Artificial intelligence does not need to be louder, flashier, or more complex to matter. According to TM Roh, President and CEO of Samsung Electronics, AI's real success will come when it fades into the background and becomes a dependable part of daily life.
In a thought piece outlining Samsung's long-term AI philosophy, Roh argues that the industry is at risk of focusing too heavily on short-term excitement instead of long-term usefulness. Drawing on the idea that transformative technologies eventually become invisible, he positions AI not as a novelty but as infrastructure, something users rely on without having to think about it.
From Invention to Infrastructure
Roh notes that most new technologies follow a familiar curve. They begin as expensive, experimental innovations before becoming reliable, affordable, and broadly usable. When that transition happens, the technology often stops attracting attention, not because it is less powerful, but because it simply works.
AI, he suggests, is now at that inflexion point. While awareness and usage are already high, the more important question is whether AI is genuinely practical in everyday scenarios. For AI to function as infrastructure, it must deliver consistent results without demanding constant input, learning, or adjustment from users.
Practical Use Over Technical Brilliance
According to Roh, usefulness matters more than raw capability. Early AI wins, such as language translation, exposed this gap clearly. Systems that perform well only in major global languages or controlled environments fall short in real-world use, where dialects, accents, and cultural nuance matter.
The same applies to accessibility features like real-time captions, image descriptions, and simplified summaries. These are not optional enhancements, he argues, but core requirements if AI is to help people understand, communicate, and act with confidence.
Trust, Privacy, and Everyday Adoption
One of the biggest barriers to AI becoming universal infrastructure is trust. Roh highlights that AI increasingly operates in highly personal spaces, from messages and photos to financial and health-related data. If users feel they must give up control or privacy to benefit from AI, adoption will slow.
For AI to scale responsibly, security, transparency, and user choice must be built in by default. These elements, he says, should not be positioned as premium features but as foundational expectations.
The Next Phase: Agentic AI
Looking ahead, Roh points to agentic AI as the next major shift. Unlike today's assistants that respond to queries, agentic systems will complete tasks end-to-end, organizing information and handling routine actions with minimal user intervention.
If designed well, this could further reduce friction in everyday digital interactions. However, the same principles apply: the best AI will be the least visible, working quietly in the background rather than demanding attention.
Roh's central argument is that AI's true value will not be measured through benchmarks or model comparisons. Instead, it will be reflected in small, everyday moments, which is when technology helps people understand more, participate more easily, and move through the world with less effort.
For Samsung, and the wider industry, the challenge is no longer teaching people how to use AI. It is building AI that people do not have to think about at all
