Apple Now Spends Millions a Day on Conversational AI

Apple is ramping up its spending on conversational artificial intelligence, spending “millions of dollars a day,” according to a new report from The Information that offers some insight into Apple’s machine learning and artificial intelligence development.

Although Apple’s AI chief John Giannandrea has said he’s “skeptical” of AI chatbots, he created a conversational AI team four years ago to explore the possibilities of the technology.

Apple has a “Foundational Models” team of just 16 people working on conversational AI. However, Apple is spending millions of dollars per day training its language models. It’s a process that doesn’t come cheap, as it requires quite a bit of expensive hardware. OpenAI’s Sam Altman has been quoted as saying his company spent more than $100 million to develop GPT-4.

In July, Bloomberg’s Mark Gurman reported that Apple is currently working on developing its own “Apple GPT” artificial intelligence project, aimed at competing with OpenAI’s ChatGPT. Gurman says the project has become a priority at the Cupertino firm over the last few months, as chatbot services and AI features in apps have been on the rise.

The chatbot arena has become quite competitive since OpenAI first debuted ChatGPT in 2022. Both Google and Microsoft offer their own chatbots, but until recently we haven’t seen any signs that Apple plans to debut a chatbot replacement for Siri any time soon.

Apple’s Ajax platform is based on Google’s Jax machine learning framework, which runs on Google Cloud. While Apple is said to have tried out OpenAI’s technology, it decided against actually signing a contract with OpenAI.

Gurman’s sources have told him that Apple is working to develop its own “Ajax” framework for large language models. Some company employees currently have access to an internal chatbot that some engineers call “Apple GPT.” Apple employees must request special access to the Apple chatbot, and the app’s output cannot be used to develop new customer-facing products.

However, Apple has not yet developed a “clear strategy” for developing AI-related products for Apple customers. The Apple chatbot can currently answer questions using the data the company has trained it with.

Apple is admittedly lagging behind other tech firms when it comes to AI-based technologies, which have the potential to change how users interact with their phones, computers, and tablets.

The report says Apple’s main artificial intelligence goal is the development of a voice assistant similar to Siri to automate tasks that require multiple steps. There is a possibility that Apple’s Siri team could make use of the AI technology to include multi-step voice-controlled automation ready for use in next year’s iOS 18 operating system.

While Apple users can currently make use of multi-step actions on their iPhones, they must manually create these workflows by using the Shortcuts app.

In August, a report by Taiwan’s Economic Daily News said Apple manufacturing partner Foxconn will be the exclusive supplier for servers that Apple will dedicate to training and testing artificial intelligence services. Foxconn already provides most of Apple’s data center servers.

Foxconn will build the servers in a facility in Vietnam. The firm currently also provides servers for several other companies’ AI applications, including Amazon, Nvidia, and OpenAI.