Apple, which will announce the artificial intelligence features it will bring to iPhones in June, continues to make moves in this regard. Apple researchers have published several different types of artificial intelligence models in recent months. A new one has been added to this.

Researchers,”OpenELMThey published their language model called “” on Hugging Face Hub. These models give us an idea of ​​how generative artificial intelligence will be used in devices.

Models allow artificial intelligence to work directly from the device rather than through the cloud

These models, which stand for “Open Source Efficient Language Models”, are used in applications such as writing emails. “very productive” on text-oriented tasks They were described as. Also open source and They can be used by any developer.

OpenELM models come in four different sizes and are truly “uvula modelThey are at a level that can be classified as “. So that the dimensions are 270 million parameters, 450 million parameters, 1.1 billion parameters and 3 billion parameter. The parameters show how many variables the models can understand from training datasets during the decision-making process.

Let’s give an example to help you understand how small it is. Microsoft recently introduced Phi-3, which it calls “my smallest artificial intelligence model.” The size of that model was 3.8 billion parameters. Apple’s numbers drop to 270 million. So what does it mean that they are so small? They are less expensive and phones, laptops They can be optimized for use on devices such as.

In other words, these models provide clues about how artificial intelligence on iPhones will work. Its built-in usage eliminates the need for the cloud. We have seen Apple increasing its strides in artificial intelligence on devices. A few days ago, Bloomberg’s Mark Gurman wrote that in addition to cloud Features that work entirely on the device He also stated that he developed it. These models also support this claim.