ChatGPT, the viral chatbot that took the Internet by storm, rolled out an updated version last Tuesday, March 14.

OpenAI, the startup that owns ChatGPT, announced in a blogpost that the new version of ChatGPT is more powerful and predictable than previous versions, and can also analyse images and handle much bigger blocks of text.

The company introduced a more powerful version of ChatGPT known as GPT-4. The chatbot is “more creative and collaborative” than previous versions when it comes to tasks like writing screenplays and composing songs.

The updated version of ChatGPT was launched this Tuesday through data-sharing partnerships, which companies like Duolingo Inc., and Morgan Stanley are paying to gain access to. 

The new version will also be provided to ChatGPT Plus subscribers, who pay $20 per month for faster service. The new version is not available yet to the free service. 

During a presentation, OpenAI explained how the tool could be used to summarise articles into sentences with every letter beginning with “Q” or explain obscure tax codes.

ChatGPT can also better analyse images for information. For instance, you can scan a photo of ingredients and then receive suggested potential dishes and recipes that you can make from those ingredients, the company said. However, the ability to analyse images has not yet been widely launched yet.

Greg Brockman, OpenAI president and co-founder, said that GPT-4 makes fewer errors than its predecessor GPT-3.5, but it still works when used by people who can check its work.

“It isn’t perfect but neither are you and together it’s this amplifying tool that lets you just reach new heights,” Brockman said.

For Keith Peiris, co-founder of the AI startup Tome, GPT-4 feels like a smart 10th grader, while GPT-3 and GPT-3.5 were sixth graders. Peiris said the new version of ChatGPT can analyse 25 pages of text compared with three pages in earlier versions.

AI analysts say that there are likely problems with generative AI. Although ChatGPT and other chatbots are accurate on topics where they have taken in high quality information, they are still capable of spewing out sexist and racist answers as well as conspiracy theories and misinformation. It can also lead to factual mistakes.

On the other hand, OpenAI said they consulted with more than 50 experts in AI, security, and safety to develop GPT-4. The updated version is 82 per cent less likely to respond to requests for “disallowed content.”