OpenAI, the company behind the sensational ChatGPT has released GPT-4.
The company said the new version of the chatbot boasts of “human-level performance” in university-standard exams.
OpenAI said GPT-4, the next gen AI-powered chatbot, marked a “milestone” in the development of deep learning, which imitates the way humans acquire knowledge.
“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,” the company said in a blog post on Tuesday.
OpenAI, backed by Microsoft, said the new version of its chatbot is a “multimodal” model can generate content from both images and text.
In an online demonstration, GPT-4 created a real website based on a hand-drawn mock-up.
OpenAI also disclosed that the new version of the chatbot can pass the bar exam for prospective lawyers with a score in the top 10 percent of applicants.
OpenAI also said the chatbot can also beat 90% of humans who take the evidence-based reading and writing section of the Scholastic Assessment Test and the verbal section of the Graduate Record Examination which is used for admission to postgraduate education.
The company added that GPT-4 is much less likely to give out inaccurate, offensive, insulting answers than ChatGPT.
“We spent six months making GPT-4 safer and more aligned. GPT-4 is 82 percent less likely to respond to requests for disallowed content and 40 percent more likely to produce factual responses,” OpenAI said.
OpenAI, however admitted that GPT-4 is “not fully reliable” and can still produce unexpected answers known as “hallucinations”
ChatGPT which was launched in November last year stook the world by surprise, and it prompted tech giants like Google, Microsoft, Huawei, Alibaba to deepen their focus on AI.