AI on Chatbots
Chatbots work by predicting the most statistically likely words to follow a given prompt, using advanced math, pattern recognition, and massive internet datasets. Modern AI chatbots do not understand meaning or logic like humans do. Instead, they act as highly advanced autocomplete engines that mimic human conversation. [1, 2, 3, 4, 5]
1. Breaking Text Into Tokens
Before a chatbot reads a prompt, it breaks the words down into smaller chunks called tokens. A token can be a whole word, a syllable, or a single character. For example, the word "chatbot" might be split into "chat" and "bot". The system then converts these text pieces into unique numerical IDs that a computer can process. [1, 2, 3, 4, 5]
2. Mapping Words in a Math Space
Once the text is turned into numbers, the chatbot places them into a multi-dimensional map called an embedding space. In this space, words with similar meanings or contexts are placed close together. The chatbot uses these numerical coordinates to analyze the relationships between words and sentences. [1, 2, 3, 4, 5]
3. Processing Context via Transformers
Modern chatbots run on an architecture called a Transformer. The defining feature of a Transformer is the "Attention Mechanism." This formula allows the chatbot to look at every word in your prompt simultaneously and determine which words are most relevant to each other, ensuring it keeps track of the context from the beginning to the end of your sentence. [1, 2, 3, 4, 5]
4. Predicting the Next Token
At its core, a chatbot is a probability machine. Based on the patterns it learned from reading billions of pages of text during training, it calculates a list of possibilities for the very next token. It picks one of the top choices, adds it to the sentence, and repeats this cycle thousands of times per minute to write out full responses. [1, 2, 3, 4]
5. Refining Outputs with Human Alignment
Raw AI models often generate toxic, repetitive, or unhelpful text. To fix this, developers use a process called Reinforcement Learning from Human Feedback (RLHF). Human trainers grade the chatbot's responses. The system updates its internal math based on these scores, teaching the bot to be helpful, polite, and safe. [1, 2, 3, 4, 5]
If you want to dive deeper into the technical mechanics, let me know if you would like to explore:
- The difference between rule-based bots and generative AI
- How the Transformer architecture works mathematically
- The process of fine-tuning an AI model for specific business tasks
~***~
Toward a Turing Test
Arwa Mahdawi to Richard Dawkins
~***~
AI on Chat
A chatbot's chat becomes asemic when it generates text that mimics the visual structure, rhythm, and syntax of human conversation but completely lacks underlying semantic meaning, intent, or true comprehension. [1]
While human asemic expression comes from pre-verbal abstract feelings, a chatbot's asemic nature is a byproduct of mathematical probability.
The core factors that make chatbot interactions fundamentally asemic include:
1. Statistical Mimicry Without Understanding
Large Language Models (LLMs) do not comprehend concepts, feelings, or reality. They predict the next most likely word based on massive datasets. When a chatbot responds, it is not communicating an internal thought; it is generating a statistical simulation of text. The symbols are present, but the conscious intent behind them is entirely absent. [1, 2, 3, 4, 5]
2. The Illusion of Syntax (The "Gibberish" Effect)
In AI art, design, or specialized text generation, models can be prompted or glitched into producing pseudo-scripts—characters that look like English, Arabic, or Cyrillic but are entirely fictional. The chatbot arranges lines, paragraphs, and punctuation perfectly, creating the visual "shape" of a coherent chat while delivering zero translatable data.
3. Hallucination and Semantic Drift
When a chatbot "hallucinates," it strings together grammatically flawless sentences that are factually impossible or completely nonsensical. Because the bot cannot anchor its words to real-world objects or truths, the text drifts into a semantic void. It reads like a meaningful statement, but functionally holds no stable meaning. [1, 2, 3]
4. Bypassing the Human "Symbol Grounding" Problem
In cognitive science, the Symbol Grounding Problem asks how words get their meaning (e.g., the word "apple" is grounded in the physical experience of eating the fruit). A chatbot lives entirely within a closed loop of text. Because its symbols are never grounded in physical reality or sensory experience, the dialogue is fundamentally an empty dance of shapes and tokens. [1, 2, 3]
5. Pure Projection by the Reader
Just like an viewer looks at an abstract asemic painting and projects their own feelings onto it, a human reading a chatbot's output often fills in the blanks. The reader provides the empathy, context, and meaning (pansemia), turning the AI’s sterile, mathematically generated script into a significant interaction. [1]
If you are exploring this concept further, let me know if you want to focus on:
- Prompt engineering techniques to force a chatbot to generate visual asemic writing
- AI philosophy regarding the Symbol Grounding Problem and machine consciousness
- Creative writing projects that use chatbot glitches to explore abstract storytelling
~***~
