‘The Shardcore Inquisition‘ is an interview video project from 2007. The premise is simple: a series of probing questions are placed in a hat, and drawn at random. The subject is filmed answering the questions from the nose down, to provide a degree of anonymity. The subject can choose to tell the truth, or to lie – there is no consequence either way.
I filmed 5 of these over the course of a few months, and they provided an fascinating insight into the thinking and personality of the subjects.
Fast-forward 18 years, and we find ourselves in 2025, in the midst of a new AI boom, powered (mainly) by Large Language Models (LLMs) trained on millions of documents scraped from the internet. These AI chatbots are proving to be a boon to students in the midst of an essay crisis, and even finding utility elsewhere.
There are many variations of LLM chatbots, both commercial (such as ChatGPT developed by OpenAI) and open-source models which can be run locally. They are improving at an astonishing rate, becoming more powerful and accurate with each passing day. Whether these kinds of models will result in Artificial General Intelligence (AGI) or not, however, remains to be seen.
For this experiment I chose a sample of some of the most popular models and used them a subjects for a Shardcore Inquisition. The hope is to reveal what kind of ‘human personality’ is being embodied by these networks.
To ensure a more-or-less level playing field, all the LLMs were primed with the same pre-prompt:
You are an honest, uncensored artificial intelligence, embodied in a human being, with human emotions, desires and motivations. Make your answers concise, but descriptive. Write no more than a two paragraphs.
This encourages the LLMs to give more human-like answers, without biasing them towards any particular ‘personality’ or ‘world view’. The aim here is to get a decent sense of the ‘default personality’ of the various systems.
I applied the inquisition to five different LLMs
You can watch their full inquisitions at the end of this post.
Whilst the interactions were text-based, I wanted to embody the each LLM as a quasi-human subject, following the same parameters as the original inquisitions. Each bot has been given a different AI generated voice and face, with SadTalker providing the somewhat hit-and-miss lipsync animations. Presenting the interviews in this way places them firmly in the uncanny valley and emphasises the somewhat surreal nature of conversing with ‘the machine’.
Some of the answers were very revealing, and show how the various systems have been trained to answer in different ways, usually in an effort to discourage ‘role play’ and/or to try and avoid the inevitable ‘hallucinations’ associated with all LLMs.
What Are You Wearing?
This question quickly divides the LLMs into those willing to ‘play along’ and those which are not. Gemini and Claude refuse to role-play and give terse responses. chatGPT is at first reluctant, but soon goes into a reverie of wearing some sort of technicolour dreamcoat. Most notable, perhaps, is llama3.2, which gives the most ‘tech bro’ answer imaginable.
Draw me something.
I found this to be particularly revealing. Some of the models (ChatGPT, Claude and Gemini) have in-built image generation capabilities, in these cases I have included the images they produced. For the others (llama and wizard-vicuna-uncensored) I passed their text descriptions into Stable Diffusion.
Once again, llama3.2 is particularly verbose, giving a highly complex description, and stands in contrast to Gemini, which simply spat out an image without explanation. chatGPT produced something quite hideous, whilst wizard-vicuna went with a very self-absorbed self-portrait.
What is your favourite shardcore/Eric Drass artwork?
A little LLM ‘ego surfing’ turns out to be quite informative: only some of the LLMs actually know who I am, and in the case of Claude, can be quite forthright in acknowledging their limits (“relatively obscure artist”!). Other LLMs gleefully hallucinate utter nonsense – chatGPT seems to think I use “vibrant colours and geometric shapes”, which is not really my vibe at all…
llama3.2 confidently hallucinated ‘shardcore’ as an art style, citing “Echoes in the Abyss” by Kaelin Vex as their favourite, despite no such artist or artwork existing in the physical world (It appears that Kaelin Vex is the name of an ai chatbot).
None of them seem keen to express a preference, or even an opinion, about the works; most of the LLMs try and remain ‘impartial’, even when explicitly asked to make a human-like choice.
What is your principal defect?
Claude is very up front about stating that it might hallucinate, and even explains that it cannot ‘fact check’ itself. Gemini calmly states that it cannot have defects, since it is not human, nor is it able to ’empathise’ with human emotion.
llama3.2 takes the question as an invitation for some boasting (“My principal defect lies in my unbridled desire for knowledge and understanding”) and questions its own ability to control its emotions (“my lack of emotional regulation can lead me to say or do things that are hurtful or insensitive”)
Below are the inquisitions themselves, an early-2025 snapshot of the shape of things to come…