Great job 👏

It hinges on how we define violence. For example some people argue being put in jail is violence and there are likely analogous methods you could create for the digital world.

One censorship technique modern regimes have used is to flood the public discourse with so many variants of the truth and lies and everything in between that it becomes near impossible to fact check everything [1]. It makes the average person give up and defer to their favorite pundits.

1. https://en.wikipedia.org/wiki/Firehose_of_falsehood (this page discusses Russian use, but honestly it’s done by all of the usual suspects)

nostr:npub1s05p3ha7en49dv8429tkk07nnfa9pcwczkf5x5qrdraqshxdje9sq6eyhe always hammers the point that prices fall to the marginal cost of production. Over time the vast majority of digital content and services will be provided by bots whose marginal cost is the energy it takes to run them.

And the cost of energy priced in bitcoin has been rapidly falling as more miners compete for fewer sats.

“It might make sense to get some…”

- Satoshi

The idea with all these impersonator accounts is that they can make it near impossible for newcomers to find the real accounts in a sea of fakes.

Imagine coming to Nostr for the first time and seeing your favorite bitcoiner’s name repeated 20 times with the same follower lists (different npubs of course) and replicated content. If you want to zap them, get advice, etc which npub do you choose?

WoT can still work if you can trust a subset of the npubs but it becomes tricky. Simple heuristics like the time that content is published can help but as bots get faster and older content drops off relays it becomes near impossible to tell the difference. The timestamp in Nostr events is truncated to the second (according to NIP-01) which is plenty of margin for a bot to act.

We’ll overcome, but we’ve got work to do.

Regardless if the allegations are true, having so many bitcoin under one roof is a problem.

Indubitably.

“Thanks for the ride, Bob!” yelled Alice while speeding off in her Uber driver’s car before he had a chance to put up the gas pump.

Bob learned a very important lesson that day: not your keys, not your car 🥁

Bitcoin is the best money we’ve got. If you want digital tokens then sats are the best we’ve come up with. If you want a decentralized app for whatever reason (e.g. censorship resistance) then you don’t need a blockchain for that.

P2P is all you need and for that we have Nostr, tor, etc. Make your app open source and if it’s needed then people will run it decentralized.

Decentralized finance is mostly a gimmick. Imagine borrowing money to buy a home and then years later you quit making payments. A DeFi token protocol can’t physically repossess the home and you don’t need it to facilitate communication or market making.

Don’t fall for shitcoins folks!

It’s the Bitcoin-stablecoin duality

It’s no coincidence the big surveillance ad network companies with large government contracts and political ties are blessed by the money printer. Governments crave control over the social sphere and don’t want to compete with other nations. This might be an arms race between nations over influence online regardless of productivity.

The internet has become a new frontline for war. Motivated actors will use bots to both promote their culture and ideologies as well as drown out or demoralize others.

That was always the case with propaganda but the tools and infrastructure we have today enables it at a scale none of us are prepared for. Webs of trust communicating over neutral networks like Nostr are part of the solution but we’ll need more.

Since AI models and agents scale better (in theory) than human teams the most motivated and productive among us will increasingly lean on them as part of their work. If in the future governments play their same traditional role then they'll necessarily rely on AI surrogates and proxies: digital diplomats.

Then it becomes a race to see who can digitize their culture and ideologies faster and deploy globally. We're seeing the early stages of this play out in chip manufacturing as well as data center deployment and open-weight LLMs. Expect this to accelerate.

I think this is why we're seeing the big tech companies in the largest economies who have large network effects and are known to have extensive ties to their governments aggressively pursuing an "AI at all costs" strategy.

So what comes next? We'll continue to see dominance by the countries with the best telecommunications/electrical/chip manufacturing infrastructure of course, and we can probably expect further erosion of social media networks since they're the "digital frontline" so-to-speak. But what else?

#ai #diplomacy #geopolitics #asknostr

Function signatures are more of a suggestion. You can call a function with any number of arguments and with arbitrary types. And you can add strings to ints but multiplication gives NaN. Lots of arbitrary design choices.

Part Cyborg, part honey badger. Good luck in Saudi, Cris!

nostr:note1dq8grhc3rgtwag4h94hqpx78ktew772afm6xqgw9g78fvt9eamkss8lhtv

The truly dedicated ones decorate their yard with zombie snowmen and reindeer to get a head start on Christmas while they’re at it.

Time to make it dead simple to make a LoRa repeater and drop them all around. Make it near impossible to enforce a ban or license restriction. Communication is a right.

nostr:note17vv9h0rxxaxs4z7vjfk4nel8678cr8syqvd4kwt4x0jmrtn36ztqvrsplx

You’re right; it will take more than just data prep and likely more than 4 years to mature.

For example we need an architecture that allows for versioning of components, auditability, benchmarking for regression testing, etc. It also needs verifiable outputs so we can prove the output was generated without tampering.

And there are supplementary components like vector data stores so that users can store context for long running tasks.

We also need a stable, scalable architecture for such a system. Sparse mixture of experts could be an option. Over time the network could add new “experts” and prune unused ones.

nostr:note12zcjle0vnv7tk4s8kuqjxvp6z3xslhrtjpcm07zja3cehxletv8q7n2v6m

There’s a new alternative to DiLoCo [1] for training large scale AI models over the internet called DisTro [2]. It enables low latency training on low bandwidth communication channels (ie. slow internet).

Methods like these are a crucial component for enabling a decentralized AI system that rivals the big tech companies and nation state actors.

The next step is to figure out monetary rewards for contributing to training and inference. The tricky part is to weed out bad training data in a decentralized way. Perhaps we could use something like a “mempool” for training data batches?

1. https://arxiv.org/abs/2311.08105

2. (PDF) https://github.com/NousResearch/DisTrO/blob/main/A_Preliminary_Report_on_DisTrO.pdf

#ai #llm