Apple to take help of Shutterstock to train its AI model

Apple has entered into a partnership with Shutterstock, a stock photos and videos library, to utilize its data for training its artificial intelligence (AI) models. The agreement suggests that Apple is making strides towards developing its proprietary foundation model. Despite previous speculations regarding the integration of Google's Gemini AI and Baidu's Ernie Bot (in China) into the iPhone 16 series for AI functionalities, it seems Apple's long-term strategy involves incorporating its own native AI model.

Apple has partnered with Shutterstock
 

Sources familiar with the matter, as cited by Reuters, revealed that Apple has inked a deal with Shutterstock to access its licensed data, encompassing images, videos, and music, for AI training purposes. Similar agreements have reportedly been established by other tech giants like Meta, Google, and Amazon. Jarrod Yahes, the Chief Financial Officer of Shutterstock, disclosed that these deals initially ranged from $25 million to $50 million each, with subsequent expansions in size for most firms.

In addition to the Shutterstock deal, Apple has also secured agreements with Defined.ai, an AI data firm based in Seattle, which licenses data to tech companies for training their AI models. According to Daniela Braga, the CEO of Defined.ai, the typical rates for licensing data are around $1 to $2 per image, $2 to $4 per short-form video, and $100 to $300 per hour of longer films, while the market rate for text stands at $0.001 per word.

These developments suggest two intriguing possibilities. Firstly, the inclusion of Gemini AI and Ernie Bot in the iPhone 16 series may serve as interim solutions until Apple finalizes its in-house AI models. Therefore, even if the initial models come with outsourced AI capabilities, they are likely to be replaced once Apple officially unveils its proprietary AI technology. Secondly, the partnership with Shutterstock indicates that Apple's AI endeavors extend beyond text-only capabilities to encompass images, music, and videos.

While the exact purpose of developing these AI models remains unclear, it is reasonable to infer that they are multimodal, potentially involving aspects of computer vision and the generation of images and videos. Notably, Apple Research has released two separate papers outlining the AI models currently under development.