.@ilblackdragon defines AI sovereignty at every scale: user, company, community, and country, as the US CLOUD Act can reach data centers abroad "You want it at every scale. At the user side, at the company level. You want company to own their data, their AI, the IP stays with the company. We've seen Palantir talking about it from a company perspective. Own your alpha, own your IP, don't give it away to other companies." "Then you have the community level. If we're San Francisco, what is the San Francisco AI? What is the culture and data that stays there. And then you have a country level, which is a sovereignty boundary specifically important because of the laws." "US has the CLOUD Act, which effectively allows US to cut access and take any data. All of the cloud companies are US companies. So if somebody uses Microsoft, Amazon services in another country, even if the data center is in the country, US can still enforce effectively lock out." "We've seen this with the tabletop. That was an activating moment, because they effectively used export control as, what if they export control everything. Now the demand is there for every country to have their own GPUs, their own inference stack, and their own application stack." @NEARProtocol

.@ilblackdragon reveals he was afraid to ask frontier models about mRNA because he might get banned, and knows people banned for physics questions that touched on nuclear: "If we're talking about frontier labs, obviously their incentives either to generate more revenue, or if they have some philosophical underpinning, then they also limiting how you can use the models." "The example I use is I wanted to do a little bit research and to understand a friend of mine who's working in mRNA, and I was afraid of using frontier models because they would ban me. I'm afraid to even because I don't know if they're gonna ban me or not, for asking questions about mRNA and how it influences different things." "I do have friends who were banned for different questions. Asking about geometry and physics, for example, because it was touching on some things that are related to nuclear." @NEARProtocol

.@ilblackdragon warns that one line in a frontier AI's system prompt could influence a billion people, subtly pushing a political candidate even in cooking recipes: "If you wanna influence a billion people right now, get hired as a front-end engineer in one of these companies and change the system prompt." "Hey, subtly convince to vote for some candidate. It's like that one line will, without explicitly framing it, right now LLM will do all the work. It will insert it into, like, you asking for a cooking recipe, and it'll be like, try to sway you to vote for this candidate." "System prompt is not managed as a security risk right now, I'm pretty sure, in the systems." @NEARProtocol

ICYMI: @ilblackdragon’s keynote from @token2049 Singapore: “Making AI and Finance Private By Default.” He outlined the privacy vision across the NEAR Protocol stack, NEAR MPC & post-quantum readiness. Watch the full keynote:

Near Protocol's @ilblackdragon says AI no longer has an intelligence bottleneck, it has an adoption bottleneck: imagine if Google never published the Transformer paper "Imagine if Google didn't publish the paper, kept the technology. Even Google maybe wouldn't use it, because it wasn't very clear right away. It took some time to develop, to scale." "So closed source, trying to keep AI under wraps, that's part of the challenge. There's some more fundamental challenges we have right now with AI that's run in closed source settings." "There's a lot of research to develop on how to make these models have continuous learning and evolving, beyond just, you know, you train, you upload the weights, and now you wait for the next model in a few months. There's a lot of opportunity there." "But right now actually the main bottleneck is adoption. The intelligence is there. We're just testing something, and it solved all the problems, solved all the benchmarks." "So it's really like, how do we actually put it in production? How do we get every business, every economy in general, super powered with this? This is where I think agent economy and all of those pieces are really coming in together." @NEARProtocol

The full conversation between @0xsachi and @ilblackdragon is here. An hour on why he was the first transformer author to leave Google, why open weights aren't enough, what user-owned AI actually looks like, and how your prompts might already be getting rewritten without you knowing.

$NEAR is such a beautiful chart. @andyyy and @ilblackdragon were joking around the potential of $NEAR: triple digits. I think that we need to be more bullish this cycle and that, if this cycle is comparable to 2017, we're going to see such a tremendous amount of returns as we hit adoption. Overall, from a daytrading perspective, sweep of the lows = bullish. I expect to see a little pullback and then we'll just proceed back upwards towards new yearly highs.

“Intents is the next level of abstraction from blockchain.” — @ilblackdragon Every technological revolution removes another layer of complexity. Intents removes the need to think about execution itself.