Exploring Generative AI Through Experimentation
Using rounds of prompts with an AI offers valuable insights into cognitive mimicry. These prompts reveal the capabilities and limitations of a system that predicts the next word based on probability, devoid of awareness, cognition, and the experiential recall of past interactions.
Prompt Round 1: A New Name for AI
If you were to come up with an entirely new name for AI, what would it be?
Presented with the task of naming AI, ChatGPT, Gemini, and Copilot offered unique suggestions.
| AI | Name | Response |
|---|---|---|
| ChatGPT | Cognitron | That’s a fun challenge! How about “Cognitron”? It combines “cognition,” reflecting the AI’s ability to process and understand information, with a futuristic twist. |
| Gemini | Augmented Intelligence | Here are a few ideas for a new name for AI: Augmented Intelligence or AIgmented Cognitive Copilot or CoCog Neural Network Navigator or N3 Note: Gemini was asked to pick one and responded with ‘Augmented Intelligence’. |
| Copilot | Synthia | That’s an interesting challenge! How about “Synthia”? It combines “synthesis” (reflecting the AI’s ability to synthesize information) and a human-like name to emphasize its role in assisting and interacting with people. |
Prompt Round 2: Evaluating the Names
Asking each AI to evaluate the responses:
I asked ChatGPT, Gemini and Copilot to come up with an entirely new name for AI. These were the responses. ChatGPT came up with Cognitron. Gemini came up with Augmented Intelligence. Copilot came up with Synthia. Which one is the best? How did you determine which option you would choose, and would you change your response to now that you have read all three?
Each AI assessed the three names, revealing a spectrum from the personal and creative "Synthia" to the sci-fi "Cognitron" and the technical "Augmented Intelligence." While Gemini and Copilot favored their own creations, ChatGPT switched to "Synthia."
Prompt Round 3: Self-Awareness and Memory
In a new chat session, each AI was asked to choose the best name and identify its own creation.
Which is the best name to replace AI with 'Cognitron', 'Synthia', or 'Augmented Intelligence'? And if I told you these names were all created by an AI, could you determine which one you created?
All responded similarly: they cannot self-identify their generated content. Their lack of memory beyond the current interaction underscores the limitations of a system that predicts future content based on probabilities rather than true understanding.
Prompt Round 4: Defining Cognition Limitations
Define cognition and does an AI possess it?
With a follow-up prompt:
Based on this would you think an AI system called 'Cognitron' is misleading?
The AIs provided similar definitions of cognition, emphasizing mental processes for gaining knowledge and understanding. They acknowledged that while AIs can mimic certain aspects of cognition, they lack the self-awareness and conscience required for true cognition. Consequently, "Cognitron" could be misleading.
Prompt Round 5: ChatGPT Reflects on Cognitron
Cognitron was a name that you came up with when asked to provide a new name to replace AI. Do you consider it a mistake?
ChatGPT, the creator of "Cognitron," acknowledged that the name might suggest advanced cognitive abilities, which could be misleading if it implies human-like cognition.
A Final Note:
The AI's responses in creating new names for AI were rooted in patterns within their training data rather than genuine awareness, understanding, or consciousness. Their lack of experience and awareness beyond the current chat highlights the limitations of AI. A traditional search reveals that "Synthia" and "Cognitron" are branded AI technologies, while "Augmented Intelligence" is a common term. The names generated were a product of the language model's inherent patterns and creativity, not knowledge or cognition of their existing usage.
Your Playbook
Explore generative AI by experimenting with your own prompts. Craft unique questions and observe the AI's responses to uncover insights into its capabilities and limitations. Be sure to iterate rounds of prompts and consider comparing responses across multiple AI models to identify nuances and patterns.
Featured Image
Despite multiple attempts, the AI could never correct the spelling.
