Almost two years after OpenAI released its industry-changing GenAI chatbot ChatGPT, Apple unveiled a series of on-device AI offerings on Monday (10 June).
During Apple’s annual Worldwide Developer Conference, the company announced plans to fit every new iPhone, iPad and iMac with on-device GenAI, under the umbrella of Apple Intelligence.
The new system will use machine learning (ML) to personalise messages and notifications depending on the user’s behaviours and preferences, as well as offering writing tools across every app to aid in proofreading and clarity.
“AI has to understand you and be grounded in your personal context like your routine, your relationships, your communications and more. It is beyond AI. It is personal intelligence,” said Cook.
Apple also revealed that ChatGPT will be integrated into its voice assistant Siri for responses that cannot be carried out solely by Apple Intelligence. According to the company, the technology will be free to use and will not require users to create an account. Apple also claimed users’ requests and data will not be logged.
How has Apple fallen behind in the AI race?
Apple’s announcements come amid concerns that the company and its devices had fallen behind in the AI race, trailing behind Google and Microsoft, which quickly reacted to ChatGPT’s release with their own AI plans.
Microsoft surpassed Apple as the most valuable technology company in the world earlier this year, following the large AI Copilot push into all of its applications and multi-billion dollar investment into OpenAI.
Apple has also seen global sales of its devices, particularly of its iPhones, slow over the past year.
On-device AI privacy could set Apple apart
In an attempt to set itself apart from the rest, Apple CEO Tim Cook’s keynote address kept privacy at the forefront of the company’s announcements.
As AI grows and is rapidly deployed into more areas of society, business and industry, governments and regulators have rushed to govern the emerging technology.
At the beginning of April, the UK and US signed a memorandum of understanding in a partnership to tackle AI safety and ethics. Last month, OpenAI announced it had formed a safety committee ahead of training its biggest AI model yet.
This has set the stage for Apple to market itself as a responsible tech giant taking on AI and delivering a safer way for users to interact with language models.
Apple Intelligence will carry out most of its computer processing on-device rather than in data centres, where a user’s personal information is more susceptible to hackers and bad actors, the company said.
For more advanced AI requests that require more computer processing, Apple has created a cloud network completely powered by semiconductors. According to the tech giant, this is a more private option as the data is not stored or accessible by anyone, including the company.
However, Sam Ward, head of research and development at enterprise AI company Enate, said it is not clear what processing will be done on the device itself and what will be sent to the cloud, especially with the implementation of OpenAI’s technology in Siri.
“It could simply be some sort of data preparation or it could include features like anonymisation, which would elevate some privacy concerns,” Ward said.
“Naturally there are concerns about the data that is being shared with OpenAI and Apple,” he added.
Suneet Muru, associate analyst at research and analysis company GlobalData, told Verdict that Apple’s competitive position in AI hinges on how well it can weave Apple Intelligence into existing user workflows.
“GenAI has typically felt like a stand-alone application, dropped into user interfaces without any real integration,” Muru told Verdict. “The natural question was, therefore, which company would transition GenAI out of the demo phase and make it a more meaningful part of existing app ecosystems.
“With Intelligence, Apple might have done just that.”
The future of GenAI with Apple in the race
According to forecasts from GlobalData, the global AI market will be worth more than $909bn by 2030.
In the GenAI space, revenues are expected to grow from $1.8bn in 2022 to $33bn in 2027, a compound annual growth rate of 80%. GenAI will impact every industry and become a catalyst for broader AI capabilities such as ML and computer vision, according to the research company.
Muru believes it is unlikely that GenAI features will be the sole reason to use a particular device for long periods of time. Discussing Apple’s position in the AI race following its announcements, Muru said that Google will remain a competitive force.
“Gemini is tailored to Google’s ecosystem the same way Apple Intelligence is tailored to Apple’s ecosystem,” Muru said. “It is unlikely that AI features will ever be the reason to use a particular device in the long run.
“GenAI integration, in its current state, simply complements the already existing user interface, so doesn’t give users many reasons to jump ship.”
Greg Hanson, group vice-president of EMEA North at Informatica, said that while the launch of Apple Intelligence will succeed in putting AI into the hands of consumers, a myriad of challenges remain for GenAI.
“[Apple Intelligence is] set to change the way that users discover and consume data through simple conversational interactions,” Hanson said.
“However, it is important to remember that these are still the early green shoots of Gen AI,” he added. “A lot still needs to be ironed out – users need to be educated on how to best ask questions that generate valuable answers and will want greater transparency about how their personal data will be used to feed and power AI models.”
As companies continue to train more powerful large language models and bring new AI services to consumers, Hanson warned that there are no shortcuts in AI development.
“Consumers and regulators will demand transparency, interpretability and ‘explainability’ from AI systems,” Hanson said. “Developing an application that lives up to the promise of its name will require holistic, high-quality and well-governed data foundations with adequate data access management and privacy controls baked-in.”