Source: Xiaohongshu Author: Mindset Is What Matters Most. This is the September 23 episode of the Ezra Klein Show, with the full version running nearly two hours. Huang explained his multi-layered theory and his thinking on AI threats. It is extremely meaningful and inspiring for ordinary people and worth spending time to listen to in full.

A study in China covering 26,811 middle and high school students recorded frightening results: after using AI, students' homework scores increased by 18% and completion time was shortened by 30%, but their monthly exam scores dropped by 20%. Students completed homework more efficiently through AI, but they did not truly learn the knowledge. The host placed this set of data in front of Jensen Huang.

Jensen Huang's answer was unexpected. He admitted that long division, multiplication tables, and manually calculating square roots are being forgotten, but then asked in return: does that really matter? He even joked about himself: he cannot remember his own postal code, nor his phone number, but life still goes on.

This is not to say that knowledge is useless, but that every time technology moves one step forward, people's attention moves up one level. After calculators became widespread, people no longer obsessed over hand calculation; as software and chips became increasingly complex, engineers could not possibly remember the underlying details of every transistor. When Jensen Huang had just graduated, the first chip he participated in designing had only about 200 transistors, and he was almost familiar with every one of them. Today's computers contain millions of trillions of transistors, and engineers long ago stopped studying them one by one, instead standing at a higher system-level position and thinking about how modules are combined and how they work together. Only by remembering fewer underlying steps can people devote their energy to bigger problems.

Jensen Huang calls this change moving from a "transistor thinker" to a "systems thinker." This is also how he understands AI. AI automates tasks, not purposes. CT scan reading can be handed over to machines, but doctors' responsibility to diagnose diseases and help patients has not disappeared; code can be generated by agents, but engineers still have to define problems, organize systems, and verify results. If you merely let it write answers for you, what you save is time, and what you may lose is ability; if you let it take over repetitive calculation and execution while people remain responsible for understanding, judgment, and integration, then it pushes us toward higher efficiency and a higher level.

Jensen Huang believes that in the future there will be fewer and fewer graduates who do not know how to use AI. But AI also has no mysterious will. In his words: there is no willpower there, only electricity. Tools will become stronger and stronger. In the end, what creates the gap is not who remembers more calculation steps, but who understands the entire system and can use tools better and more efficiently.

Personal reflection: the biggest difference between humans and animals is that humans can use tools, and social progress is also based on not spending energy reinventing the wheel, but building higher-level applications on top of existing technology. For example, thirty years ago students still had to learn to use an abacus, but now the abacus has been eliminated, and even calculators are no longer popular. In a few more years, students who cannot use AI may find it hard to graduate. I highly recommend everyone watch this interview, and do not watch other people's summaries or short video overviews. Watch the full version, and you will have a different experience.