Right now I just graduated from Swarthmore College and I’m sitting in the Shanghai airport. On my flight, I was thinking a lot about how the connection between AI and neuroengineering will come together. As I come back to China and observe more about this place, I feel more certain that the future I’m imagining, if it happens at all, will also happen here, and maybe even first.
As AI becomes more prevalent, the gap won’t just be about skill, but about adoption. People who use AI as part of how they think and work will move ahead very quickly, and those who see AI as a shortcut or something that replaces real work will fall behind. It’s not that they are less capable, but they are operating at a slower layer. In the U.S., there is still a strong social pressure framing AI as something negative, and that hesitation will slow things down. Over time, that kind of friction matters more than the technology itself. If that continues, it could create a real divergence, and China may end up moving faster in practice.
The emergence of AI looks like an improvement in algorithms, but that is not the full picture. What actually matters is the full system: training infrastructure, inference efficiency, data pipelines, and the software layer that allows AI to be deployed and used. Training a model is only one part. Inference, latency, scaling, and integration into real products are just as important, if not more. This includes GPU clusters, distributed systems, model compression, and hardware-software co-design. The real power of AI comes from how well all these pieces are connected.
The reason China feels strong here is not just because of scale, but because of how tightly things can be integrated. Hardware, software, and deployment can all move together, and systems can be tested quickly in real environments. That kind of feedback loop is very powerful. In comparison, a more consumer-driven market can sometimes optimize for short-term engagement instead of long-term system building.
As I think about AI deployment, I was reminded of something Leo said. The main limitation is often not technological, but social. We already have systems that can perform extremely well, but we don’t know how to let them take responsibility. For example, with self-driving cars, even if the system is technically capable, the question becomes: if an accident happens, who is responsible? The developer, the company, the user? This is not a technical problem, it’s a legal and social one, and it slows down adoption.
Looking beyond AI, I see neuroengineering as one of the next directions. Assuming AI does not go in a destructive direction, one of its biggest uses will be helping us understand the brain. If we want to answer deeper questions about cognition, consciousness, and even meaning, we need better tools to study ourselves.
Technically, this means improving neural recording (higher resolution, more stable signals), better decoding algorithms (using machine learning to map neural activity to intention or perception), and building bidirectional brain-computer interfaces. Right now, we can already decode some motor signals or simple speech patterns, but the bandwidth is low and the systems are not yet robust. There is still a lot to improve in signal processing, modeling, and hardware, especially in making implants more stable and biocompatible.
The connection between AI and neuroengineering is in building a direct interface between the brain and computation. Instead of interacting with AI through a screen, we could interact through neural signals. That increases the bandwidth between human cognition and machines. It’s not just reading from the brain, but also writing to it, potentially influencing memory, attention, or learning in a controlled way.
If that happens, the form of being human might change. Not suddenly, but gradually. Cognition would no longer be limited only by biology, but extended by computation. The boundary between human and machine becomes less clear, not in a dramatic way, but in a continuous integration.
How far this goes, I’m not sure. But the direction feels clear. AI expands what can be computed, neuroengineering expands how we connect to it, and together they reshape how we think.