Mark Cuban: Beware of AI Replacing Trial-and-Error Skills
Mark Cuban wrote about the current discussions surrounding AI and mathematics education, drawing parallels to the acquisition of "domain knowledge" in the business world.
He suggested that in companies, the most important ability for new employees is not how much specialized knowledge they have mastered, but rather their ability to "figure out how to figure things out".
He emphasized that domain knowledge is accumulated through the repeated experience of "finding solutions on one's own", rather than being directly told the answers.
Based on this, he warned that companies must carefully continue to encourage and train employees to independently explore problem-solving skills, to avoid skipping or replacing this critical growth stage with tools.
He compared this business observation to the ongoing discussions in the AI/mathematics education field—the core concern is the same: when tools (whether AI, calculators, or formulas) can directly provide answers, do people still have the opportunity to truly exercise independent problem-solving thinking processes.
From an organizational perspective, the implied stakes of this viewpoint are: if companies use AI to directly replace the process of junior employees "figuring out problems", they may indeed improve efficiency and reduce labor costs in the short term, but in the long term, they will cut off the path for cultivating senior domain experts, leading to a gap in the future talent pipeline; whereas companies that insist on allowing newcomers to experience the "trial-and-error" process are more likely to maintain their ability to nurture senior talent, gaining a structural advantage in the competition between AI and talent development.
Source: Public Information
ABAB AI Insight
Mark Cuban's own career experience exemplifies the logic of "figuring out problems through practice"—he is not a formally trained technical expert, but started from selling software and running a bar, accumulating cross-domain business judgment through continuous trial and error, ultimately founding Broadcast.com, which was acquired by Yahoo in 1999 for about $57 billion in stock. This experience itself reflects his assertion that "domain knowledge is accumulated through on-site exploration".
As an investor on Shark Tank and owner of the NBA's Dallas Mavericks, Cuban has long preferred startups whose founders have practical experience in "figuring out business models from scratch", rather than just teams with prestigious backgrounds or purely theoretical knowledge. This is a concrete manifestation of his capital allocation logic that values "trial-and-error skills".
This aligns with the long-standing logic of cultivating "T-shaped talents" in Silicon Valley, as well as the historical practices of institutions like Goldman Sachs and McKinsey, which have insisted on using high-intensity practical rotations instead of purely theoretical training to train junior analysts—true professional judgment is often not directly given by classrooms or tools, but learned through repeated trial and error in real scenarios. The industry is currently at a stage where AI tools are massively intervening in basic information retrieval and preliminary analysis tasks, potentially compressing the window for such "practical trial and error".
This is essentially the hidden risk brought by technological substitution: when AI begins to replace the process of "figuring out problems" originally undertaken by junior employees, companies may seem to improve short-term efficiency, but in reality, they may cut off the necessary path for cultivating future senior domain experts. The mechanism behind this is that domain knowledge is not static information imparted once, but a dynamic capability formed through repeatedly facing uncertain problems and independently seeking paths. Once this learning process is taken over by tools, the talent pipeline within the company will experience structural gaps.
ABAB News · Cognitive Laws
Answers can be provided by tools, but judgment cannot.
A generation that skips trial and error will become a generation of gaps.
Domain knowledge is not learned; it is discovered through exploration.