Hybrid AI is the way ahead to make artificial intelligence more practical on smartphones: Samsung’s Won-Joon Choi

While Samsung sells smartphones across all price categories and questions around privacy of hybrid AI dependency may gain prominence, other brands like Apple and Google Pixel, despite launching only premium smartphones, have taken a similar approach.

Samsung is also looking at bringing full-range Galaxy AI services to devices in the budget-friendly Galaxy A series smartphones as well.

Artificial Intelligence (AI) in smartphones raises many questions across industries that current generative AI tools won't have an answer for. Of course, there's the talk around personal data privacy but more importantly how mobile phone makers should approach AI while integrating ChatGPT-like services on handsets? The reliability, speed and practicality of generative AI tools on smartphones depend a lot on the hardware prowess.

While on-device AI seems ideal, it may not be economically viable for all smartphone brands to deliver useful AI tools in mobile phones across all price points, specially in budget and mid-level categories. At the same time, not everyone would be willing to increase their smartphone buying budgets every year to get a powerful flagship smartphone. Now, the dilemma– how to offer useful AI services on affordable smartphones- a segment which constitutes the largest set of mobile buyers, in countries like India?

Bring AI to smartphones across all price points

Samsung has a clear answer– “Hybrid AI”. Not just for its flagship Galaxy S or foldable Galaxy Z series, Samsung is also looking at bringing its Galaxy AI services to devices in the budget-friendly Galaxy A series smartphones as well. To offer a seamless experience- some part of AI processing happens locally on the user's device while certain complex functions require the user's smartphone to be connected to the internet so that it happens over cloud. For Samsung, forming a clear “case-by-case” policy as to what happens on-device and what data processing happens on the cloud is the priority.

“Our users are now experiencing the tangible benefits of Galaxy AI, innovated to provide an entirely human-centric mobile AI experience – and that was just the beginning. Within months, we have democratised Galaxy AI, extending access to more models and form factors. It's a feat only possible through close collaboration with our industry partners and our distinct approach to hybrid AI to bring our users the most practical, reliable and versatile mobile AI experience possible. An experience that has the potential to change lives,” said Won-Joon Choi, Executive Vice President and Head of Mobile R&D Office, Mobile eXperience Business, Samsung Electronics at a brainstorming session post the recent Galaxy Unpacked 2024 event in Paris.

Google supplies Android OS, hardware comes from partners like Qualcomm and then Samsung talks about hybrid AI where user’s personal data may be processed in servers that they will have little control over.

Google supplies Android OS, hardware comes from partners like Qualcomm and then Samsung talks about hybrid AI where user’s personal data may be processed in servers that they will have little control over.

Won-joon Choi heads Samsung's ​flagship R&D team in the mobile Communications division and ​helped the company become ​the world's first device maker to ​commercialize 5G network-enabled​ smartphones in 2019.

At the AI session, Jenny Blackburn, Vice President of User Experience for Gemini Experiences & Google Assistant, said, “To create something that works for everyone, using different devices in a range of situations, that's something Google and Samsung have been figuring out together since the beginning of the Android platform. It was no different when it came to Gemini and Galaxy AI. It all comes down to zooming in on how people use their devices.”

 

How can users trust hybrid AI models?

Google supplies Android OS, hardware comes from partners like Qualcomm and then Samsung talks about hybrid AI where user's personal data may be processed in servers that they will have little control over. The next big question- how can this be trusted when there are too many “players” involved in a hybrid setup.

“We need regulations that can adapt as AI evolves. Continuous research and monitoring like what is provided by the OECD.AI Policy Observatory, will ensure informed decisions and help policymakers to keep up with developments. We work directly with leaders in government and major tech companies alike. Samsung is well-placed to actively participate in discussions about AI best practices that promote human-centric, safe, secure and trustworthy AI,” said Lucia Russo, Economist and Policy Analyst in the Artificial Intelligence (AI) unit, OECD Digital Economy Policy Division.

While this statement itself may not be enough to explain how trustworthy hybrid AI will be, what's interesting is that tech companies are willingly accepting that AI policies need continuous reviews and monitoring.

“Our responsibility as leaders of mobile AI is to closely work together in partnership with industry and regulatory leaders to build AI that brings positive impact to society. At Samsung, we understand our users. What they want and how they behave. It's those insights that have allowed us to create truly human-centric mobile AI, while also protecting users' right to privacy. Right now, this is our foremost concern because in this age of hyperconnectivity, privacy is more important than ever,” said Daehyun Kim, Executive Vice President of the Samsung Research Global AI Center, Samsung Electronics.

While Samsung sells smartphones across all price categories and questions around privacy of hybrid AI dependency may gain prominence, other brands like Apple and Google Pixel, despite launching only premium smartphones, have taken a similar approach. For example, Apple's optional ChatGPT feature for generative AI content along with features like NightSight Video on Pixel flagship phones also has dependency on hybrid AI and third-party partnerships.