A court in Chile has allowed Amazon to build a new data center in Huechuraba near Santiago after local residents and environmental groups failed to stop the project. This forms part of Amazon’s plan to invest more than $4 billion in data centers across Chile. People were worried about high electricity and water use, plus new power lines that could affect the hills, but the project passed all environmental checks and any new power lines will be reviewed later. Amazon says the data center will use special designs to protect the environment. Chile is becoming a popular spot for big data centers because of its clean energy but buiding data centers are destroying the nature

@Pirat_Nation I’m from Chile, and we need at least 100 times more datacenters like this one They should pay 0 taxes, in fact, we should pay them to build them.

@Metaculususer @Pirat_Nation I'm from Chile and also extremely pro-datacenters but we should obviously tax them like any building, that's one of the main ways Chile benefits from foreign investment lol

@austinc3301 @Pirat_Nation Fair, But compute is an extremely valuable resource, It might just run out, CPU's are running out right now In my opinion we need to push a bit and agresivly build for the future, and put big or near total temporary tax incentives But fair point of view.

@Metaculususer @Pirat_Nation I do agree we should incentivize it in the short-term, but in the long-term the whole point is for Chile to benefit; the problem isn't whether demand for compute will be met (it will be), but whether we get to take our share of the surplus

@austinc3301 @Pirat_Nation Just hosting the datacenters is good enough. Even if they generate 0 tax revenue. (I highly doubt that.) In a case of war, tariffs or uncertainty, having datacenters already build is key. Local compute. You can see it as Defense spending if you like so.

@Metaculususer @Pirat_Nation if you host them without any taxes, then all you get is the jobs benefits and that's probably negative on society on net

@Metaculususer @Pirat_Nation 'defense spending' makes no sense, insofar as compute is valuable, we don't get access to it, amazon sells it to the highest bidder, and so we gain no advantage from having it happen here than it is if they had built in peru

@austinc3301 @Pirat_Nation In case of war, ask for Amazon to use it, and in the worst case we can just expropriate it 'National interest' At simple as that. is in OUR territory and all the building is ready, all it's set up.

@Metaculususer @Pirat_Nation How exactly would we use it for war? we don't have access to frontier models, so AI is not really a thing we could do ourselves, and it's not like war will be determined by how many vCPUs we have available

@austinc3301 @Pirat_Nation We don't have frontier models but we have the compute, compute is absolutely necessary. There are Chinese Open Source models (V4), that aren't as far from the frontier, we can finetune them with compute. Palantir uses tons of datacenters, Military operations need compute and AI

>There are Chinese Open Source models (V4), that aren't as far from the frontier, we can finetune them with compute. This is simply not true; they're extremely far from the frontier, at least 6 months behind, and China does not have enough compute to actually keep up scaling pre-training indefinitely, so they're actually falling more and more behind each month. See e.g. Epoch's analysis via ECI epoch.ai/data-insights/…

@austinc3301 @Pirat_Nation That chart is outdated as fuck. Kimi-2.6 is like 6 months behind from frontier models. Even cursor finetuned it and uses its as it's flaship model. Don't look at the chinese models as they are trash. They aren't

@Metaculususer @Pirat_Nation I went and re-generated the chart with kimi 2.6, and voilá, the gap actually widened since!

@austinc3301 @Pirat_Nation Well, that's right, I know they are getting behind on some benchs But that doesn't invalidate my point. The Chinese models are good as fuck. They are free. And having a national reserve of compute is good. Your own chart shows it at GPT-5 level. Academia runs on Chinese AI

@Metaculususer @Pirat_Nation These rankings simply don't correlate with AI talent at all, because almost all top AI talent in the US is in industry, not in academia. The Chinese models are good, sure, but they simply are well behind US models, and the gap is widening, not closing.

@austinc3301 @Pirat_Nation Don't underestimate them. The US has a big wallet for AI talent, But many of them stay in china, for example in efficiency in inference? Bruh the chinese dominate the US. Have you seen Bytedance Seedance 2.0? That Txt-2-vid model crushes everything from the us.

@Metaculususer @Pirat_Nation the only reason you're impressed by Chinese papers on efficiency of inference is because all the efficiency of inference improvements the US makes are mostly kept as trade secrets; Anthropic and OpenAI do inference very efficiently, we just don't know how they do it

@austinc3301 @Pirat_Nation Even dario admired V3 efficiency when it launched. He had to make a public statement. (I don't really trust him) And just to add, there is also the chance DeepSeek doesn't make everything they make public. Deepseek offers V4-pro for free in their web reddit.com/r/LocalLLaMA/c

@Metaculususer @Pirat_Nation Not sure why you removed the TechCrunch link, but I remember that Dario also said in that statement that the improvements had been exaggerated and that:

@austinc3301 @Pirat_Nation It cutted off, idk. Yeah, but the point is that he had to make a public statement. Dario is a CEO, and he would never make a comment that would harm the Venture capital flow. reddit.com/r/LocalLLaMA/c…

@Metaculususer @Pirat_Nation well, if that's true, then I don't see how you hoped to justify your point by citing him it has to work both ways: you can't have it support your statement when the quote agrees with you, and not have it also work against you when the quote disagrees

@austinc3301 @Pirat_Nation Well, I agree with the quote, I think that maybe the issue is the difference in the amount of trust we put on Dario. If someone trust him blindly, then yeah, Chinese models aren't a big deal. I do not, I see him as a bubble enabler. And this statement is a 'This model is great'

@Metaculususer @Pirat_Nation again, you're the one who cited dario, not me!

@austinc3301 @Pirat_Nation That doesn't change that the Chinese have a great hold on The AI academic space. If you don't like that subjective analysis of mine. You can always look at csranking