OpenAI has officially announced the latest version of its hugely popular natural language bot ChatGPT.
As expected, GPT-4 is 'multimodal' which means users can enter different kinds of input - video, sound, images, text – and still get human-like answers in natural language.
This represents a pretty substantial update on the previous version, which caused such a sensation when it was released in November 2022. Indeed, ChatGPT was the biggest tech story of the year and prompted a series of responses from the world's leading tech companies.
Microsoft is using a version of the product as the backbone for its new Bing search engine, while Google is reported to be developing its own similar chatbot.
ChatGPT provoked much discussion about the future of search and therefore the business model that underpins so much online activity. Observers asked: if you can get a workable and readable answer to your question from ChatGPT, why bother browsing multiple search results?
The new version will intensify these conversations. Thanks to its multimodal training, ChatGPT should be able to understand and produce language that is more likely to be accurate and relevant. Indeed, OpenAI claims that GPT-4 is 40 percent more likely to provide factual responses. The new version can also remember and act on more than 20,000 word prompts.
Meanwhile, ChatGPT continues to have an impact on the business world. As part of its news update, OpenAI revealed new partnerships. It says the subscription tier of the language learning app Duolingo, Duolingo Max, will now offer English-speaking users AI-powered conversations in French or Spanish. It also confirmed that payment processing company Stripe is using GPT-4 to answer support questions from corporate users.