikalpa that can be self-dissolving; thus resulting in an undisturbed mind ’chitta vritti nirohda’. Again, these distinctions are very subtle and easily overlooked, but I would argue that these are subtly incorrect implications that would never appear in the writing of an expert on the topic, as they are clearly misleading and are actually worse because they are so subtle.

If the AI overview said ’ashudda vikalpa’ are flying pig-pony-men, you would, of course, immediately disregard this absurd definition and seek out a more authoritative one. It is precisely because this definition is so close to being reliable and accurate that a reader not fully acquainted with the subject matter would not seek out another definition. They would consider this definition to be “good enough” and most likely move on with their yoga or reading or whatever activity they were performing before seeking to define the term. Indeed, for this AI overview to be truly useful, you must already be an expert or at least well enough acquainted with the subject matter to be able to notice and correct these subtle mistakes. This leaves us with a pressing question: What is the use of a source of knowledge that presupposes that one is an expert in the knowledge that they are seeking? You can tell people that it is just a jumping-off point for further research, but this is NOT how people actually work in the real world. People, myself included, are lazy, busy, scattered-brained, or under pressure from a tight deadline almost all of the time. How often do you, in the course of your day, have the kind of time I have devoted to debugging this AI overview? Almost never.

Perhaps this issue will one day be resolved, and the AI overview will be able to reliably, on command, define virtually any term on a moment’s notice, but this is clearly not the case in its current iteration. I believe we need to seriously ask the question: Is the proliferation of “AI” in its current state truly a net good for society? If it is an engine that is perpetuating and indeed creating misinformation then surely it cannot be. If this is the case, can it truly be ready for use by the general public and integration into virtually everything?