OpenAI introduces CriticGPT AI model

We are seeing the first practical benchmarks for AI vision: 1) A challenging real-life chart benchmark on chart reading Charxiv, shows humans get 80% right. Claude 3.5, the best LLM, gets 60% 2) Chatbot Arena compares which AI vision answers humans prefer. GPT-4o wins this one. pic.twitter.com/jeBgObA4kE

👇Astonishing! LLM-based AI that can’t reason and that is known for hallucinations and boneheaded errors needs a lot of “handholding” in business applications • Data from wrong year • Incorrect answers • Inconsistent answers • Doesn’t know which sources are gold standard pic.twitter.com/JlKs0yKHwR

OpenAI plans to use CriticGPT to help human trainers spot mistakes and improve ChatGPT, but this new tool has some limitations. www.siliconrepublic.com/machines/openai-criticgpt-chatgpt-ai-errors-hallucinations

OpenAI Builds AI to Critique AI CriticGPT is intended to help identify hallucinations as models grow more sophisticated spectrum.ieee.org/openai-rlhf