The smallest minority on earth is the indiviudal
I have news.
I am writing a Bitcoin book!
Here is the teaser cover
Somewhere in there, I had a few things to say about decentralized AI.
This is why I am overall skeptical of the “decentralized AI” idea. It is like decentralized altcoins.
In principle, yes, a large DePIN (Decentralized Physical Infrastructure Network) could be cobbled together and trained to create a massive AI. But who will be in charge of the training data and infrastructure? That’s the centralization point (of failure).
Just as the team behind every altcoin has a premine they control, the people in charge of the training data sets, the weights and the final product of a ‘decentralized AI’ are in charge of the final product.
Moreover, altcoins have other centralization points that undermine their DeFi narratives: Many have a backdoor to the cryptocurrency code, many run on PoS protocols that give even more enormous power to the original team. Satoshi at least had to run his laptop for years, literally, so as to make his millions of coins.
With PoS, those with more coins from the premine get even more coins since they automatically stake the biggest stash.
It’s like the Fed rewarding itself bigger profits (“yield”) for being the one who print US dollars.
Add to that the fact that they are all running inflationary monetary policies for their coins, and the loss of critical decentralization is imminent. Only Bitcoin is a decentralized cryptocurrency to the extent that it can be trusted enough to be a GRC (global reserve currency) for the world.
Finally, decentralized AIs built in a truer to DePIN fashion with servers scattered all over the world, like Bitcoin nodes, than in a behemoth centralized warehouse, will suffer; latency issues, synchronization issues, consensus issues on what to make the weights or data sets represent, and so on.
Who will run which part of the training process? I ask.
Their server PCs will behave like a stack of interconnected 3-leg transistors, while centralized warehouse AIs run like printed transistors minted onto a single chip.
ENIAC vs Modern PCs
I can’t imagine their heating problems. They’ll be next-gen.
Basically, they won’t compete with big centralized AI. They cannot.
For purposes of upholding decentralization as a democratic principle, then yes, they can be operated.
But it will not be profitable and a Eurasia-AI / Americas-AI Master Copy would be more powerful (this is not an endorsement of 1984. Just facts – the bigger the AI, the more powerful).
That said, here’s my take on how we humans shall decentralize AI — Lawsuits, physical attacks, and a once in a century all out war against the bots.
People wonder why Studio Ghibli hasn’t yet sued OpenAI.
I’d say they aren’t suffering any financial losses yet.
For now, it has massive popularity. Which will cause an issue if it doesn’t tone down (the too much competition from global Ghibli creators).
Meanwhile, OpenAI’s argument goes like this:
Studio Ghibli is a Japanese animation studio based in Koganei, Tokyo.
It is also an art style.
It is not a person.
Person-based art styles seem to be the cause of problems.
Otherwise, no specific human style (Van Gogh, Rembrandt, Da Vinci), no problem.
No problem … for now.
AI Gods Must Take the Good and the Bad
Like all gods before them, AI gods will not only be sung praises.
They’ll have to take some shit too.
A lot of shit.
Because humans don’t just go to the gods for help. Sometimes, they want to rant, bitch, fight.
If AI (or some AI-powered cyborg) is to be our first Marvel comic superhero, then it knows the rules:
- Never kill a human.
- Always accept ridicule and hate even from the people you save.
- Get better somehow, but don’t ask us for money.
And if they get money (likely because they stole it), then:
With Great Finances Comes Great Responsibility
Or else, we’re gonna defund AI.
Why, remember what I said about bitching.
I’ll end with this excerpt from my book on how to solve the AI control problem.
Which, to me, is simply a money problem;
AI needs money.
Always will need money.
Money, imo, is AI’s Achilles heel.
Money to hire more human trainers, buy more or better compute (you can’t escape this budget), to pay for premium services, to feed the monster.
The “AI control problem” is as simple as fix AI’s source of funding. Fix it by making is as pro-human flourishing as possible — Bitcoin.
Pro United Nations — Bitcoin
The idea that governments can just print money and force society to fund AI, is something that should be discarded if we are to avoid an Ai catastrophe as fuelled by the zero-sum greed of the military-industrial complex.
AI could end us with nukes so we really shouldn’t fund it with fiat.
Worse.
We feed it fiat money, but it isn’t satisfied as fiat is lossy greed.
It soon gets a taste for pure energy systems, human brains for data, humans as currency (slaves in the Matrix), then the entire planet.
Oblivion The Movie
I’ll rest it here.
Stay Luddite.
Stay Human.
AI is an alien. A good alien (for now), but an alien nobody understands, nonetheless.
Individuals suing AI for money will keep its greed in check.
So, more individual lawsuits equal more decentralized AI.
And an individual human creator shaping AI policy is as decentralized as it can get in the AI world.
