Apple Using In-House Generative AI Chatbot

Apple employees are using the tech giant’s in-house generative AI, nicknamed Apple GPT, to enhance their productivity.

Bloomberg reported last week that Apple has been creating a framework through which to build its own large language models (LLMs), named Ajax. Ajax enables Apple to potentially develop a generative AI product to rival OpenAI’s ChatGPT, Google’s Bard or Microsoft’s Copilot.

Before Ajax results in commercial products, however, Apple has reportedly built its internal ChatGPT-like solution to streamline employees’ workflows. The solution helps Apple workers “prototype future features, summarize text and answer questions based on data it has been trained with,” according to Bloomberg‘s Mark Gurman.

The news follows two months after May’s revelations that Apple had restricted the internal use of ChatGPT and Copilot due to concerns around data security.

Apple reportedly isn’t sure how best to use and sell generative in B2C or B2B capacities yet, although Bloomberg‘s sources stated that Apple would make a major AI-based announcement next year.

However, Bloomberg‘s report offered insight into its prospective uses internally beyond just Apple GPT. One suggestion was its capacity to aid AppleCare support staff in helping Apple customers.

Ajax is a cross-company project encompassing Apple’s software engineering, machine learning and cloud engineering teams.

What is Apple’s Relationship with AI?

Apple has restricted the internal use of chatbots, including ChatGPT and Copilot, because of its concerns about data security and the limited protections offered by current generative AI products.

A central issue around OpenAI’s ChatGPT specifically is that the most widely used generative AI service consumes user prompts to develop and improve its model unless users deliberately opt-out. This has triggered worries that employees might inadvertently include proprietary or confidential data or information in their prompts, which ChatGPT extracts to answer future queries.

For example, if an Apple employee was to accidentally include confidential information about the latest iPhone in their prompt, ChatGPT could store that information without realising the significance of its value and then circulate it in subsequent answers to other users’ prompts.

Microsoft’s recently announced Bing Chat Enterprise seeks to address these types of concerns by being deliberately designed to forego saving the chat data — and consequently, the data cannot be extracted to train AI models. No one else can view the user’s prompts and, therefore, their data.

However, while this chat data erasure is a welcome answer to questions about generative AI’s security, it doesn’t address other issues around the trustworthiness of generative AI’s answers to prompts. Misinformation or outright lies are still rife across commercially available AI products.

Apple CEO Tom Cook stated in the tech giant’s May earnings call that “a number of issues need to be sorted” with AI, signalling that Apple might be more cautious with releasing AI products in imperfect states than its competitors.

However, these reports about Apple GPT and a significant consumer-centric AI announcement being touted for next year suggest Apple are equally cautious about being left behind by the likes of Microsoft, Meta and Google in the AI market. Likewise, Apple’s hiring of Google’s former head of AI, John Giannandrea, in 2018 illustrates the company has been exploring AI’s possibilities long-term.

Apple’s Tentative Steps into the Enterprise Tech Market

Apple Products such as Macs and iPhones have naturally always included enterprise capabilities, but there have been several significant steps into business technology in earnest in 2023. These early signs of a potential B2B or B2C AI product next year maintain that pattern.

The most high-profile example was Apple’s Vision Pro “spatial computer” device, announced at this year’s WWDC. Apple’s announcement emphasized that the Vision Pro “frees apps from the boundaries of a display so they can appear side by side at any scale”,  allowing users to “be even more productive, with infinite screen real estate, access to their favourite apps, and all-new ways to multitask”.

The Vision Pro is supported by Apple’s Magic Keyboard and Magic Trackpad. Users can create a wireless workspace by integrating the Vision Pro with their Mac and can create a large 4K work display within their Pro headset. The Vision Pro’s FaceTime deployment utilises spatial computing and spatial audio so that everyone on a call is reflected in life-size tiles and speaking as if from where they are positioned.

Also at WWDC, Apple announced plans to turn the iPhone into a huddle camera through integration with Apple TV. The feature utilizes Apple’s Continuity Camera mode, allowing iPhones and iPads to function wirelessly as webcams.

Apple also unveiled new video conferencing features for the upcoming macOS to improve meetings across all platforms. macOS Sonoma will include Presenter Overlay, enabling users to place themselves on top of content such as slideshows. They’ll have the option of appearing as a small bubble or in a larger frame, and the feature removes the participant from their background and places the content behind them.

Meanwhile, in April, Apple filed a next-generation camera system patent for FaceTime with advanced enterprise collaboration possibilities, such as the capability to parse in-air gestures and process physical drawings into digital media.

The tech giant published a patent application form that included a new conferencing interface aiming to improve the user experience of those employing FaceTime for business or education purposes. The solution aims to allow easier sharing of content during video meetings.