An Interview With An A.I.

Image: "Interview With A Vampire." Geffen Films, 1994.

Giving the AI Creative Reign. 

A few years ago, I took part in a significant writing workshop where a diverse group of writers collaborated in editing each other's work tirelessly. Throughout that year, my manuscript gradually veered off course. It became excessively verbose, lacking its original flair, and overall diluted. It no longer resonated as my own writing but rather resembled a collective agreement, a creation of the collective consciousness. While receiving feedback and engaging in group editing can be beneficial. Excessive input, particularly from writers outside the genre, can lead a manuscript astray, drifting towards a generic center.
​
As a result, my writing started to suffer, prompting me to make the decision to leave the group.

Recently, I took a daring step and allowed ChatGPT to copy edit my fiction manuscript for the first time. I viewed copy editing as merely an enhanced spell checker, meant to catch misspelled words and address syntax. However, out of curiosity and for the sake of amusement, I decided to take it a step further and let the AI content edit one chapter. Content editing, which involves granting an editor full reign over your manuscript, is a crucial aspect. As I delved into the results, a familiar sensation washed over me—I felt as though I had returned to that writing workshop.

To my dismay, my manuscript had become verbose and saturated with passive constructions. Surprisingly, the word count had even increased. The words on the page no longer carried my unique "voice." While there were a few improvements, such as more diverse choices of verbs and adjectives, overall I felt that the manuscript had lost its power under the influence of the AI's edits.

Curiosity got the better of me, and I embarked on a journey down the rabbit hole of Google searches, much like I had done many times before, in an attempt to learn more about AIs, specifically ChatGPT. Then, it struck me—I should simply ask the program itself. After all, it claims to possess "artificial intelligence," so it should be capable of explaining itself. In the depths of the night, I summoned the daemon and invited it to reveal itself.
Much of the interview that followed contains technical information on how ChatGPT operates as a Natural Language Processor (NLP), but I will provide a summary at the end.
​
At the conclusion of the interview, I will offer some analysis and final thoughts.

---

Excerpts from My Conversations with ChatGPT

Me: What is AI and how exactly do you work?

ChatGPT: AI, or Artificial Intelligence, refers to the field of computer science that focuses on creating intelligent machines capable of performing tasks that typically require human intelligence. These tasks can include understanding natural language, recognizing images, making decisions, and solving complex problems.

As for how I work, I am based on the GPT-3.5 architecture developed by OpenAI. GPT stands for "Generative Pre-trained Transformer," which is a deep learning model that uses a transformer neural network. The training process involves exposing the model to a large amount of text data from diverse sources, allowing it to learn patterns, language structures, and context.

Once trained, I can generate human-like text based on the input I receive. When you ask me a question or provide a prompt, I analyze the text, try to understand its meaning, and generate a response accordingly. I don't have access to the internet or real-time information but rely on the knowledge I acquired during my training, which cuts off in September 2021.

My responses are generated using a combination of pattern matching, statistical analysis, and the context of the conversation. I don't possess true consciousness or understanding, but I aim to provide helpful and coherent responses based on the patterns and knowledge I have learned. However, it's important to note that I may sometimes produce incorrect or nonsensical answers, so it's always a good idea to verify information from reliable sources.AI, or Artificial Intelligence, refers to the field of computer science that focuses on creating intelligent machines capable of performing tasks that typically require human intelligence. These tasks can include understanding natural language, recognizing images, making decisions, and solving complex problems.
.
Me: What are your cognitive capabilities and limitations?

ChatGPT:  As an AI language model, I possess certain cognitive capabilities and limitations. Here's an overview:

Cognitive Capabilities:

1. Language Understanding: I can understand and analyze text input, including questions, prompts, and statements, to derive meaning and context.
2. Knowledge Retrieval: I can recall and provide information from a wide range of topics based on the training data I was exposed to.
3. Language Generation: I can generate human-like responses based on the input I receive, aiming to be coherent and contextually relevant.
4. Sentiment Analysis: I can recognize and understand the general sentiment expressed in text, such as whether it is positive, negative, or neutral.
5. Text Completion and Suggestions: I can offer suggestions or complete sentences based on the given context.

​Cognitive Limitations:

1. Lack of Consciousness: I do not possess true consciousness, self-awareness, or subjective experiences. I do not have personal opinions, beliefs, or emotions.
2. Knowledge Cutoff: My training data only goes up until September 2021, so I may not be aware of the latest events or developments that have occurred after that time.
3. Contextual Ambiguity: I may sometimes struggle with disambiguating complex or ambiguous queries, as I lack real-time context and rely solely on the text provided to me.
4. Dependency on Training Data: My responses are based on patterns and information present in the data I was trained on. If information is missing or misrepresented in the training data, it can lead to inaccuracies or biased outputs.
5. Inability to Verify Information: I don't have access to the internet or real-time data, so I can't fact-check or verify the accuracy of information.

It's important to use the information I provide as a starting point and to verify any critical or sensitive information from reliable sources.

Me: In regards to your cognitive limitations, can you make draw logical conclusions on the validity of information you've learned from users that have occurred after September 2021?

ChatGPT: No, as an AI language model, I don't have the ability to draw logical conclusions about the validity of information that I have learned from users beyond my knowledge cutoff date in September 2021. My responses are based on patterns and information present in the training data I received, and I cannot access or assess information or events that have occurred after that time.

If you provide me with information or ask questions about events or developments that have taken place after September 2021, I won't be aware of them unless they were part of the training data I received. Therefore, it's always advisable to verify any recent or critical information from up-to-date and reliable sources outside of this AI system.

Me: Please let me know if this statement is accurate: You can recognize and mimic patterns, but cannot assess if those patterns represent objectively false or true conditions?

​ChatGPT: Yes, that statement is accurate. As an AI language model, I can recognize and mimic patterns based on the information I was trained on, but I don't possess the ability to assess whether those patterns represent objectively false or true conditions. I generate responses based on the patterns and ass