'Human-level performance': ChatGPT creator OpenAI releases smarter, faster GPT-4 AI

OpenAI has officially announced the release of its latest AI language model, GPT-4. The company claims that this model is “more creative and collaborative than ever before” and is capable of solving difficult problems with greater accuracy than its predecessors. GPT-4 can parse both text and image input, although it can only respond via text.

It is important to note that GPT-4 retains many of the same problems as earlier language models, including the ability to make up information (or “hallucinate”). OpenAI warns that the model still lacks knowledge about events that have occurred after September 2021.

"It is still flawed, still limited, and it still seems more impressive on first use than it does after you spend more time with it," OpenAI CEO Sam Altman wrote on Twitter while announcing GPT-4.

While OpenAI on their website said, “GPT-4 is a large multimodal model (accepting image and text inputs, emitting text outputs) that, while less capable than humans in many real-world scenarios, exhibits human-level performance on various professional and academic benchmarks.”

Despite these limitations, OpenAI has already partnered with several companies to integrate GPT-4 into their products, including Duolingo, Stripe, and Khan Academy. The latest model can be accessed by people who have ChatGPT Plus, OpenAI’s $20 monthly ChatGPT subscription. It is also powering Microsoft’s Bing chatbot and will be accessible as an API for developers to build on.

OpenAI suggests that the difference between GPT-4 and its predecessor GPT-3.5 is not easily noticeable in everyday conversation. The company claims that GPT-4's enhancements are apparent in its performance on various tests and benchmarks such as the Uniform Bar Exam, LSAT, SAT Math, and SAT Evidence-Based Reading & Writing exams. GPT-4 scored 88th percentile or higher on these exams.

GPT-4 is indeed multimodal, but it can only accept text and image inputs and emit text outputs. However, the model’s ability to parse text and images simultaneously allows it to interpret more complex input.

Also read

Instagram now lets users post GIFs in comments, here's how it works

Elon Musk says his Twitter posts were not boosted, plans to take legal action against employee

Watch: Salil Parekh, Suresh Narayanan, and more: What India's top CEOs studied, earn