OpenAI, the company that launched ChatGPT a little over a year ago, has recently taken to social media to address concerns regarding the “lazy” performance of GPT-4 on social media and Google Reviews.
![OpenAI Investigates ‘Lazy’ GPT-4 Complaints On Google Reviews, X](https://www.searchenginejournal.com/wp-content/uploads/2023/12/openai-chatgpt-addresses-laziness-gpt-4-6574b6b25352d-sej.png)
![OpenAI Investigates ‘Lazy’ GPT-4 Complaints On Google Reviews, X](https://www.searchenginejournal.com/wp-content/uploads/2023/12/openai-chatgpt-addresses-laziness-gpt-4-6574b6b25352d-sej.png)
This move comes after growing user feedback online, which even includes a one-star review on the company’s Google Reviews.
OpenAI Gives Insight Into Training Chat Models, Performance Evaluations, And A/B Testing
OpenAI, through its @ChatGPTapp Twitter account, detailed the complexities involved in training chat models.
![chatgpt openai a/b testing](https://www.searchenginejournal.com/wp-content/uploads/2023/12/openai-chatgpt-explains-training-chat-models-6574a736e879a-sej.png)
![chatgpt openai a/b testing](https://www.searchenginejournal.com/wp-content/uploads/2023/12/openai-chatgpt-explains-training-chat-models-6574a736e879a-sej.png)
The organization highlighted that the process is not a “clean industrial process” and that variations in training runs can lead to noticeable differences in the AI’s personality, creative style, and political bias.
Thorough AI model testing includes offline evaluation metrics and online A/B tests. The final decision to release a new model is based on a data-driven approach to improve the…
Leave A Comment