![[lego.png|478]] *a lego orrery i built a while back* There seems to be a broad consensus around AI eating the world but so little about how it will impact human cognition. Kids growing up today will never know a world without AI. It is just like how we didn't know a world without the Internet. In the mid-2000s, our parents had to learn how to use Google but we just knew it. Now, today's 5 year olds will just know claude/ chatgpt and whatever comes next. ### the parallel Here's what being an "Internet Native" actually meant for our generation - Access to near infinite information: we leapfrogged books and ramped faster on knowledge than our peers who did not. - Constantly learning things outside of school: we constantly searched for tutorials on Youtube, accessed niche internet builder forums, went on deep wikipedia rabbit holes. - Lower friction to curiosity: most friends I made throughout elementary to high school were internet enthusiasts and constantly shared cultural moments and information that otherwise would not have been accessible to us. Ultimately, we were one of the fastest learning generations relative to any other generation in human history. ### why ai is a bigger jump than the internet was The biggest blocker for us was the prerequisite to understanding the information itself. A 12 year old trying to understand how to build RC planes also needs to understand various foundational knowledge around the best practices. This was something that was largely inefficient, reliant on outdated wikis and scouring niche forums to build the muscle for it. AI gives access to understanding at your own level of intellect and accrued knowledge. The difference between reading 10 wikipedia pages vs having a tutor explaining it to you at your own level is a game changer. The closest example is having your personal tutor at adaptive difficulty that constantly revises your baseline upwards. For that, the implicit assumption that LLMs are better than the 90th percentile human in that field is closer to the truth now. It is in fact infinitely patient and personalized to the temperament and thinking pattern of today's generation where divergent starting points can converge to expert level understanding. As opposed to our generation, where a certain style of thinking and behavior was rewarded and whose impact compounded over time. ### what it means for learning Broadly, the floor on human cognition will rise yet again but by orders of magnitude. The average kid will know more, reason better and learn faster than the average kid from our generation. This is not because they are smarter but the tools are so much better. The tools when used correctly and in the right form factor will eventually deliver the most impact in the early stages of peak neuroplasticity in the current generation. ![[chess.png]] *via [Raskerino chess](https://raskerino.wordpress.com/2022/03/09/comparing-fide-ratings-over-time/)* Think of it this way, the average chess player today is better than the average chess player from the 1950s. It's meaningfully different when you lead from raw intellectual horsepower versus when you augment it with the right training tools and databases. ### what changes and what doesn't We are moving closer to a reality where learning speed, breadth of knowledge and the ability to go deep on niche interests will continue to compound at unreal pace. The baseline reasoning ought to improve when you can constantly reach first principles and later reason upwards and cross functionally into other fields. That also means that taste, judgment, agency and knowing what questions to ask become the most impactful things to guide the current generation on. This is largely due to the fact that these things will not undergo any meaningful change with the advent of AI. This also makes a case for optimizing for varied and unique experiences early in life. Think: a trip to Europe and exploring Art Museums will do more good than constantly scouring Behance and overanalyzing UI/UX approaches. Access to unique and differentiated experiences will yield diverse starting prompts that can be accelerated with LLMs broadly. ### the core counter argument > But won't kids get lazy/ dumber/ dependent? This seems to be the biggest worry I have seen parents mention. This is also something I have heard my own parents and teachers say to me throughout schooling. Over reliance on tools are often seen as crutches that delay inevitable reckoning with reality around true skill. Historically, similar reservations have been made about Calculators and mathematical aptitude, around using Google and research potential, around GPS and spatial navigation skills. These are good arguments indeed and should not be trivialized as being a luddite. Every generation defining technology raises the bar on human cognition, the floor always goes up and not down. But it also indicates that the fabric of reality itself is in flux. So the worry itself isn't new but has been historically wrong. The job I am working on right now did not exist when I was born. The same appears more likely in the future and being infinitely curious is the only hedge to it. ### what next? The gap between an AI-native kid and us will feel very similar to the gap between us and someone who grew up without the internet. It is about skills that atrophy and the skills we nurture that defines this divergence. One thing I've been mulling over: there's an obvious shape for AI-assisted learning startups here: an LLM layer for the kid, an observability and curation layer for the parent. The pitch writes itself. But the execution risk is real and I haven't fully thought it through yet. Curating _how_ the model teaches (when to give answers vs make them work for it) is harder than it looks, distribution is brutal when the buyer and the user are different people, and there's nothing stopping the foundation labs from shipping a "kids mode" natively. The version that probably works is narrower than "AI tutor for kids", but I don't have a strong take on what that narrower version is yet. We are also the last generation that had to learn this as adults which itself is a weird thing to sit with.