Tech Desk
Jul 18, 2024
Apple has confirmed that its OpenELM model, an open-source language model, has not been used to power any of its AI (artificial intelligence) or machine learning features, including Apple Intelligence.
However, this clarification came after the investigation revealed that the Cupertino-based tech giant along with other tech companies has used the dataset of 'YouTube subtitles' to train their AI models.
Earlier, companies including Apple, Anthropic, and NVIDIA were accused of using this dataset to train their AI models; none of the statements were confirmed.
This dataset was comprised of over 170,000 videos containing well-known YouTubers such as MKBHD and Mr. Beast.
Read more: Apple's new patent reveals system for ranking Wi-Fi networks in iOS
However, Apple has stated that this dataset was used only for further research and to make contributions to the large language model being developed in the open source.
Apple's OpenELM model was released in April and can be found on its Machine Learning Research website for download. According to the reports, the model was developed only for research purposes, so it did not empower any one of its Apple Intelligence features.
Apple said it would train its Apple Intelligence models on licensed data, with the public data captured by the model supplied by OpenELM's web crawler. It is not its intention to replicate a new version of OpenELM's model.
