Mamdani's Parrot

Photo by kevin.muellr on Unsplash

The title of this article is of two worlds, one of animal kingdom and the other of foundations in Machine Learning. This article is not a tutorial or walk-through into either one of those subjects. I would like to share my thoughts on my observations of what I’ve come to see of human nature emergent from latest developments.

Mamdani is one of the geniuses behind a foundational algorithms that we use for intelligent systems called Mamdani-Sugeno Fuzzy Inference, part of fuzzy logic, often found in modern refrigerators and washing machines.

A parrot is a bird, often green but it does come in multiple colours.

I want you to imagine a parrot, Parrot extremely good at Parroting. Meaning mimic human words and sounds. Sometimes at a near perfection. Now imagine through selective breeding or genetic engineering we manage to create a parrot with near perfect memory and near perfect mimicry. We’ll call this the Super Parrot. Imagine if this super parrot was to memorize entire scripture, whether its the Bible, Quran, Bhagavad Gita or Dhammapada, and now the parrot is able to recite verses at near perfection. Does the parrot now qualify to be a Priest, Imam, a Guru or a Monk?

Is mimicry intelligence? as we established in the thought experiment, we would need the most parrot with highest cognitive capacity to pull this off. But does the parrot gain equal footing in human society? As per the cambridge dictionary Parroting is the act of repeating without true comprehension. If this super parrot has no comprehension we can write it off from giving voting rights in our society. But at this point how would we prove that the Parrot is truly Parroting and not a divine intellect?

At the present moment only way to prove or test this as per my knowledge is the Turing test. Unfortunately like with all tests lot of humans fail this too 1.

But lets assume the parrot passes the Turing test. Does it now qualify to be a human? If we were to give it voting rights, would it be ethical?

This is the point where we humans are at, we do have a Parrot and its not biological. Its the LLM’s a.k.a Large Language Models. There’s many parrots and these are very power hungry parrots (ChatGPT, Claude, Mistral, LLAMA etc.. )

Unlike parrots we humans trust the tokens generated from these systems as literal truth. If a parrot were to repeat the things these LLm’s were spewing you wouldn’t take it seriously. But here we are. What compels us to trust these more freely is surprising as well. There were many times I pondered these factors, perhaps its the word choices, or the grammar or tonality of the words. Whatever that may be its managed to fool most of us into believing them.

LLM’s are far superior to Mamdani-Sugeno Inference, yet its not really Artificial Intelligence. The presence or absence of a language model is not a good indication of intelligence. Would a deaf / mute person be considered less intelligent? Its not wrong to call LLM’s a type of Lossy Super compression of data. There’s research to back this up. There’s plenty of resources to understand how LLm’s work, but in summary its merely approximating what word to say at each rotation. It cannot say things that it wasn’t trained on, these systems are extremely good at connecting the dots of our existing knowledge, but let it come up with new information it’ll generate outright slop. But there’s a bigger danger.

As an engineer I believe a simple truth, machines are machines, and machines can break. But this is not the common knowledge to the masses. Some people and governments have begun to accept Ai as ultimate truth. Sometimes punishing people who’s yet to commit a crime but due to potential they might have they get flagged early on 2 3. The authorities will say the system reported so and so and will not question the result. I’ve seen entire accounts get auto banned and getting locked out from Microsoft and AWS alike just because a resource was downloaded slightly differently or using a different browser than the usual [].

Now forget Ai / LLM for a moment, imagine people trusting a Super Parrot without questioning, imagine an entire government process being run by Super Parrots. Imagine being falsely accused of a crime because the Parrot didn’t like your skin colour or your race.

We’ve dug ourselves a hole where we’re becoming slaves to Mamdani’s Parrots. Its ultimately your choice, use LLM’s or any Ai as a tool, not as a false Idol. Or make it your false prophet.

References and Inspirations

https://dictionary.cambridge.org/dictionary/english/parroting

https://www.mathworks.com/help/fuzzy/types-of-fuzzy-inference-systems.html

https://arxiv.org/abs/2505.09598

https://arxiv.org/abs/2309.10668

https://theconversation.com/why-microsofts-copilot-ai-falsely-accused-court-reporter-of-crimes-he-covered-237685

https://gulfnews.com/world/americas/false-alarms-and-arrests-the-controversial-role-of-ai-in-school-safety-1.500225810

https://niklashageback.com/2020/03/23/why-so-many-humans-would-fail-the-turing-test/