For years, the artificial intelligence conversation has largely centered on one question: how to build bigger and more capable models. But as AI increasingly moves beyond screens and into robots, autonomous systems, and real-world machines, a different question is emerging. Can model scale alone solve the challenges of reliability, safety, and decision-making in physical environments, or does Physical AI require a fundamentally different approach?
Physical AI Is Challenging a Common Assumption About AI
The recent rise of generative AI has made deep learning the dominant public face of artificial intelligence. Large language models, multimodal systems, and increasingly powerful foundation models have driven much of the industry’s attention, investment, and technical ambition.
However, some engineers working on physical systems argue that equating AI with deep learning risks oversimplifying the broader field.
According to Raymond Kim, founder and CEO of AidALL, a Seoul-based robotics company focused on autonomous mobility technologies, the industry’s current understanding of AI is increasingly shaped by market narratives rather than the full history of intelligent systems.
“Since ChatGPT, ‘AI’ has become almost synonymous with large deep-learning models,”
Kim told KoreaTechDesk in an exclusive interview.
“That equation is a market shortcut, not a definition.”
Kim then points to technologies such as fuzzy-logic control systems, automotive anti-lock braking systems (ABS), aircraft fly-by-wire systems, and Mobileye’s Responsibility-Sensitive Safety (RSS) framework as examples of AI-related systems that have operated in safety-sensitive environments for decades without relying on modern deep-learning approaches.
The distinction becomes increasingly important as AI systems begin interacting directly with the physical world.
“The shortcut works inside a screen, but it breaks the moment a system steps into the physical world,”
Kim said.
Why Physical AI Raises Different Technical Questions
Physical AI generally refers to AI systems embedded in machines that perceive, reason, and act within real environments. This includes autonomous robots, industrial systems, mobility platforms, and other intelligent machines operating outside controlled digital settings.
Unlike purely software-based applications, Physical AI systems must function under strict constraints involving power consumption, latency, environmental uncertainty, and real-time decision-making.
Industry attention around Physical AI has grown significantly over the past year. NVIDIA has described Physical AI as a new category of intelligence that enables autonomous systems to understand and interact with the real world, while researchers and technology companies continue exploring architectures optimized for robotics and edge computing.
For Kim, the challenge is not finding a single superior model.
“At AidALL, we treat this as an architectural question, not a philosophical one.”
Moving Beyond Deep Learning Alone
One of the emerging themes in Physical AI research is the growing interest in combining multiple AI paradigms rather than relying exclusively on deep-learning systems.
Kim said AidALL’s internal architecture combines deterministic computation, neuromorphic design principles, and a neuro-symbolic reasoning layer.
“We combine deterministic computation, neuromorphic design principles, and a neuro-symbolic reasoning layer into a single system,”
he explained.
Neuro-symbolic AI has attracted increasing attention among researchers seeking to combine the pattern-recognition strengths of neural networks with the reasoning capabilities of symbolic systems. Research organizations such as IBM have described neuro-symbolic approaches as a way to improve explainability, reasoning, and decision transparency in complex AI systems.
Meanwhile, neuromorphic computing draws inspiration from biological neural systems to improve efficiency and reduce computational demands. Researchers increasingly view neuromorphic architectures as a promising direction for robotics, edge devices, and resource-constrained environments where energy efficiency is critical.
These approaches do not necessarily replace deep learning. Instead, they suggest that future Physical AI systems may require a broader toolbox.
“The shift in Physical AI isn’t about finding a new model. It’s about asking which paradigm fits which problem.”
Why Reliability Matters as Much as Intelligence
The conversation becomes more complex when AI systems are expected to make decisions that directly affect physical outcomes.
Public AI benchmarks often focus on model accuracy, reasoning performance, or task completion rates. Yet real-world deployment introduces additional requirements that benchmark scores may not fully capture.
The International Organization for Standardization’s ISO/IEC TR 5469 guidance on AI and functional safety notes that machine-learning systems can create challenges around explainability, repeatability, and assurance in safety-related applications.
Kim argues that some critical system properties cannot simply emerge through additional data or larger models.
“The most repeated misconception is ‘scale solves it,’”
He believes certain requirements sit outside the normal scaling curve of contemporary AI development.
“Determinism isn’t something you reach by adding statistical resources. It’s a property of the computational structure itself.”
This perspective reflects a broader discussion emerging across robotics, autonomous systems, and edge AI communities. While larger models continue to improve perception, language understanding, and generalization, engineers increasingly face questions about traceability, predictability, and reliability when those systems interact with physical environments.
A Broader Debate About the Future of AI
The discussion around Physical AI is not ultimately about choosing one AI paradigm over another.
Deep learning remains central to many of the industry’s most significant advances. At the same time, researchers and engineers working on real-world systems are increasingly examining how different computational approaches can complement each other.
That shift is already visible in areas such as robotics, autonomous mobility, industrial automation, and edge intelligence.
And for companies building systems that must operate beyond controlled digital environments, the next phase of AI development may depend less on model size alone and more on how different forms of intelligence are combined into reliable architectures.
The Architecture Question
The history of AI has often been described through breakthroughs in models. But today, Physical AI may push the industry toward a different conversation.
As intelligent systems move into machines that navigate, assist, and act in real environments, the central challenge may no longer be how large a model can become. It may be how reliably an entire system can perceive, reason, and respond when conditions are unpredictable and mistakes carry real consequences.
And in the end, for engineers building the next generation of Physical AI systems, that distinction is rapidly becoming a defining engineering reality—one that will shape which technologies succeed in the physical world and which remain confined to the lab.
Key Takeaway
- Physical AI is shifting attention beyond model scale toward system architecture, reliability, and real-world operation.
- Raymond Kim of AidALL argues that AI should not be treated as synonymous with deep learning, particularly in physical systems.
- Deterministic computation, neuromorphic computing, and neuro-symbolic AI are emerging as important complementary approaches in Physical AI development.
- Research and industry interest in Physical AI continue to grow as AI expands into robotics, autonomous systems, and edge computing.
- According to Kim, some critical properties such as deterministic guarantees cannot be achieved simply through larger models or more training data.
- The broader industry debate is increasingly focused on which AI architectures are best suited for real-world deployment rather than model scale alone.
🤝 Looking to connect with verified Korean companies building globally?
Explore curated company profiles and request direct introductions through beSUCCESS Connect.
– Stay Ahead in Korea’s Startup Scene –
Get real-time insights, funding updates, and policy shifts shaping Korea’s innovation ecosystem.
➡️ Follow KoreaTechDesk on LinkedIn, X (Twitter), Threads, Bluesky, Telegram, Facebook, and WhatsApp Channel.
