OpenAI, the company behind ChatGPT, recently launched an enhanced version of the technology that powers the popular chatbot. GPT-4 is currently available to ChatGPT Plus subscribers, while there’s a waitlist for developers who want to incorporate its API into apps and services.
Microsoft also confirmed that its AI-powered search engine Bing
runs on GPT-4. The tech giant has removed the waitlist, allowing anyone to test the new Bing Chat.
“If you’ve used the new Bing preview at any time in the last five weeks, you’ve already experienced an early version of this powerful model,” Microsoft’s Yusuf Mehdi, Corporate Vice President & Consumer Chief Marketing Officer, wrote in a blog post
Bing users might not notice the difference between the new and the old version. According to OpenAI, the difference can be imperceptible in casual conversations, but ChatGPT Plus users will be able to take full advantage of its new capabilities.
OpenAI explained its latest version GPT-4
“is more reliable, creative, and able to handle much more nuanced instructions” than its predecessor. The real power of GPT-4 comes out when presented with more challenging tasks whose complexity “reaches a sufficient threshold.”
For instance, on a simulated bar exam that US law school graduates have to take before professional practice, GPT-3.5 scored in the bottom 10% while GPT-4 placed among the top 10% of test takers.
GPT-4 is multimodal, meaning it can generate content from both text and image prompts, but image inputs are still being tested and aren’t publicly available. Its researchers dedicated six months for GPT-4 to produce “the best-ever results (though far from perfect) on factuality, steerability, and refusing to go outside of guardrails.”
Microsoft’s Bing, like similar AI tools, is prone to hallucinations. In the past, users have reported harmful outputs which prompted the tech giant to impose a number of restrictions on its use. OpenAI has implemented additional safety measures on GPT-4
, making it 82% less likely to respond to disallowed prompts, and scoring 40% higher on factuality tests than its forerunner.
The tech giant also incorporated the technology into Dynamics 365 business apps
, with more Microsoft products expected to get the ChatGPT treatment soon.