Apple and Google have both entered the generative AI fray, showcasing features that will bring AI features to all our smartphones. While this means that you can probably do much more with your upcoming phones, it also brings some risks to the table.
What can your phones do with AI?
On 10 June, Apple unveiled ‘Apple Intelligence’—a suite of features that bring generative AI to iPhones, iPads and Macs. As a user, you can transcribe recorded calls, generate emails and summarize notes, erase objects from images, create illustrations and animations at will, and make emoticons. In September last year, Google had already unveiled similar features, which also include live transcription of audio recordings. Samsung, too, is in the fray—the Korean electronics firm offers a native AI model running on its flagship devices to process audio recordings and phone calls, as well as summarize webpages.
Are all these features running locally?
Not all of them so far. Apple clarified that while Siri, its on-device digital assistant, can process basic queries on your device, a broader query may need to access OpenAI’s ChatGPT based on the GPT-4o multimodal AI model. Google’s Pixel phones rely on an internet connection to access large AI models hosted on cloud servers for many of its features. Samsung, too, does the same—the company offers a setting where users can choose to only use local AI features such as live transcribe a voice note, or access an on-cloud AI model to generate summaries of recorded voice notes.
Will all phones get generative AI soon?
Apple’s AI features will come only to its ‘Pro’ iPhones for now, although all of Google’s latest phones support its AI chops. Samsung only has AI on its flagship Galaxy S24 series, but may expand to new devices next month. It all depends on your phone’s processor—for now, only flagship ones by Qualcomm and MediaTek will support local AI processing.
So, what is the real trade-off here?
Privacy experts have raised concerns around what could happen with all the personal data that each local AI model can see on any phone. Apple claims that no personal ‘contextual data’ will be transferred online, and privacy will be maintained on OpenAI’s servers. Google and Samsung claim they do not collect any sensitive personally identifiable information from smartphones, but they do take ‘anonymized metadata’. Audits are yet to take place on whether all the personal information is really local.
Is privacy the only concern?
No. AI is prone to hallucinations and even misinformation, while local ‘narrow’ AI models may not have the full context of understanding, so texts and images generated by them could significantly vary. Local AI models don’t need the internet. But without internet they could turn into black boxes of unexplained decision making. Equally, with internet, privacy would be a major concern. This is why most companies are playing safe with the basic features, which may be of limited value. (edited)