[AI Talent War] Who Are They?
AI Super-Talents… Undergraduate in China, PhD in the US
Core Competencies for Future AI Talent: What Should We Teach?
Hard Skills and Soft Skills: Cultivating Powers that AI Cannot Replace
Insight Bridge AI’s Perspective: The Ability to Judge Right from Wrong, Critical Thinking as Future Competency
Silicon Valley is waging an unprecedented war to secure top-tier AI talent. The scale of this war is beyond imagination. According to AI talent recruitment specialist Harrison Clark, salary packages for mid-level AI engineers have skyrocketed from $400,000-$900,000 (approximately 550 million to 1.24 billion won) in 2022 to $500,000-$2 million (approximately 690 million to 2.77 billion won) recently.
For top talent, packages reach around $7 million (approximately 9.7 billion won), with some exceeding $10 million (approximately 13.8 billion won). Meta, which established a superintelligence research lab, recently made statements about offering up to $100 million (approximately 138.4 billion won) in compensation to recruit OpenAI talent.
This global trend brings a sense of crisis to the Korean government, job seekers, and parents with children. If Korea cannot escape the situation where the US and China monopolize top talent, both individual and national competitiveness will inevitably decline.
This ultimately brings us back to the question: “How can I raise my child to become a super-talent?” How can we raise our children to become future AI leaders earning billion-dollar salaries?
Why Bet $10 Billion on One Super-Talent?
The meaning of salaries ranging from 10 billion to 100 billion won is not simply “money” but rather “strategic investment” to secure future technological hegemony. In cutting-edge AI fields like superintelligence research, the ability to create and advance technology itself is required, beyond simply applying technology well. This is being treated not as an HR issue but as capital expenditure (Capex).
“Creative thinking brilliant minds” themselves are becoming the scarcest resource. These are the so-called “Foundational Minds.” They are not simple developers. They are the select few who determine the direction of AI evolution through next-generation large model design, alignment, and multimodal interfaces.
Why go to such lengths? The answer is simple: “One person” can create greater utility than computing resources worth hundreds of billions of won. AI cannot answer the question “why.” But humans can. AI processes data, but it’s human imagination that designs “what to create.”
Who can design superintelligence? Companies are generously investing in talent who ask these questions and find solutions.
Their value assessment criteria are neither sales nor patents. While the core of AI technology lies in algorithms and data, how to combine them and where to apply them depends on one person’s creativity, intuition, and experience.
Their thinking and execution can save tens of thousands of GPUs. AI is unpredictable and we don’t know why it produces certain results. Therefore, having experienced that moment of AI technological innovation is crucial.
Sam Altman, CEO of OpenAI, emphasized the importance of exceptional talent by saying, “An engineer who is 10 times better is great, but an engineer or researcher who is 10,000 times better is even more so.” A Databricks AI VP compared top-level AI talent to superstar basketball player LeBron James, estimating that there are fewer than 1,000 researchers worldwide capable of building frontier AI models.
In other words, such amazing opportunities are open only to extremely limited talent. A small number of experts can become superstars and enjoy wealth and fame. However, the majority cannot. There are observations that polarization is intensifying, with recent college graduates in development and research unable to find jobs at all.
National AI Capability Rankings. South Korea ranked 6th overall, following the United States and China. Rankings were based on criteria such as talent, infrastructure, environment, and research capabilities. (Source: The Global AI Index, Tortoise Media)
Undergraduate in China, PhD in the US?
Top AI talents generally received education in the world’s top 1%, and among them, they are the 0.1% who can design AI’s evolutionary direction. They are people equipped with intuition beyond algorithms, judgment beyond data, and philosophy beyond hardware.
How are the “top AI talents” confirmed in the field being cultivated?
According to a global AI research report published by the China Investment Promotion Office (IPTO) under the United Nations Industrial Development Organization (UNIDO), China had the most global top 100 AI talents (57 people). The US was second with 20 people, while Korea had only 1 person included.
This was a comprehensive evaluation of publication volume and citation counts of approximately 96,000 major AI papers published from 2015 to 2024 and about 200,000 researchers who wrote them. An analysis of data from 44 people who joined Meta’s superintelligence research lab showed that 21 were Chinese nationals, accounting for half (47.7%) of the total. Including Chinese-Americans with Chinese surnames, approximately 60% (26 people) were identified as Chinese-origin talent. US nationals numbered 11, making it the second-largest group.
