OpenAi, parent of ChatGPT, released a new tool aiming to make ChatGPT more reliable in its answers. Hello CriticGPT.

With this seemingly becoming “self-aware”, I can only imagine that this is what AGI’s (Artificial General Intelligence) early development looks like.

-Admin

CriticGPT, developed by OpenAI, is an advanced AI model designed to enhance the reliability of AI-generated content by assisting human reviewers in detecting and critiquing errors in code produced by ChatGPT. This model, part of the GPT-4 family, aims to address the increasing complexity of evaluating sophisticated AI outputs as large language models evolve.

Training and Performance CriticGPT’s training involved a dataset with intentionally inserted bugs, allowing the model to effectively recognize and flag various coding errors. This approach led to remarkable results, with CriticGPT catching about 85% of bugs compared to the 25% identified by human reviewers. Additionally, its feedback was preferred over human critiques in 63% of cases involving natural language model (LLM) errors, showcasing its superior performance in error detection. To further enhance its capabilities, researchers developed the Force Sampling Beam Search (FSBS) technique, which improved CriticGPT’s ability to provide detailed code reviews while minimizing false positives.

Applications and Limitations While CriticGPT is primarily focused on code review, it has also shown potential in identifying errors in non-code tasks, highlighting its versatility in improving AI outputs. However, the model’s effectiveness diminishes with longer and more complex tasks, as it was trained on relatively short responses. Despite its impressive performance, CriticGPT still produces some false positives and requires human oversight to ensure accuracy. Additionally, the model struggles with detecting errors spread across multiple code strings, making it difficult to identify the source of certain AI hallucinations.

Future Integration Plans OpenAI plans to integrate CriticGPT into its Reinforcement Learning from Human Feedback (RLHF) pipeline, providing human trainers with an AI assistant to help review and refine generative AI outputs. This integration aims to enhance the overall quality and alignment of AI systems with human expectations. By leveraging CriticGPT’s capabilities, OpenAI anticipates improving the efficiency and accuracy of their AI training processes, potentially leading to more reliable and sophisticated AI models in the future.

Overall, CriticGPT represents a significant advancement in AI error detection and quality assurance, offering valuable support in the continuous improvement of AI-generated content.

Reprinted with Permission from Discovr.Ai