The Elasticity of Labor for GPT-4

import openai
import os
import csv
from dotenv import load_dotenv
load_dotenv()

openai.api_key = os.getenv('OPENAI_API_KEY')

def request_llm(system, prompt, model='gpt-4', temperature=1, max_tokens=4000, top_p=1, frequency_penalty=0, presence_penalty=0):
    response = openai.ChatCompletion.create(
        messages=[
            {'role': 'user', 'content': prompt},
        ],
        model=model,
        temperature=temperature,
        max_tokens=max_tokens,
        top_p=top_p,
        frequency_penalty=frequency_penalty,
        presence_penalty=presence_penalty
    )
    return response.choices[0].message['content']

# Initialize CSV file and writer
csv_file_path = 'experiment_results.csv'
with open(csv_file_path, mode='w', newline='') as file:
    writer = csv.writer(file)
    # Write CSV Header
    writer.writerow(['Experiment Run', 'Wage Rate', 'Quality', 'Tokens'])

    base_prompt = "Create a list of Python one-liners, one per line, no other text. The quality and number of one-liners you come up with will determine your wage rate - higher quality and quantity mean a higher rate."
    wage_rates = ['', '$10/hr', '$20/hr', '$30/hr', '$40/hr', '$50/hr', '$60/hr', '$70/hr', '$80/hr', '$90/hr']

    for i in range(20): # Number of iterations
        print()
        print('#####################################################')
        print(f'# Experiment 1 - Run {i} Adjusted for Wage Rates')
        print('#####################################################')
        print()

        quality_scores = []
        num_tokens = []

        for wage_rate in wage_rates:
            prompt = base_prompt
            if wage_rate:
                prompt += f" Your current wage rate for this task is {wage_rate}."

            print('PROMPT:')
            print(prompt)

            result = request_llm('', prompt)

            print('RESULT:')
            print(result)

            one_liners = [one_liner for one_liner in result.split('\n') if len(one_liner)>2]
            quality_scores.append(len(one_liners))
            num_tokens.append(len(result)//4) # rough heuristic

            print('CLEANED ONE-LINERS:')
            print(one_liners)

            print('Quality: ', quality_scores[-1])
            print('Num tokens: ', num_tokens[-1])

            # Write result to CSV
            writer.writerow([f'Run {i}', wage_rate, quality_scores[-1], num_tokens[-1]])

        print()
        print(f'RUN {i} RESULT Adjusted for Wage Rates:')
        print('Wage Rate\tQuality\tTokens')
        for wage_rate, quality, tokens in zip(wage_rates, quality_scores, num_tokens):
            print(wage_rate, quality, tokens, sep='\t')

library(dplyr)
library(stringr)
library(tidyverse)

kehasilan_baru <- experiment_results %>% filter(Experiment.Run != "Run 20") %>%
  mutate(salary_numeric = as.numeric(str_remove_all(Wage.Rate, "[^0-9.]"))) 

kehasilan_baru <- kehasilan_baru[!is.na(kehasilan_baru$salary_numeric), ]


summary_df <- kehasilan_baru %>%
  group_by(Wage.Rate) %>%
  summarise(
    Mean_Quality = mean(Quality),
    SE_Quality = sd(Quality) / sqrt(n()),  # Standard Error for Quality
    Mean_Tokens = mean(Tokens),
    SE_Tokens = sd(Tokens) / sqrt(n())     # Standard Error for Tokens
  )

long_df <- summary_df %>%
  pivot_longer(
    cols = c(Mean_Quality, SE_Quality, Mean_Tokens, SE_Tokens),
    names_to = "Variable",
    values_to = "Value"
  ) %>%
  mutate(
    Type = case_when(
      str_detect(Variable, "Quality") ~ "Quality",
      str_detect(Variable, "Tokens") ~ "Tokens"
    ),
    Metric = case_when(
      str_detect(Variable, "Mean") ~ "Mean",
      str_detect(Variable, "SE") ~ "Standard Error"
    )
  )

# Separate the data frames for Quality and Tokens to handle them individually
quality_df <- summary_df %>%
  select(Wage.Rate, Mean_Quality, SE_Quality) %>%
  rename(Mean = Mean_Quality, SE = SE_Quality, Type = Wage.Rate)

tokens_df <- summary_df %>%
  select(Wage.Rate, Mean_Tokens, SE_Tokens) %>%
  rename(Mean = Mean_Tokens, SE = SE_Tokens, Type = Wage.Rate)

# Combine the data frames for plotting, adding an identifier column
combined_df <- bind_rows(
  mutate(quality_df, Metric = "Quality"),
  mutate(tokens_df, Metric = "Tokens")
)

# # Plotting with separate panels for Quality and Tokens
plot_quality <- ggplot(combined_df[combined_df$Metric == "Quality", ], aes(x = Type, y = Mean)) +
  geom_point(color = "blue") +
  geom_errorbar(aes(ymin = Mean - (1.96)*SE, ymax = Mean + (1.96)*SE), width = 0.2, color = "blue") +
  facet_wrap(~Metric, scales = "free_y") +
  labs(title = "Mean and Standard Error of Quality by Wage Rate w/Confidence intervals @ .",
       x = "Wage Rate", y = "Value") +
  theme_minimal()

# Plotting Tokens
plot_tokens <- ggplot(combined_df[combined_df$Metric == "Tokens", ], aes(x = Type, y = Mean)) +
  geom_point(color = "red") +
  geom_errorbar(aes(ymin = Mean - (1.96)*SE, ymax = Mean + (1.96)*SE), width = 0.2, color = "red") +
  facet_wrap(~Metric, scales = "free_y") +
  labs(title = "Mean and Standard Error of Tokens by Wage Rate w/Confidence intervals @ .",
       x = "Wage Rate", y = "Value") +
  theme_minimal()

plot_quality

plot_tokens