Following the widespread adoption of its well-liked artificial intelligence (AI) chatbot ChatGPT, Microsoft-backed startup OpenAI has begun rolling out a new potent AI model GPT-4. GPT-4, a human-like technology that is sometimes referred to as ChatGPT’s successor, is more imaginative and collaborative.
On creative and technical writing tasks, such as writing songs, screenplays, or figuring out a user’s writing style, GPT-4 can generate, edit, and iterate with users.
“We’ve built GPT-4, the next milestone in OpenAI’s endeavor in scaling up deep learning. GPT-4 is a big multimodal model (accepting picture and text inputs, outputting text outputs) that, although less competent than humans in many real-world settings, demonstrates human-level performance on several professional and academic benchmarks,” OpenAI noted in a blog post.
“We’ve spent 6 months iteratively aligning GPT-4 using lessons from our adversarial testing program as well as ChatGPT, resulting in our best-ever results on factuality, steerability, and refusing to go outside of guardrails,” the company said.
The business is making text input for GPT-4 available through ChatGPT and the API. In a casual conversation, the distinction between GPT-3.5 and GPT-4 can be subtle, OpenAI said.
When the task’s complexity reaches a certain level, the difference becomes apparent: GPT-4 is more dependable, inventive, and able to handle far more sophisticated instructions than GPT-3.5.
It claimed that in some instances, the most recent technology from OpenAI represented a significant upgrade over the GPT-3.5 predecessor.
According to OpenAI, the new model performed in the top 10% of test takers in a simulation of the bar exam, which US law school graduates must pass before engaging in professional practice.
Although in casual conversation the two versions may seem identical, “the difference comes out when the intricacy of the job approaches a significant threshold,” according to OpenAI, which noted that “GPT-4 is more dependable, creative, and able to handle far more subtle instructions.”
