A long-awaited update to the artificial intelligence (AI) model used by the chatbot sensation ChatGPT has been released, according to the startup that created it. GPT-4, the company’s most recent AI innovation, is a sizable “multimodal model,” according to a blog post by OpenAI, that “exhibits human-level performance on multiple professional and academic standards.”
The business used a simulated bar exam to demonstrate the potential of the AI, noting that GPT-3.5 performed about as well as the worst 10% of test-takers but GPT-4 passed with a score in the top 10%.
The distinction between GPT-3.5 and GPT-4 are more pronounced when the complexity of the task reaches a sufficient threshold where the latter is more reliable, creative, and able to handle much more nuanced instructions than GPT-3.5, according to the OpenAI.
“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 (though far from perfect) on factuality, steerability, and refusing to go outside of guardrails.”
Billionaire investor Elon Musk, who has also co-founded a brain-chip company Neuralink that develops implantable brain–computer interfaces, responded to a tweet about ChatGPT-4 acing exams, wondering what would be left for humans to do.
“We better get a move on with Neuralink!” he added.
