OpenAI, the creators of ChatGPT, has launched a new and more powerful AI bot – GPT-4 – which it admits is so advanced it could be ‘harmful’.
The bot can now accept input in the form of images as well as text, but still outputs its answers in the form of text, which means it can provide detailed descriptions of the images.
And if asked, “What’s so funny about this picture?” GPT-4 will reply. Among other things, it can instantly calculate people’s tax liability and can pass the Uniform Bar Exam.
“We created GPT-4 as the latest milestone in OpenAI’s efforts to extend deep learning,” OpenAI said in a blog post. “GPT-4 is a large multimodal model (it accepts image and text inputs, and emits text output), although it is less capable than humans.” In many real-world scenarios, it displays human-level performance on various professional and academic standards.”
The company has warned that the model is still prone to ‘hallucinating’ with inaccurate facts – and can be persuaded to spew out false or harmful content.
It can now pass legal exams with results in the top 10% (GPT-3, which previously supported ChatGPT, can pass exams, but with results in the bottom 10%).
GPT-4 is now available to ChatGPT subscribers: free users will still encounter the older GPT-3.5.
It also supports Microsoft’s Bing chatbot (and the company revealed it’s been doing this for six weeks).
“We’re excited to confirm that the new Bing runs on GPT-4, which is for search,” said Microsoft’s head of consumer marketing, Youssef Mahdi. The latest model from OpenAI”.
The company says that GPT-4 is being integrated into other products from companies such as Duolingo and Stripe.
OpenAI said, “GPT-4 and later models have the potential to significantly impact society in both beneficial and detrimental ways. We are collaborating with external researchers to improve how we understand and assess potential impacts, as well as build assessments of the dangerous capabilities that may emerge in future systems.”
The ability to accept images means that users can now request ChatGPT for screenshots and other media.
Speaking this week at SXSW, one of the founders of OpenAI said that fears that AI tools will take people’s jobs are overblown — and that AI will free up humans to focus on important work.
According to analytics firm LikeWeb, ChatGPT averaged nearly 13 million daily users in January, making it the fastest-growing Internet app of all time.
It took TikTok about nine months after its global launch to reach 100 million users.
OpenAI, a private company backed by Microsoft Corp, made ChatGPT available to the public for free in late November.