The risk of AI is in centralization of AI. In the same way the risk of the web was in centralization of the web.
Those very real risks manifested in very real ways on the web.
I have young kids. I'm not exaggerating when I say I wake up every day and diligently build the future I was promised for them.
I've got somewhere around 10 years, conservatively, until Instagram/TikTok/YouTube/Facebook/etc. try to get my daughter to kill herself. I've got around 10 years until she is a statistic, hopefully the statistic that survives.
Social Media wasn't always Social Media. We weren't promised Social Media. We were promised Social Networking.
Social Networks were supposed to connect us with our communities. Bring the world together in a way previously impossible. Allow the free exchange of information and ideas. That future is still possible.
But it turns out expecting some dude in Palo Alto to pay to be the free Proxy and Archivist of the entire world's social communication is a hard problem to solve. It's a hard engineering problem. It's a hard financial problem. It's a hard legal problem. It's a hard ethics problem.
Social Networking companies didn't solve that problem. They failed and pivoted into Social Media.
Social Media divides us. Social Media isolates us. Social Media believes my friends, my family, and my community "aren't engaging enough" and litters my timeline with randos it's trying to hype to keep me engaged so they can monetize the experience.
I go to my PTO meetings and parents talk about their kids being isolated and depressed. They talk about body image issues and lack of purpose.
I go to my community gatherings, business networking groups, and gym. People talk about how lonely they feel. How angry they feel. How depressed and anxious they feel.
Instead of the Social Networks we were promised, we have these Social Media monstrosities that take randos on the internet and try to make them famous in front of other randos on the internet in a variable rate reward slot machine.
Social Networks on P2P technology have been viable for at least 5 years now, but the industry hasn't caught up. It's possible to join a p2p social network from your browser tab, pair it with your phone, have them both simultaneously be able to generate posts and sync over a gossip protocol without conflict. And it's possible for your social group to backup your content. All of this is still possible with the Application Service Provider architecture of having a "backup pinning server" that will make your data highly available.
Identity is solvable in a way that allows for full control of your identity without relying exclusively on public/private key cryptography - and permits social recovery of your identity when your keys are lost or compromised [1]. End users don't need to understand private keys or public keys. They just know they have an account, and that their friends can send them a friend request.
I'm going long on p2p applications. I think they aren't just viable, but that they can out compete the user experience that current generation Application Service Providers offer while simultaneously breaking the incentives that are destroying the societies my children will inherit.
AI is no different. The call to arms about AI regulation and alignment often degrades to advocating for a world where my children do not get access to AI because it's "too dangerous" for them - while someone else's kids belong in the privileged AI class.
The trajectory forming for AI strongly signals to me that it will be the single most profound means of production our species has produced. Any philosophy that advocates taking that away from my kids can go play in its own corner. I'm not taking back the web for my kids just to let them take away AI in return.
If these things interest you, and you're interested in doing hardcore research on productionizing real, no bullshit, peer-to-peer networks both for end users and enterprises, my email is in my bio. I want to talk to you. We need more good people in this space, and we have budget for it.
My team is building peer-to-peer networks for AI, but the solutions that break AI out of it's ivory walls aren't AI specific solutions. These peer-to-peer building blocks are general purpose and the next generation of the web is going to run on them.
This is the other other road ahead [2].
We will succeed or I will die trying.