Apple has reportedly claimed that the open-source Efficient Language Models (OpenELM) AI model it recently released was not used for its Apple Intelligence or any of the company’s machine learning features. The Cupertino-based tech giant open-sourced the OpenELM model in April. The reported statement comes just days after an investigation revealed that the AI model was trained on datasets containing subtitles from hundreds of thousands of YouTube videos. Notably, these datasets were also used to train AI models from Anthropic, Salesforce, Nvidia and more.
Apple reportedly denies using OpenELM for Apple Intelligence
Speak Speaking to 9to5Mac, the iPhone maker said that OpenELM does not power either AI features under the Apple Intelligence branding or other machine learning features. The company also added that OpenELM was created as a contribution to the research community to help them advance the development of large language models, according to the report.
Because the AI model was designed for research purposes, it wasn’t used for any of Apple’s internal AI innovations, the company told the publication. Notably, OpenELM is still open source, and its model weights are accessible via the company’s Hugging Face list.
At the time of release, Apple made the OpenELM family of AI models available in two flavors: pre-trained and instructional. These models are further divided into different parameters of 270 million, 450 million, 1.1 billion, and 3 billion.
AI models trained on YouTube data
A recent investigation accused Apple’s OpenELM and multiple other AI models from major AI companies of using caption data from YouTube videos for training. The report found that companies including Anthropic, Salesforce, Nvidia and others were using a publicly available data repository called the Pile, which contained data from YouTube videos from major content creators such as Marques Brownlee, MrBeast, CarryMinati and more.
While Apple denied using the AI model in its devices and operating systems, Jennifer Martinez, a spokesperson for Anthropic, told Proof News, the publication that conducted the investigation: “The Pile contains a very small subset of YouTube captions. YouTube’s terms include direct use of its platform, which is different from use of the Pile dataset. As for any potential violations of YouTube’s terms of service, we must refer you to the Pile authors.”
