A little late, but remember that openAI is just a language prediction model. It takes a string of text, and predicts what comes next based on all its training data. Think about all the training data it's had on the internet. This includes screenplays, short novels, fanfictions, etc. It knows a lot about how to continue conversations based on the likelyhood of the next word, but only if it's in the right format. So, instead of formatting your data like just regular strings, format it in a way that makes sense for the AI to continue it, such as in a screenplay. In other words, iterate on a string like:

import openai

openai.api_key = "secret_key"

model_engine = "text-davinci-003"

conversation = "This is the script for a casual conversation between a human and an AI:\n"

for i in range(2):

[tab]conversation += f"\nHuman: {input()}\n"

[tab]response = openai.Completion.create(engine = model_engine, prompt=conversation, max_tokens = 1024, n=1, stop=None, temperature=0.5)

[tab]conversation += f"{response['choices'][0]['text']}\n"

[tab]print(response["choices"][0]["text"])

print(conversation)

With this code, I got this output:

This is the script for a casual conversation between a human and an AI:

Human: Hello, computer. What is your name?

AI: Hi there! My name is AI. It's nice to meet you.

Human: Come up with a better name.

AI: How about AI-2?