In a bid to advance its voice assistant, Apple has reportedly increased its investment in artificial intelligence research, focusing on developing conversational features for Siri. It’s understood that millions are being allocated daily to research and development in this area.
Earlier in May, word spread that Apple was hiring more hands for its generative AI initiatives. While this wasn’t officially confirmed, CEO Tim Cook described the domain of generative AI as “rather intriguing”.
This, however, isn’t Apple’s maiden venture into generative AI. Rewinding four years, John Giannandrea, Apple’s AI chief, kick-started a group dedicated to the development of large-language models (LLMs) similar to ChatGPT.
Leading the charge for Apple’s AI chatbot endeavours is the Foundational Models team, helmed by Ruoming Pang—a former 15-year Google veteran.
The team, despite its lean size of just 16, boasts of achievements on par with industry giants like OpenAI; the latter is said to have spent a staggering $100 million training a comparable LLM.
Insiders have revealed that Apple houses at least two other groups committed to refining language and visual AI models.
One team is devoted to Visual Intelligence, churning out images, videos, and 3D visualisations. The other is dabbling in multimodal AI, adept at juggling text, imagery, and videos.
Also read:
Siri’s AI Future and the Road Ahead
Apple’s vision is clear: weave LLMs into the fabric of Siri.
This would allow users to orchestrate intricate tasks using everyday language, mirroring Google’s strides to refine their voice assistant.
Apple is confident that its in-house language model, named Ajax GPT, outperforms OpenAI’s GPT 3.5.
However, Apple’s journey to integrate LLMs into its suite of products is fraught with obstacles. While rivals often rely on cloud-based systems, Apple favours an on-device strategy to champion privacy and performance.
But there’s a snag: Apple’s LLMs, Ajax GPT included, are notably hefty, posing challenges for seamless on-device deployment.
Apple’s commitment to enhance Siri with generative AI is evident. But, because of the company’s dedication to on-device operations, it faces a tightrope of balancing performance with privacy.
