It’s kind of an artificial state from what I understand. They have a one dimensional economy and they’re just propped up by US and Israel as a staging area against Iran and an oil extraction facility. UAE, Bahrain, and Qatar seem like most likely to disintegrate as a result of the war.

Exactly, that would be the prerequisite. That said, given that these bases have already been demolished, it might not be that big of an ask.

I mean now that they can see the US protection isn’t worth the paper it’s written on, there’s not much other choice.

I mean the costs are literally an order of magnitude lower for either using Chinese services or even just running your own on prem models. And, while top tier models like Claude are better, they’re just not so much better to justify the astronomical cost. I really don’t see how the whole business model US companies developing proprietary models have can survive even in the near term, let alone the long run.

I think at some point you can be fairly sure that the model performs well enough. And the simplest thing it can do is literally just act as a translator layer on top of the model. So, if you give a query, it’ll reformulate it in a way the model is known to respond well to. You can do a random sample test to see that you’re generally getting the results you expect too.

At the end of the day, models shouldn’t be treated like oracles in the first place, it’s a useful tool for helping point you in the right direction, or work through a problem. But it should always be the human making a decision in the end, and doing their own due diligence to verify the information.

I find if there’s one critique that I have of Chinese news sites is that they almost never bother with linking to the primary sources. Like I do more diligence in my substack posts. 😄

These AIs have American values firmly embedded in them it turns out.

Yes, you absolutely can. That’s precisely what LoRAs are for. You can completely change the way the model responds by adding a layer on top. All the core knowledge stays the same. I’ve actually done this myself. I rented some time on runpod to train a LoRA on Lovecraft that I applied to a base Qwen model.

It’s a lot worse than that, it’s not that the model doesn’t understand the question. It chooses the answer based on the persona it interacts with. It’s not a capability limitation.

The difference is that DeepSeek has open weights and anybody can download and run the model themselves. And you can tune the model any way you like, so even if DeepSeek had some baked in biases, anybody can publish a new version without them. That’s why developing this stuff in the open is so important.

I expect that we’re going to start seeing more aggressive push to ban Chinese tech in G7 countries in the coming years. And as a result, there’s going to be a bifurcated market for technology globally. The global south will be using advanced Chinese tech, while the west will be stuck with legacy technology increasingly falling behind the rest of the world.

but where’s the profit in that?

it just fundamentally doesn’t work, Ted Postol gave a few interviews explaining how it’s completely unworkable, this one is a good run down

https://www.youtube.com/watch?v=D9Wzlqh7bIs

pretty much :)

nah, everybody knows the golden dome is unworkable, it’s just another scheme to line the pockets of the kleptocrats

exactly

Jokes aside, this was a great interview explaining how Iran absolutely could build an atomic weapon if they actually wanted to. https://www.youtube.com/watch?v=JtUobr7xGz4

The whole aspect of the imperial decline where they continue making wrong decisions based on the assumptions of their own superiority has to be the most hilarious aspect of the whole thing.

yeah basically, I bet it’ll be pretty handy for construction work

The war in Ukraine was one of the most consequential blunders for the empire because it led to a whole alternative economic system being established. And not only can’t the west weaponize it, it’s also completely opaque to the west removing the advantage western companies had being privy to trade flow data. I recall how for a while there were a whole bunch of articles talking about China’s trade slowing down, and then it turned out that they were based on SWIFT numbers. What was actually happening that China’s trade through the western financial system was being reduced.