Apple has reportedly entered a deal with the stock photos and videos library Shutterstock to use the data to train its artificial intelligence (AI) models. The report also confirms the speculations that the tech giant is likely in the advanced stages of developing its in-house foundation model. Even as earlier reports have mentioned that the iPhone 16 series could be equipped with Google’s Gemini AI and Baidu’s Ernie Bot (in China) for the AI features, it appears the long-term plan for Apple is to integrate a native AI model.
In a report, Reuters cited unnamed people familiar with the matter, to reveal that Apple has struck a deal with Shutterstock to procure its licenced data — pictures, videos, and music — to train its AI models. However, Apple is not the only one as Meta, Google, and Amazon are also reported to have struck similar deals. Shutterstock Chief Financial Officer Jarrod Yahes told Reuters that the size of the deals tech companies initially entered ranged from $25 million (roughly Rs. 208 crores) to $50 million (roughly Rs. 416 crores) each. Most firms have already expanded the size of the deal, as per the CFO.
This is not the only deal the Cupertino-based tech giant has struck to procure data for its AI model. As per the report, it also has agreements in place with Seattle-based Defined.ai, an AI data firm that licences data to tech firms training their AI models. The CEO of Defined.ai Daniela Braga told Reuters, “Companies are generally willing to pay $1 to $2 per image, $2 to $4 per short-form video and $100 to $300 per hour of longer films. The market rate for text is $0.001 per word.”
If the information is to be believed, it points to a couple of interesting possibilities. First, the company’s rumoured inclusion of Gemini AI and Ernie Bot is likely a stop-gap till it can fine-tune its own AI models. This means even if the iPhone 16 series comes with outsourced AI capabilities, they will eventually be replaced when Apple officially announces its in-house AI. Second, the deal with Shutterstock highlights that Apple is not only focusing on text-only capabilities but also images, music, and videos.
While it cannot be said whether it is just to develop computer vision or to have the models generate images and videos as well, it is safe to assume the under-development AI models are multimodal. Notably, Apple Research has published two separate papers on the family of AI models it is working on.