Apple Uses 'Fake' Data To Train Its AI Model

Apple Develops AI with fake data (photo: dock. Apple)

JAKARTA Not long ago, it was revealed that Apple was trying to catch up in the field of Artificial Intelligence (AI). One of the efforts Apple made was to take advantage of synthetic data.

According to a Bloomberg report, reported via 9to5mac, researchers at Apple rely heavily on artificial or fake data sets. This data is licensed by third parties and is specifically designed to train AI models.

In order to get maximum results, Apple also increases its fake data using an iPhone. They compare fake data obtained from third parties with the language in the user's email series stored on the cellphone.

This comparison action will not harm its users. The reason is, Apple does not include user information into the AI model they developed.

Currently, many companies are trying to maintain the privacy of their users, including Apple. Therefore, the use of synthetic data or fake data is not something new. Apple is not the first company to rely on this type of data.

Several other major technology companies such as OpenAI, Microsoft, and Meta have also done the same. They have succeeded in developing sophisticated AI models using special data that are not taken from their users.

Synthetic data is much safer to use because it is generated by an algorithm. This data is made internally so that technicians can guarantee the accuracy of the label, adjust the data to various scenarios, and keep protecting the user's personal data.

Apple had discussed how they used synthetic data. The company explains that it created thousands of emails with the same topic, then compared its emails to messages received by users locally.

After that, technicians analyzed which synthetic data was the most relevant. Synthetic data can improve the performance of the AI model if done properly. However, technicians must carefully curate the data to reduce hallucinations in their AI model.