ReALM: Apple offers an AI model similar to GPT-4

a apple foot [PDF] A new paradigm of artificial intelligence promises to revolutionize artificial intelligence capabilities crab – it's called kingdom (Reference decision as language modelingor reference accuracy such as language modeling).

When running on the device, it seeks to make Siri smarter, allowing the assistant to understand “entities” on the user’s screen, as well as conversations and background processes, such as alerts and music, to respond more accurately.

are they: a screen (what is displayed), conversation (data relevant to the conversation, from previous interactions or from the virtual assistant) and bottom (Background processes, which affect the context of the interaction).

A recently published paper explores ReALM and shows that it can outperform existing systems, supporting its effectiveness in enhancing Siri's usefulness through advanced linguistic modeling.

We demonstrated significant improvements over the existing system with similar functionality across different reference types, with our more compact model achieving absolute gains of over 5% for on-screen references. We also made comparisons with GPT-3.5 and GPT-4, where our more compact model achieved similar performance to GPT-4, and our larger models significantly outperformed it.

Moreover, testing Standard Against OpenAI's ChatGPT 3.5 and ChatGPT 4.0 shows that ReALM achieves similar performance to GPT-4, with larger models significantly outperforming it.

Our goal is to have both variables predict the list of entities from the available set. In the case of GPT-3.5, which accepts text only, our input consists of the prompt only; However, in the case of GPT-4, which also has the ability to place images in context, we provided the system with a screenshot of the historical accuracy task shown on the screen, which we found helped improve performance significantly.

Even with fewer parameters than GPT-4, everything indicates that ReALM outperforms it at textual tasks and understanding user commands, despite GPT-4 having the ability to put images in context.

It is very likely that we will see news related to these technologies next June, at WWDC24. This year's event is expected to focus specifically on integrating artificial intelligence features into Apple's main operating systems. Let's wait!

via 9to5Mac