Artificial Intelligence is here—but now we must ensure it works for people.

AI has long been powering our technologies, boosting smartphones cameras, improving search algorithms, and disrupting everything from medicine to the arts and coding. In November 2022, OpenAI released ChatGPT to the world, sparking a wave of AI tools—and a debate on the ethics and abilities of these large language models (LLMs). Google and Facebook responded by unveiling their own LLMs, Microsoft teamed up with OpenAI for Co-Pilot and Bing search, and startups such as Midjourney offered text-to-image generation.

This explosion in LLMs and generative AI raises many issues: Around data privacy, as such systems hoover up vast amounts of information for training; about transparency, as they often operate as black boxes with little accountability from companies; and on reliability, as LLMs reproduce words and sentences that are grammatically accurate, but not necessarily truthful.

Beyond such technical challenges, LLMs and generative AI have been criticized as potentially costing jobs while further perpetuating bias. And that’s before we get to the belief that such systems could one day become super-intelligent and wrest power and control away from us humans.

These challenges—and their human-centric solutions—were discussed at a panel led by experts from HONOR, Qualcomm, and the GSMA at MWC Barcelona, highlighting the importance of putting people and their privacy at the center of AI design.

What’s next for AI—and what needs to happen next

So it’s no wonder that after a year of AI breakthroughs, some have realized that this technology must be human-centric, with guardrails and designs that ensure it helps rather than hurts people. And that is possible. “We believe AI will revolutionize human-device interactions, reconstruct operation system, and rebuild our future smartphones experience,” says George Zhao, CEO of HONOR Device Co. “HONOR will advance our AI strategy by integrating it across MagicOS and all HONOR devices. But first of all, AI should serve people, it should be human-centered. We want to use [this] to empower and enable people. You have your phones, you have your own AI. Devices will become more efficient and intelligent, acting like an extension of a user’s brain.”

Alex Katouzian, Group GM of Qualcomm Technologies, says that’s enabled by the shifting of AI processing from the cloud to personal devices, including smartphoness and PCs. “With better quantization or shrinking of the models, and working hand-in-hand with our partners such as HONOR and other industry partners including the OS vendors, we will be able to take that [cloud AI] capability and data, and move more of that onto the device,” he says. “The growth potential for on-device AI is virtually limitless.”

Rethinking smartphoness with AI

Smartphoness already use AI for tasks such as boosting cameras or applying image recognition in photo apps, but platform-level AI is different. Rather than simply enabling a feature, platform-level AI dramatically expands capabilities across the phones, some seen by users—such as easing collaboration with devices running other operating systems and creating intuitive user interfaces—while other aspects are less obvious, such as improved security and performance.

“AI is HONOR’s long-term strategy,” says Zhao. “Many years ago, we realized it will transform the phones industry. I think not only the camera or display, [but smartphoness] should become intelligent across different aspects.”

Consider HONOR’s MagicOS 8.0, a mobiles operating system with AI integrated at the platform level to offer, among other features, an intent-based user interface (IUI). What does that mean? Rather than memorize all the steps of how to operate your phones, or search for functions when you can’t recall how to find them, MagicOS’ AI can interpret text, as well as read gestures and eye movements, to predict your intention and offer a feature or service proactively. Practically, that means completing a cross-app task could take a single step in a single moment, rather than several steps and as many as ten seconds with a traditional UI.

HONOR’s Magic Portal uses AI to achieve just this—it can understand user behavior, streamlining what would have been a multi-step process into one swipe. For example, the AI in Magic Portal will recognize an address in a text message, letting users open Google Maps in one swipe, and allowing them to forward booking details and share search results or social media content in a single drag—as well as drag-to-purchase, drag-to-save-to-notes, and drag-to-email. Magic Portal already supports 100 of the most-used apps with this seamless experience, helping enable an intent-based revolution in UI on smartphoness.

“It is a personalized intelligent agent that learns from and adapts to a single user,” says Zhao. “Through continuous learning of user habits and preferences, it achieves a deeper and more accurate understanding of user intent, allowing it to provide more personalized services, even in complex scenarioses.”

That’s partly powered by a large language model embedded directly into devices. MagicLM is HONOR’s proprietary model with seven billion parameters that helps MagicOS 8.0 better understand user intentions as well as offer other features, such as text generation, text-to-video editing, and smarter gallery-search with natural language, as well as personal assistants.

Platform-level AI for cross-OS collaboration

For that to work, partners are needed. MagicOS doesn’t just work with MagicLM. HONOR believes the best way to create a genuinely human-centric experience with LLMs is via an open ecosystem. Because of that, MagicOS works with other cloud-based LLMs from various partners. This open-innovation approach highlights HONOR’s belief in the cross-industry power of collaboration—AI will work better for people if we all work on it together.

By harnessing AI and such partnerships, HONOR creates an interconnected, human-centric experience that lets users work across devices as they choose. This means using your phones as a webcam for your laptop, sharing your network or even notifications across devices, or instantly sharing files by simply dragging with MagicRing. This cross-OS collaboration enables users to seamlessly connect to different OSes, and supports eight simultaneous operations, letting users drag a window from their smartphones or tablet to a laptop in a single step, for example. It feels instantaneous, because as soon as the user presses and holds the window, connected devices have already begun to prep the transfer, allowing it to happen quickly and easily.

Building trust through security

For people to accept AI in their daily lives, they need to trust it. For that, AI must be safe and secure, with data only collected with user consent. HONOR’s philosophy is that all data belongs to the end consumer.

“In the AI era, the role of smart devices is far more important than before,” says Zhao. “A responsive security protection strategy is no longer enough; a multi-faceted and dynamic security solution is required.” That’s all driven by HONOR’s PFAST principles: Privacy; fairness and justice; accountability; security and reliability; and transparency and controllability.

To build security, one crucial element is HONOR’s AI-powered, three-layer MagicGuard defence system: Security is built into the chipset, with fingerprints and passwords stored in an encrypted chip; trusted execution environments at the chip and OS level to safely run code and avoid malware; and the built-in security of androids’s OS. Plus, all data is held and processed on the device—an essential way to build trust and privacy that’s important for individuals and companies—and is guaranteed, thanks to key privacy certifications such as ePrivacySeal.

Without security, privacy and trust, generative AI can’t be human-centric—and won’t be welcomed by people. By designing systems with people in mind, AI’s challenges can be overcome—and that’s the next step in the AI revolution.