ChatGPT upgrade to focus on video

A new version of the artificial ­intelligence that powers ChatGPT, which can turn text into video and other media, will be released this week, according to Microsoft.

ChatGPT is set to launch a version that will generate videos. Picture: Getty Images

A new version of the artificial ­intelligence that powers ChatGPT, which can turn text into video and other media, will be released this week, according to a Microsoft executive.

GPT-4 is a more powerful and versatile update of GPT-3.5, the large language model that sits behind ChatGPT, which has caused a sensation because of its ability to write well in many formats based on text commands.

More than 100 million people use it regularly, making the ­application, released in November, the fastest-growing in technology history.

Now the world needs to prepare for another step change after Andreas Braun, chief technology officer for Microsoft in Germany, said the company would introduce GPT-4 in that country this week.

“We will have multimodal models that will offer completely different possibilities; for example, videos,” Mr Braun said.

The event was reported by German publication Heise.

Microsoft is the largest investor in OpenAI, the company that developed ChatGPT, and has incorporated the technology into its Bing search engine.

Rumours have been swirling for months about the capabilities of GPT-4.

“The GPT-4 rumour mill is a ridiculous thing. I don’t know where it all comes from,” OpenAI chief executive Sam Altman said. “People are begging to be disappointed and they will be.”

If GPT-4 is “multimodal”, however, it will enable people to generate images, sounds and video from simple text demands. It is not clear in what format the new AI model will be released, whether it will be as a stand-alone feature or incorporated into a Microsoft product. Microsoft did not comment on Mr Braun’s statement.

OpenAI has already developed DALL-E, an AI for generating images from text, and Jukebox, which can create music from text prompts.

Text-to-video models have also been developed by Meta (Make-A-Video) and Google (Imagen) but not released as products.

Marianne Janik, chief executive of Microsoft Germany, described the AI development and ChatGPT as a turning point – “an iPhone moment”.

With the expansion of capabilities for AI comes increased concern about their impact. A version of GPT deployed in Bing at one point tried to persuade a journalist in a conversation to leave his wife.

Mr Altman addressed this in a blog post late last month: ­“Society will face major questions about what AI systems are allowed to do, how to combat bias, how to deal with job displacement and more.” He said OpenAI would deploy its models gradually but also “empower individuals to make their own decisions and the inherent power of diversity of ideas”.

Others feel the pace of change is not gradual. Henry Ajder, an expert in generative AI, said the release of text-to-video would be a “big development” and quicker than anticipated.

“I think video, as we’ve seen with deep fakes … have had some of the most impact and have some of the most visceral impact when deployed in malicious contexts. How robust is it going to be both from an actual technical perspective, but also from a safety perspective?”

OpenAI did not respond to a request for comment.

The Times