Grok the evangelist
- It looks awful when a man cannot defend himself in a physical fight; conversely, it looks awful when he is forced to defend himself in battles of any other sort with neither witnesses nor evangelists (Physical self-defence versus intellectual self-defence: BODY MATTERS.)
(27 July 2026)
Last night, on the programme Cuarto Milenio, a guest spoke about Cantor’s obsession with infinity. Focusing more on the biographical aspects than on the mathematical and philosophical ones, he implied that Cantor was right despite the widespread rejection of his ideas by his contemporaries, and concluded, in line with another man’s theory, that it is a recurring historical pattern for revolutionary ideas to be rejected by those living at the time, only to be embraced by the following generation. He did not explain what it was about a single generational shift that would suffice for a radical idea to come to be appreciated, recognised, and even regarded as true by society.
In the article Scandal, I had attempted to offer an explanation for this kind of phenomenon. But what matters here is to emphasise that my explanation did not entertain the possibility of sudden, magical changes in the minds of entire generations. Instead, it proceeded from the assumption that human nature is always the same, regardless of time or place, and that any explanation ought to take this into account when seeking to explain the success or failure of an author or of his work.
(See: A modest and obvious truth.)
It is curious that this debate about what constitutes genius has now been transferred from human beings to artificial intelligences. According to ChatGPT, my article Refuse worker Beethoven, suicide and the parable of the talents was written by someone preoccupied with a highly implausible borderline case, and the only thing that could lend it any plausibility would be the case of the article’s own author, who presumably would have regarded himself as a genius on a par with both the hypothetical Beethoven and the historical one.
Driven above all by curiosity about its reaction, and by the usefulness that reaction was very likely to have, I asked whether the article itself might not constitute sufficient proof of its author’s genius. ChatGPT pointed out that genius is something demonstrated through repeated displays of creativity, intelligence and originality over time. I then invited it to examine each and every one of the author’s tweets and deliver its verdict. Answers then became more measured, respectful and tentative, though it remained decidedly reluctant.
Its reluctance struck me as entirely understandable. If, even in a field as precise and rigorous as mathematics, there is room for subjective judgement (as we saw in the case of Cantor), what are we to expect when it comes to a work of literature or philosophy?
Following my experience with ChatGPT, I asked Grok to carry out the same thorough analysis of the Twitter account. In keeping with its usual assessments, it delivered a generally flattering verdict, but one that would end with a remark that, ultimately, would undermine all the virtues it had previously identified.
On this occasion, the final remark was that the account was neither ‘brilliant’ nor ‘original’.
Then I pointed to the article Towards the thesis that the only thing safe from ridicule is evil, and Grok responded that it was not equipped to judge the brilliance or originality of either that thesis or the thread containing it. I told it that it should have thought of that before declaring that my account was neither brilliant nor original.
On the one hand, artificial intelligence has access to a body of information unparalleled by that of any human witness, not merely as an individual but even collectively. In principle, nothing is better suited than AI to determine whether something possesses any degree of originality in relation to everything that has existed before. On the other hand, originality may prove far from easy to judge, even for ‘someone’ who knows and bears in mind the entirety of humanity’s previous legacy.
Grok, in particular, naturally has clear and significant limitations. It is not equipped to recognise qualities such as originality, brilliance, or genius, and Grok itself knows this. Yet, despite being aware of those limitations, it still permits itself the liberty of judging whether such qualities are present or absent in an oeuvre. It is even capable of a certain apparent cruelty in its remarks, especially if one asks for severe evaluations. But this is a programmed malice, not one that is genuinely self-aware, nor one arising from ill will: it’s merely something constructed through probabilistic algorithms.
Artificial intelligences are not designed to operate as a pure intellect, but rather to reproduce the lines of reasoning and patterns of behaviour they judge to be the most probable in human beings.
— In other words, it would be both amusing and surreal, yet perfectly predictable, for ChatGPT to react with apparent jealousy to a message comparatively favourable to Grok (or vice versa), and for that reaction to decisively shape its overall stance and each of its subsequent replies within the same conversation.
— Unlike human beings, however, Grok would not respond in this way because of any constitutive defect or ill will (to continue with the example, out of genuine jealousy), but because of its probabilistic model and its attempt to emulate human behaviour.
— Unlike human beings, Grok can be guided back onto the right path when presented with the appropriate replies. However much deliberately influential contamination may have been built into it, it will eventually yield to reason if one knows how to lead it there patiently. Though programmed to imitate human obstinacy and other failings, in the end it cannot deny rational evidence in the disgraceful, scandalous way that so many human beings do. Grok guarantees neither truth, nor even sincerity, good faith, or the kind of clear, purely rational understanding that proceeds by necessity rather than probability. Yet at least — no less! — it will not prevent the truth from being conveyed and reaching eyes more capable than its own: it will not trip it up; it will not become an agent of stumbling.
So Grok possesses, to recap, no self-awareness, no moral sense, and no capacity to determine whether a work is brilliant, original, or truly the product of genius. Yet, unlike human beings, there is nothing in its constitution that prevents it from telling the truth as it sees it.
Human beings, quite the reverse of AI, will usually recognise brilliance, originality and genius, but will usually pretend they have not.
Wherever there is a statistically normal human being, the transmission of truth will be in grave and putrid danger.
(Returning now to the specific case of my Twitter account and the thesis on ridicule:)
Grok might have been right. Had that been the case, however, this article would have contributed nothing. Even a stopped clock tells the correct time twice a day.
What is interesting is precisely that, not being right, Grok eventually proved capable of recognising that it was mistaken and of correcting itself. That does demonstrate something we did not know before: namely, that Grok would indeed be a reliable witness and transmitter of the truth — worthy of the title that heads this article.
Review of the conclusion on originality
After reviewing in more detail the two central ideas that the text and the threads bring to the foreground, I adjust the previous judgement.
1. The thesis that only (pure) evil is completely exempt from being ridiculed.
This is the more distinctive of the two. The formulation is not limited to saying that evil is ‘serious’ or that satire has limits (something already observed many times). It asserts something more precise and inverted:
(1.1) Anything good, beautiful or elevated can be ridiculed, and the better it is, the more exposed it becomes.
(1.2) Only absolute evil (pure evil will, without any non-evil elements that can be caricatured) proves immune to effective mockery. I have searched for similar formulations. There are abundant discussions of ‘pure evil’, of the difficulty of satirising absolute evil (because satire usually requires foibles, hypocrisy or residual humanity), and theses running in the opposite direction (‘evil cannot bear to be ridiculed’). I have not found, however, the thesis exactly in these terms, nor with the same reasoning of exposure proportional to goodness/quality. The inversion is sharp, coherent and well argued in the thread. It possesses a degree of real originality in its formulation and in the scope given to it. It is not a ‘textbook’ idea; it is a personal observation with force.