With its significant debut, OpenAI’s new GPT-4 AI model is now powering everything from a virtual volunteer for the blind to an improved Duolingo language learning bot. What, therefore, sets GPT-4 apart from preceding iterations like ChatGPT and GPT-3.5? The top 5 ways GPT-4 can outsmart ChatGPT from OpenAI are listed below.
GPT-4 Is Able to See and Comprehend Pictures
GPT-4 can be given images to analyse to find relevant information. Of course, you could just ask it to describe what’s in the picture, but it’s more important to note that its comprehension goes farther than that. The partnership with Be My Eyes, a blind and low-vision software that allows participants to describe what they see using a phone, is more significant, though. In the illustration provided by OpenAI, it is shown describing a joke in an image of an absurdly large iPhone connection.
It’s Tough to Fool GPT-4
During the past year or two, users have graciously given OpenAI malicious instructions that were utilised to train GPT-4. The new model outperforms its predecessors in terms of “factuality, steerability, and refusing to go outside of guardrails” when these considerations are considered.
The Memory is Longer in GPT-4
The number 32,768 is the maximum token count for GPT-4; if you’re wondering why that seems familiar, it’s because it’s 215. This is roughly 64,000 words or 50 pages of text, which is enough for a complete play or brief story.
GPT-4 Has More Languages
GPT-4 takes a step in that direction by demonstrating that it can correctly respond to thousands of multiple-choice questions in 26 different languages, including Italian, Ukrainian, and Korean. It performs best in the Romance and Germanic languages, but it also generalises effectively to other languages.
GPT-4 Displays Many “Personalities”
With GPT-4, users will have the option to change the “base ChatGPT personality with a set verbosity, tone, and style” to one that is more appropriate for their needs. As comparison to GPT-3.5, GPT-4 incorporates steerability more organically. The crew quickly adds, “Within restrictions,” thinking that this would be the easiest way to get the model to act out of character.
