OpenAI wants to make its chatbot even more powerful
Nearly four months after OpenAI took the tech industry by storm with ChatGPT, the company is launching its next generation of the technology powering the viral chatbot tool.
In a blog post on Tuesday, OpenAI showcased GPT-4, which the company says is able to perform well on a variety of standardized tests, and is also less likely to "push the envelope" with its answers, as it says some users have previously experienced.
According to OpenAI, the updated technology passed a mock law school exam with scores in the top 10% of applicants; In contrast, the previous version GPT-3.5 only hit the bottom 10%. According to the company, GPT-4 can also read, parse or generate up to 25,000 words of text and write code in all major programming languages.
OpenAI called the upgrade the "final step" for the company. Although still "less capable" than humans in many real-world scenarios, according to the company, it "shows human-level performance in various academic and professional criteria."
GPT-4 is the latest version of OpenAI's large language model, which is trained on large amounts of online data to generate compelling answers to user queries. The updated version, now available through a waitlist, is already making its way into some third-party products, including Microsoft's AI-powered Bing.
"We're excited to confirm that the new Bing we've customized for search runs on GPT-4," Microsoft said Tuesday. "If you've used the new Bing preview at any time in the last five weeks, you've already experienced a previous version of this powerful model."
Although ChatGPT has impressed many users with its ability to create original essays, stories, and text in response to user feedback since its launch in November 2022, it has also raised some concerns. AI chatbots, including tools from Microsoft and Google, have come under fire in recent weeks for overreacting, making factual errors and engaging in what the industry outright calls "hallucinations."
GPT-4 has similar limitations to previous GPT models. "It's still buggy, it's still limited, and it still looks more impressive the first time you use it than if you've spent more time on it," OpenAI CEO Sam wrote. Altman, in a series of tweets on Tuesday, where the update was announced.
But there are noticeable improvements, he said. "He's more creative than previous models, hallucinates a lot less, and has less bias," he wrote.
Still, according to the company, "very careful consideration should be given when using language model results, especially in high-risk contexts."
The news comes two weeks after OpenAI announced that it was opening up access to its ChatGPT tool to third-party companies, paving the way for the chatbot to be integrated into many apps and services.
Their chat keeps being improved and getting smarter.