GPT-4 is available today via OpenAI’s API with a waitlist and in ChatGPT Plus, OpenAI’s premium subscription for ChatGPT, its AI-powered chatbot. According to OpenAI, GPT-4 can accept image and text inputs — an improvement over GPT-3.5, its predecessor, which only accepted text — and performs “human-level performance” on various professional and academic benchmarks. For example, GPT-3 passes a simulated bar exam with a score around the top 10% of test takers. OpenAI spent six months iteratively aligning GPT-4 using lessons from an adversarial testing program as well as ChatGPT, resulting in “best-ever results” on factuality, steerability and refusing to go outside of guardrails. OpenAI releases GPT-4, a multimodal AI that it claims is state-of-the-art by Kyle Wiggers originally published on TechCrunch
