본문 바로가기
Post · 2026.02.17

How Sim-to-Real changed robot learning

This page is a translation of the Korean original. Read the Korean original →

Lately, as I browse various IT news, robots keep coming up alongside AI as a hot topic. It naturally takes me back to the early 2000s, my time in the Samsung Software Membership program. Back then, getting a robot to simply “walk properly” was a massive challenge in itself. Developers pulled all-nighters manually tuning the parameter values of each motor one by one, and it took thousands of tests just to get the robot to take a few steps.

Now, things are very different. Thanks to dramatic advances in Sim-to-Real (Simulation to Real) technology, an AI pre-trained on tens of thousands of trial-and-error runs in a virtual environment can be deployed onto a physical robot and achieve a high level of motion within a relatively short time. Several technical hurdles still remain, but even finger control — once considered one of the hardest problems — has rapidly progressed beyond two- or three-finger grippers to the fine, dexterous movement of all five fingers.

Watching this shift, my first thought wasn’t simply “the technology has advanced,” but rather why the pace changed so much. In the past, physical experimentation was almost the entire process; now, large-scale simulation infrastructure and computing resources underpin it. And it occurred to me that this foundation was only possible because of massive capital investment.

Reflecting on my experience managing Samsung’s global R&D portfolio, it seems technology is no longer a question of “is it possible,” but increasingly one of “how fast can it be realized.” That led me to think the core engine determining that speed is resources — specifically, how capital is allocated.

Extending that thought a bit further, the market sends strong signals about the direction of technology. But capital doesn’t flow automatically. How to interpret that signal and where to focus is a matter of corporate strategic judgment.

  1. The maturity of a technology correlates strongly with the market’s capital investment.
    Robotics and AI technology haven’t evolved at this unprecedented pace simply because time has passed. The moment strong market need is confirmed, massive capital pours in, pushing the technology’s maturation speed past a critical threshold.

When you think about it, it’s simple. Companies concentrate wherever capital flows, and human and material resources follow. The speed at which a technology is realized is closely tied to how much capital is allocated to it, and how intensively.

That said, even given the same market signal, not every company moves at the same speed. Some invest capital aggressively, others watch and wait or respond selectively. What creates that difference is each company’s internal strategic interpretation and its appetite for risk.

  1. Technology strategy ultimately comes down to ‘where you spend the money.’
    When you’re running R&D, the term “technology strategy” can sound quite grand. But what I’ve felt on the ground is a bit different. Technology strategy is really about deciding where, in what order, and for how long to commit limited resources.

For example: do we build up our internal R&D capability further, or partner with outside startups or universities to move faster? Since you can’t do everything yourself, you have to make a choice at some point. That choice is precisely the starting point of technology strategy.

Another dilemma is whether to launch a product quickly to seize the market right now, or to patiently build up core technology even if it takes longer. Choosing the former may mean accepting a certain amount of technical debt; choosing the latter risks missing the immediate opportunity.

Technology strategy is a judgment about where to invest capital: whether to invest over the long term to build technology that will become a future barrier to entry, or to concentrate resources on increasing speed right now.

Companies that have built competitiveness by accumulating technology over decades show traces of patiently allocating capital in one direction. High-growth companies, by contrast, often choose speed over perfection, accepting a certain amount of technical debt to capture the market first. What matters isn’t whether debt is left on the books, but whether the company has a structure in place to manage and repay it.

And even given the same market opportunity, different companies arrive at entirely different capital allocation strategies depending on their resource structure, organizational culture, and management judgment. In the end, technology strategy is less about simply following the market’s signal, and more a process of interpreting and choosing in response to it.

A technology challenge isn’t simply engineering — it’s a comprehensive system in which budget, organization, and decision-making structure are all interlocked.

  1. Market signals carve the channel through which capital flows.
    In the past, technology tried to lead the market; now it feels closer to a structure where market need summons capital, and that capital accelerates the technology. But it’s the company’s strategic choices that determine the direction and intensity of that flow.

Every technology starts from the question, “What value will this deliver to the customer?” Wherever capital concentrates, a clear market objective exists. But how each company chooses to realize that objective, and how fast it’s willing to bet, differs from firm to firm.

Personally, I don’t think technology can be judged with a simple binary of “it works” or “it doesn’t.” But when sufficient market conviction combines with a company’s will to interpret that conviction and act decisively, technology has consistently matured faster than we expected.

If a given technology is realized more slowly than expected, it’s worth asking whether that’s purely due to technical limitations, or because market conviction and corporate strategic betting haven’t yet been sufficient.

From this perspective, isn’t an investment decision on realizing a technology not merely a cost outlay, but an interpretation of the market’s signal — and a piece of corporate strategy in itself?

※ This piece reflects my own views, though I used AI assistance to polish the writing for readability.
※ Different perspectives and opinions are always welcome.

#TechStrategy #CapitalAllocation #RNDManagement #AI #Robotics #TechStrategy #CapitalAllocation #Innovation #Leadership

Tips on this topic

Tips →

More on this topic

Continue the conversation on LinkedIn →