CNN
—
Nearly four months after OpenAI shocked the tech industry with ChatGPT, the company has released a next-generation version of the technology powering the viral chatbot tool.
In a blog post on Tuesday, OpenAI unveiled GPT-4, which the company says is capable of performing well on a battery of standardized tests, and It’s unlikely to “get off the guardrails” in response, as some users have experienced before.
OpenAI said the updated technology passed the mock law school bar exam, scoring in the top 10% of test takers; by comparison, the previous version, GPT-3.5, the score is around the bottom 10%. GPT-4 can also read, analyze or generate up to 25,000 words of text and write code in all major programming languages, according to the company.
OpenAI described the update as the company’s “latest milestone.” According to the company, it demonstrated “human-level performance on a variety of professional and academic benchmarks,” although it remained “less capable than humans” in many real-world scenarios.
GPT-4 is the latest version of OpenAI’s large-scale language model trained on massive amounts of online data to generate convincing responses to user prompts. The updated version is now available through a waitlist, and it has made its way into some third-party products, including Microsoft’s AI-powered Bing.
“We’re pleased to confirm that the new Bing is running on GPT-4, which we custom-built for search,” Microsoft explain Tuesday. “If you’ve used the new Bing preview at any point in the past five weeks, you’ve already experienced an early version of this powerful model.”
Although ChatGPT has deep impression Since its launch in November 2022, many users have been able to generate original articles, stories and lyrics based on user prompts, which has also raised some concerns. In recent weeks, AI chatbots, including tools from Microsoft and Google, have been criticized for emotionally overreacting, making factual errors and outright “hallucinating,” as the industry calls them.
GPT-4 has similar limitations to earlier GPT models. In a series of tweets announcing the update Tuesday, OpenAI CEO Sam Altman wrote: “It’s still buggy, still limited, and it still looks more impressive on first use than after you spend more time with it. profound.”
But he said there were clear improvements. “It’s more creative than previous models, with significantly less hallucinations and less bias,” he said. wrote.
Still, the company says, “extreme caution should be exercised when using language model output, especially in high-stakes settings.”
The news comes two weeks after OpenAI announced it open access ChatGPT tools from third-party companies have paved the way for chatbots to be integrated into numerous applications and services.
Instacart, Snap, and tutor app Quizlet are early partners experimenting with the tool.In January, Microsoft confirmed it was making “Billions” of investments In OpenAI, and has rolled out the technology to some of its products, including its search engine Bing.