OpenAI’s ecently released GPT-4 improves upon its predecessor, GPT-3, in key ways—for example giving more factually true statements and allowing developers to prescribe its style and behavior more easily. It’s also multimodal in the sense that it can understand images, allowing it to caption and even explain in detail the contents of a photo. But GPT-4 has serious shortcomings. The model “hallucinates” facts and makes basic reasoning errors.
Asked to compare GPT-4 to GPT-3, OpenAI co-founder and president Greg Brockman had one word: Different. “It’s just different. There’s still a lot of problems and mistakes that [the model] makes … but you can really see the jump in skill in things like calculus or law, where it went from being really bad at certain domains to actually quite good relative to humans.”
View Full Article
No entries found