Google Co-Founder Sergey Brin draws parallels between the Internet and the AI
Stage of Development
AI in 2025 is compared to the early web (around the early 1990s, especially post-Mosaic browser in 1993).
AI today is like the web just as it was beginning to scale — visible potential, rapid momentum, but still in its early phase.
Invention vs. Discovery
“it’s a discovery in the sense like we simply do not know what is the limit to intelligence …
the internet was brilliant … but it wasn’t like technically revolutionary … nobody would have questioned whether that was physically possible five years before.”
The Internet was largely an invention — an engineered system based on agreed-upon protocols (e.g., TCP/IP, HTML). The internet was like inventing money or communication channels.
AI is more of a discovery — exploring the boundaries of intelligence, an emergent phenomenon we don’t fully understand. AI could reveal new “laws” of intelligence we didn’t know existed.
Similar to quantum computing — high theoretical potential, unknown practical limits.
Predictability of Impact
“With the internet you could imagine everybody could communicate at high speed with everybody else … but with AI you don’t know where the peak is — or if there’s a peak at all.”
Internet: The impact was imaginable (e.g., websites for every business, high-speed communication).
AI: The impact is unimaginable — we don’t know if there’s a ceiling or what superintelligence might look like.
Technical Foundation
“We don’t even know what intelligence is … we don’t know how far we can take it … With the web … they were just organizing the scientists’ data and stuff and sharing it … but it wasn’t like … testing the limits of the universe.”
Internet: Not technically revolutionary — built on existing computing and communication systems.
AI: Technically profound — challenges fundamental understanding of intelligence and computation.
Philosophy
“The internet didn’t raise questions of consciousness, for example. But right, you know, if this AI is smart enough and self-aware enough — does that matter? What does that mean?”
Internet: Raised few philosophical questions; mainly technological and social.
Scale and Resources
Early Internet: Built by startups with modest funding (e.g., Google started with <$1M in seed funding). The web started small—startups and protocols spreading through viral utility.
AI Today: AI has massive momentum: billions of dollars, global talent, and exponential hardware demand, making it fundamentally different in scale, speed, and stakes.
Global Attention
“This has now gained … profound international attention. The amount of resources and … energy that are flowing towards AI is extraordinary.”
Internet (early days): Niche, tech-focused interest.
AI (now): Broad international attention, involving governments, corporations, academia — a geopolitical priority.
Momentum and Acceleration
“A lot of people, myself included, are just surprised how quickly and how far it has ramped.”
Internet: Spread quickly, but through social and business adoption.
AI: Rapid, exponential growth in capability — surprises even experts.
Human Experience
“Language models 2 years ago made so many very embarrassing errors… wow this thing actually did this correctly and that’s super cool
…
when you kind of look at the trend, you probably will be able to use it
dayto-day with reasonable reliability”
Internet: Changed how we communicate and access information.
AI: Potential to change how we think, learn, create, and interact with reality itself.
Where to focus?
“You’ll see um some kinds of reinforcement learning APIs from the top models so that you know you can send us your problems um maybe with correct solutions or guardrails and we can contribute your problems to the mix of like if you want the model to be good at that
…
it’s pretty good time um to be able to make an impact uh without having to train up a foundation model”
- You don’t need to be in a lab or have an elite pedigree to contribute.
- Tools are becoming more accessible—even if you’re not on the frontier, you can build with frontier tools.
- For the foreseeable future, human direction still needed for best results.
- Find real-world problems where AI could help, then learn how to leverage existing models (text, code, video).
- Stay future-facing. What these tools will do in 1–2 years is far more important than what they can do today.
Philosophical questions emerge from technical ones:
- How to evaluate new models?
- How to compare diffusion vs. autoregressive architectures?
- What is a good model?
References
https://www.youtube.com/watch?v=4N9MCa4hCsA