Slow and steady wins the race: With big-name companies like Microsoft and Google vocally throwing their weight and money into the conversational AI market and releasing large language models based on OpenAI’s work, it might be easy to write Apple off as uninterested or too far behind to matter. However, it is too early to write Apple off just yet.
The Information notes that Apple is working on several AI projects, increasing budget expenditures to “millions of dollars a day.” It currently has resources focused on a conversational model, an image generation system, and a multimodal AI that handles text and visual data.
Senior Vice President of Machine Learning and AI Strategy John Giannandrea heads up development on the language model. Apple initially hired Giannandrea to improve Siri, so it seems probable that Cupertino’s AI assistant will be the primary benefactor of Apple’s LLM research.
The company will likely blend all its AI systems to accomplish various tasks through voice commands. Training and integrating different AIs is expensive and time-consuming, which would explain the perceived disinterest. However, while everybody else is launching and patching models with subpar performance, Apple is patiently building something that will outperform them and offer more than just chatbot functionality.
The Information notes that Apple’s largest foundational AI model is “Ajax GPT.” It is supposedly already more powerful than OpenAI’s GPT-3.5. Ajax has over 200 billion parameters, 25 billion more than GPT-3.5 but far short of GPT-4’s 1.7 trillion.
Another reason not to count Apple out of the race is its two-billion+ user base. Leveraging a comprehensive blended AI system that exceeds performance expectations set by Bing and Bard and putting it into the hands of two billion users virtually overnight could quickly leapfrog Apple to the lead in the race.
Apple has not announced anything regarding its AI efforts. As usual, the Cupertino tech titan likes to keep the development of its products under wraps until the last second. So, there is no way to guess when we could see an LLM-driven Siri commanding other complicated AI tasks on any Apple platform.
However, it is most assuredly coming. The company is not one to dump hundreds of millions per year into a project, never to have it materialize, no matter how long it takes. The Apple Vision Pro is a perfect example of that slow-burn methodology.