Meta’s Superintelligence Lab also prominently features graduates from prestigious Chinese universities. Tsinghua University graduates make up the majority, with graduates from Beijing University, University of Science and Technology of China (USTC), and Zhejiang University among China’s top-level science and engineering universities included.
Many of them obtained doctoral degrees from US graduate schools. They received advanced research training at America’s top universities including MIT, Stanford, Carnegie Mellon University, Princeton University, UC Berkeley, and University of Illinois Urbana-Champaign. These are ultra-elite courses with acceptance rates of less than 1%. These universities currently serve as major talent suppliers for global AI research labs.
Meta CEO Mark Zuckerberg actually used Google Scholar to search researchers’ papers when forming the superintelligence team, selecting promising candidates based on citation counts and impact. It essentially started between papers.
Most are proficient in calculus, linear algebra, and probability theory, with deep interest in algorithm design. Simply having computer science undergraduate-level knowledge is insufficient to design “intelligence.”
Interestingly, when these hundred-billion-won salary talents began studying, this field wasn’t popular. When they started their doctoral programs over a decade ago, they weren’t considered mainstream. However, they devoted themselves to “non-mainstream” fields like robotics and generative AI, which have now emerged as core AI technology topics, making them the center of attention.
Despite being super-talents with doctoral degrees from top 1% universities, they concentrated their research on fields they found interesting rather than “mainstream” research that would immediately help with employment, and meeting the right “timing,” they became today’s global 0.1% talents.
Core Competencies for Future AI Talent: What Should We Teach?
This brings us back to the question. Again, it’s about “my child.” If your child is currently in elementary, middle, or high school, or college, how can they become a super-talent?
NVIDIA CEO Jensen Huang provided an answer. When a reporter in Beijing asked him a question, he said, “If I were a 20-something fresh college graduate now, I would focus on ‘Physical Sciences’ rather than software.”
He argues that understanding physics laws is important because AI development will evolve toward understanding and recreating the physical world. According to a report published by MIT, about 41% of global factory automation systems have already adopted robotic forms, and this ratio is expected to increase to 73% by 2035.
This suggests that designing AI that interacts with the physical world requires understanding, knowledge, and application abilities in basic sciences like physics.
Didi Das, an AI investor at Silicon Valley major venture capital firm a16z, listed mathematics, physics, chemistry, biology, computer science, and engineering as subjects students should study in the AI era.
“STEM (Science, Technology, Engineering, Mathematics) subjects teach reasoning, objectivity, and how to think itself. They enable you to learn everything else better,” he emphasized. “There will always be career opportunities in this field for analytically gifted individuals who have these abilities.”
Sam Altman: “My Child Will Probably Not Go to College”
In the AI era, the ability to work with AI is also important. To become an “AI-augmented professional,” one needs training to maximize creativity, strategic thinking, and problem-solving abilities using AI as a tool.
The World Economic Forum (WEF) projected that while 92 million jobs will disappear by 2030, 170 million new roles will be created. This means jobs are not simply disappearing but evolving. One core aspect of this evolution is AI utilization.
In this wave of change, NVIDIA’s Jensen Huang has provided the most realistic educational solution. He asserts that “every child should have an AI personal tutor.” He uses AI to train self-learning, questioning, and reasoning. Not just memorizing, but training to ask AI daily: “Why do you think that?”, “Isn’t there another way?”, “Can you explain with an analogy?” The core of future education he describes is not a “brain familiar with AI” but a “human who can collaborate with AI.”
Sam Altman, OpenAI CEO, a Stanford dropout and current billionaire leading the AI industry, recently revealed in an interview that “my child will probably not go to college.” This isn’t simple elite criticism. He has a more fundamental concern.
“Children to be born will grow up with AI smarter than themselves. They will grow up in a world where products and services smarter than humans are the ‘default.’ Education cannot remain the same as now,” he said.
Altman sees the existing education system as ignoring the “risk of meaninglessly spending the most productive four years.” What he emphasizes is that children need to develop execution ability, collaboration, and the sense to use AI as a tool, more than any knowledge.
