The long-awaited sequel to ChatGPT is now live and claims to do “human-level performance” on university-level tests.
OpenAI said that GPT-4, the next version of its chatbot powered by artificial intelligence, was a “milestone” in developing deep learning, which is like how humans learn.
The San Francisco-based company said in a blog post on Tuesday, “We’ve spent 6 months iteratively aligning GPT-4 using lessons from our adversarial testing programme as well as ChatGPT, resulting in our best-ever results (though far from perfect) on factuality, steerability, and refusing to go outside of guardrails.”
OpenAI, which is backed by Microsoft, said that the new version of its AI-powered chatbot is a “multimodal” model that can create content from both images and text prompts.
OpenAI President Greg Brockman showed how GPT-4 made a real website based on a hand-drawn mock-up in an online demo.
OpenAI said that the update makes it possible for aspiring lawyers to pass the bar exam with a score in the top 10 per cent of test-takers, up from the bottom 10 per cent before the update.
OpenAI said that the chatbot can also do better than 90% of people 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 to get into graduate school.
The company said GPT-4 is much less likely to give wrong, offensive, or 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, on the other hand, said that GPT-4 is still “not fully reliable” and can still come up with “hallucinations” and reasoning mistakes.
When OpenAI released ChatGPT in November, it sent shockwaves through the tech world and made people wonder about the future of everything from education to journalism to health care.
Tech giants like Google, Microsoft, Huawei, Alibaba, and Baidu are all racing to release their versions of the technology in a fierce battle to be the best in the growing AI market.