👀 Apple's new AI model outperforms GPT-4 | Is Apple Secretly Building AI Agents?

Apple has made significant advancements in AI research, potentially developing a model that surpasses GPT-4. Speculation suggests that Apple may be secretly working on building AI agents, with a focus on reference resolution, multimodal AI, and exploring the acquisition of perplexity for a potential search engine integration.

Apple has recently made advancements in AI research, potentially developing a model that outperforms GPT-4. This development has sparked speculation about Apple’s potential pursuit of building AI agents. The company’s focus on AI has become more apparent, especially after scrapping its autonomous electric car project earlier this year. Apple researchers have published a paper on reference resolution in language modeling, showcasing their progress in understanding screen context.

Rumors suggest that Apple may be considering acquiring perplexity, a company specializing in a different approach to online web search using large language models. This technology aims to summarize information found online based on various sources, enhancing the search experience. The potential acquisition of perplexity could provide Apple with its own search engine and eliminate the need to develop its own large language model from scratch. This move could give Apple a competitive edge in the AI space.

Apple’s research on multimodal AI and screen context understanding could pave the way for the development of AI agents that run on devices like iPhones and laptops. By leveraging smaller language models specifically tailored for reference resolution tasks, Apple aims to create more efficient AI agents that can interpret on-screen context effectively. This approach could lead to the integration of AI agents into everyday tasks like searching, emailing, and online shopping.

The proposed AI model, known as “realm,” has shown promising results in domain-specific queries and outperformed GPT-4 in certain tasks. By utilizing smaller language models, Apple’s approach focuses on running AI agents on devices without requiring an internet connection. This cost-effective and efficient method could revolutionize how AI agents operate, enhancing user experiences and interactions with technology.

Overall, Apple’s advancements in AI research, particularly in reference resolution and multimodal AI, indicate a shift towards developing on-device AI agents. The potential integration of these agents into Apple devices could significantly impact the way users interact with technology, offering a more personalized and efficient experience. Apple’s exploration of smaller language models for specific tasks showcases a unique approach to AI development that could set the company apart in the competitive AI landscape.