He also worries about the university system where the essence of education has disappeared. AI-dependent assignment writing, academic helplessness, and the “networking instrumentalization” of education in US universities show the crisis of the current generation becoming “non-thinking humans,” he warns.
He doesn’t fear technology itself. AI is just another tool like a calculator. However, he worries that if a generation accepts this tool without criticism and increasingly delegates thinking, humanity’s intellectual foundation could collapse.
[Conditions Required for 0.1% Talent to Realize Superintelligent AI]
- Mathematical thinking ability: Linear algebra, calculus, and probability theory are basic.
- Problem definition ability: More important than technology is how to view problems.
- Collaboration sense: Models created alone don’t exist. Collective knowledge comes from teamwork.
- Intellectual curiosity: These are people who choose “interesting problems” over money.
- Experimental environment: Outstanding talent grows only in environments where they can ask the “best questions.”
Jensen Huang, NVIDIA CEO, and a robot created by Disney Research based on NVIDIA’s technology (Source: NVIDIA)
Soft Skills Are Important to Become 0.1% Talent
Successful AI talent combines technical expertise with uniquely human capabilities. It’s important to develop the power to question critically, connect creatively, and empathize with colleagues. Particularly, one must not make the mistake of depending too much on LLMs and weakening thinking muscles.
We’re in an era where Google Gemini can write gold medal-level answers at the International Mathematical Olympiad (IMO) 2025.
An approach that divides into “hard skills” that build AI technology foundations and “soft skills” that create differentiation through uniquely human abilities is also helpful.
First, we can classify as hard skills: deep understanding of mathematics (linear algebra, probability, statistics) that forms AI’s foundation, programming languages for handling data like Python (called the common language of the AI ecosystem) and data processing libraries (Pandas, SQL), and English proficiency for natural language prompting and vibe coding.
Additionally, if one has professional skills to directly create and train AI models – such as ability to use deep learning frameworks like TensorFlow or PyTorch, and understanding of cloud platforms (AWS, Microsoft Azure, Google Cloud) for large-scale model operations as a technology stack – rapid field deployment is possible.
Non-technical capabilities, soft skills, might be even more important competencies. If technical capabilities are the “entrance ticket” to the AI era, uniquely human soft skills become the “differentiation” that makes one an irreplaceable superstar.
Particularly, the power to question and criticize, the power to connect and create, and the power to empathize and lead are predicted to be the most important meta-competencies among numerous soft skills. This aligns with “Moravec’s paradox” – that what’s easy for computers (like calculation) is difficult for humans, while what’s easy for humans (like sensory recognition or complex social interaction) is difficult for computers.
Scott Galloway, professor at NYU Stern School of Business, defined the core of such soft skills as curation, curiosity, and connectivity.
(Source: Gemini 2.5)
Insight Bridge AI’s Perspective: The Ability to Judge Right from Wrong, Critical Thinking as Future Competency
What’s happening in Silicon Valley now isn’t just salary competition. It’s a global competition to find the “brains” needed to design humanity’s next civilization. Like Leonardo da Vinci in the Renaissance, nations are in quiet war over this era’s creative geniuses.
Parents’ questions should now be different. Not “Which school should I send them to?” but “What questions should I help my child ask?”
When machines replace everyone’s average, only human extraordinariness becomes the sole differentiation. It’s self-evident that if we only develop the ability to find correct answers well through Korea’s rote learning education, we cannot compete with AI.
Parents’ roles must also change from “managers” who manage and control children to fit predetermined frameworks, to “coaches” who stimulate curiosity latent in children’s hearts, encourage failure as a learning process, and help them find problems they want to solve themselves.
AI experts also emphasize the importance of critical thinking. Experts who develop logical thinking and thinking abilities through STEM disciplines can ultimately become top AI talents who develop better AI.
Didi Das said, “Today’s most advanced AI models were ultimately reinforcement learned from human experts,” adding, “If analytically gifted individuals continue to work hard, they will be able to achieve results.”
The ability to judge right from wrong is also core to critical thinking. This ability will serve as a safety device to prevent AI from running amok or harming humanity.
See more insightful news!
