ChatGPT to get video features

OpenAI’s artificial intelligence chatbot will be able to recognize objects through a smartphone camera and give real-time reaction

OpenAI’s ChatGPT is now capable of processing video cues provided by users. The new feature is already being made available to paid ChatGPT Plus and Pro subscribers, with enterprise and educational customers expected to gain access next month.

Launched in 2022, the artificial intelligence chatbot has been steadily expanding its capabilities. Last year, its developers said the GPT-4 Large Language Model could get a higher score on the SAT exam – the standardized test widely used for college admissions in the US – than over 90% of all people.

During a livestream on Thursday, OpenAI unveiled the latest feature, which enables ChatGPT to interact with the user based on what it observes through a smartphone camera or what is currently being displayed on the screen. For instance, a human can ask the chatbot to help write a suitable answer to a message in an open app or give real-time advice on performing tasks in the real world.

In February, researchers unveiled a tool named ‘Sora’, which is “able to generate complex scenes with multiple characters, specific types of motion, and accurate details of the subject and background,” all based on prompts from users. It can extend images or videos fed to it with new material, the company said at the time, sharing a few examples of AI-generated videos on its social media accounts.  

Read more

In July, Reuters reported that OpenAI was working on an approach aimed at drastically improving the reasoning capabilities of AI models. The capabilities, described by an anonymous source as a “work in progress,” would allow ChatGPT not only to generate answers to queries, but also conduct “deep research” and navigate the internet proactively.

‘Strawberry’ should enhance the AI chatbot’s ability to find common-sense solutions that are usually intuitive to humans – something of a weak spot for ChatGPT and similar models so far.

Around the same time, a group of Russian scientists from the T-Bank’s AI Research Lab and the Moscow-based Artificial Intelligence Research Institute (AIRI) said they had developed a new artificial intelligence model capable of self-adaptation to new tasks and context without additional input by humans.

According to the developers of the model dubbed ‘Headless-AD’, it was already capable of performing five times more actions than it was initially taught.