Apple has reportedly acquired a French AI startup specializing in computer vision technologies that run locally on-device with no cloud processing, according to the French publication Challenges.
The AI startup in question went by Datakalab, a small company with less than 20 employees based in Paris. The acquisition was finalized in December, but Apple informed the European Commission of this month’s purchase. Datakalab’s expertise mainly revolved around miniaturizing Large Language Models (LLMs) to be small enough to operate locally. Datakalab also worked on computer vision and image analysis technologies that use AI for object detection and feature extraction.
Apple is certainly interested in LLM miniaturizing and compression techniques as evidenced by its earlier acquisition of DarwinAI, a Canadian AI startup that worked on similar technologies to Datakalab. Both of these acquisitions will strongly help Apple develop AI features that completely run on-device, a direction the company is rumored to pursue with AI capabilities heading to iOS 18.
Datakalab’s experience in creating high-efficiency deep learning algorithms that are low-power enough to run locally has caught the attention of several customers before Apple. The AI startup’s novel expertise earned it projects with Disney and the French government in the past. For example, its project with the French government involved automatically and passively monitoring attendees in a cinema hall to confirm that everyone was wearing a mask. Datakalab also touted the privacy of its on-device AI models for this project, something that is certain to have caught Apple’s interest.
The on-device approach means that most AI features heading for iOS 18 will likely be accessible offline and more private since users’ data won’t have to be processed server-side. On-device AI features will also run faster as they are not susceptible to the quality of an internet connection.
However, on-device AI has a few drawbacks. More sophisticated AI features like text generation and other forms of content creation still require some cloud processing. The most demanding iOS 18 AI capabilities could also be limited to the latest iPhone hardware with more capable Neural Engines and GPUs, meaning that some users with older iPhone models could miss out on accessing those features.
In all cases, the backend for cloud-processed iOS 18 AI features likely won’t come from Apple itself as the company is said to be in talks with providers such as OpenAI, Google, and Baidu to use their AI infrastructure and cloud-based machine-learning models. Recently discovered code also suggests that Apple is preparing an AI-based browser assistant for Safari that could help summarize content from webpages and more.
It’s all but confirmed that Apple will bring forward a slew of new AI-based features across the board with iOS 18, macOS 15, and more at WWDC 2024. Apple aims to introduce AI capabilities within every corner of iOS 18 as the company has largely remained silent features-wise on today’s AI frenzy that has swept the market.
Look out for AI-powered features spanning Siri, Messages, Spotlight Search, Apple Music, Xcode, and more. iOS 18 is also said to introduce considerable UI design changes across the board, with these changes said to make the update one of the biggest in iPhone’s history.
