On Tuesday, OpenAI launched GPT-4, the fourth installment in the artificial intelligence (AI) lab’s GPT series.
Like its predecessor GPT-3, GPT-4 is capable of producing fluent and fairly human-like text in response to human prompts. However, OpenAI CEO Sam Altman has noted that GPT-4 is a multimodal model, which means it can also respond to image input, in addition to text-based input. Those interested in trying out GPT-4’s capabilities can join the API waitlist to access it via API or pay for Chatag/chatGPTtGPT Plus, a premium version of ChatGPT that utilizes GPT-4 (unlike the free version, which is based on GPT-3.5).
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.
— Sam Altman (@sama) March 14, 2023
“We’ve created GPT-4, the latest milestone in OpenAI’s effort in scaling up deep learning,” the company writes on its webpage for GPT-4. “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.”
OpenAI’s practically become a household name these days, thanks to the widespread popularity of — and scorn for — ChatGPT, a chatbot that can produce long, fluent strings of text in English and several other languages. OpenAI’s team notes that GPT-4 is significantly more capable than ChatGPT, pointing to the fact that in testing, GPT-4 was able to score in the top 10% of a bar exam, while ChatGPT scored in the bottom 10% on the same exam.
In the four months since it first launched, ChatGPT’s drawn a lot of criticism for its tendency to hallucinate facts, stating false information and presenting it as the truth. Although OpenAI hasn’t completely fixed this problem in GPT-4, the company notes that this version is much less likely to present false information as fact. According to OpenAI, GPT-4 boasts an accuracy rate of nearly 80% for most subjects, while the most up-to-date version of ChatGPT has an accuracy rate that hovers at around 60%.