whereismyacc

That I can absolutely believe but the big competition is in enterprise gpt-5-size models.

best inference silicon in the world generally or specialized to smaller models/edge?

What would be the evolutionary advantage of this? I know nothing about biology so I'm struggling to see the use of such a network. Some kind of synchronization?

yes but there could conceivably be better tricks to discover

> There are no probability distributions over possible states when there is perfect knowledge of the state.

I know very little about physics but I thought that the leading interpretations of quantum physics say that the probability distribution is all we can know about a system. The entropy is not due to due to a lack of information about the quantum state, but because the outcomes are inherently stochastic?

It sounds like you're talking about information entropy which to my understanding is analogue to but not the same as entropy in physics?

You were still going to lose export markets as international competition grew, and you were still going to shift to higher value-added service jobs as the economy developed.

America got richer and outgrew the phase where tons of factory jobs made sense. It seems pretty clear to me that well-paying manufacturing job in developed countries were the product of a particular moment in time where poorer countries couldn't do it yet. Now they can. It was never going to last.

The question was about what happens in other cases.

to understand what they said, or to understand a proof of why it would be true?

any stats class would be enough to understand what they said

welp i guess i should get my sight checked

i thought it was really cool when it picked up the grapes by the vine

edit: it didn't.

Oh yeah, I don't think I would suggest actually implementing this because it's just a lot of costs and fees to emulate what otherwise would be an index without the stock

Yeah of course you can't escape the general market risk, i'm only talking about a strategy to avoid net exposure to risk specific to the TSLA stock.

I definitely don't know much about finance, but a short is essentially being obliged to sell a position at a future date, right? You match your short position against the fraction of your ETF which is in TSLA, such that you are obliged to sell in the future exactly as many TSLA shares as you indirectly own through the ETF. This way you do not need cash on hand, because you can sell a portion of your ETF to pay off the short.

Theoretically it's not risky because in the scenario that the short becomes expensive, TSLA has gone up the and in turn TSLA has made your ETF appreciate the same amount that you owe due to the short, and vice versa if it goes down.

well not if the short matches the amount you hold in tesla stock through the ETF right?

I really don't like that these models can be branded as Deepseek R1.

I've never heard about Kagi but a paid search engine just sounds great? I assume the userbase will always be small enough that websites won't bother doing SEO for it. Maybe there's some low hanging fruit there in getting less spammy results?

sounds good

They literally pay the tax but yes often the incidence ends up on the consumers.

Clearly it could never be enough to draw that conclusion but it might be very weak evidence in one direction

I've gotten the impression that:

1. The bias is mostly due to the training data being from larger models, which were heavily RLHF'd. It identified that OpenAI/Qwen models tended to refuse to answer certain queries, and imitated the results. But Deepseek models were not RLHF'd for censorship/'alignment' reasons after that.

2. The official Deepseek website (and API?) does some level of censorship on top of the outputs to shut down 'inappropriate' results. This censorship is not embedded in the open model itself though, and other inference providers host the model without a censoring layer.

Adit: Actually it's possible that Qwen was actively RLHF'd to avoid topics like Tiananmen and Deepseek learned to imitate that. But the only examples of such refusals I've seen online were clearly due to some censorship layer on Deepseek.com, which isn't evidence that the model itself is censored.

It could be that, but also many other languages don't capitalize titles the way English does. It took me a long time as an English learner to learn that it was expected. It still feels very weird to me.

Aha, for some reason I thought they provided full-size Llama through some bundling of multiple chips. Fair enough then, anyway long term I feel like providers running powerful open models on purpose built inference ASICs will be really awesome.

Neither of the deepseek models are on Groq yet, but when/if they are, that combination makes so much sense. A high quality open reasoning model, but you compensate for the slow inference of reasoning models with fast ASICs.

At least (I assume?) a nice feature of short-form video app algorithms is that every few seconds, the users swipe or like or whatever to reveal preferences. The rate at which that data comes in from users is much higher than for long-form platforms.

I'd be more worried about actually getting content on there? But if Tiktok users are fragmenting onto Instagram Reels / Youtube Shorts etc, maybe the dominant strategy becomes to upload everywhere. In which case, maybe this platform has some semblance of a chance?

Also I didn't get to use it, but I suspect decentralized is going to be pretty tricky for a short-form video app because even minor loading times cause serious friction. If every other video load comes from a different service, it'll probably feel very slow in practice?

The weights are quantized down to fewer bits in order to save on memory. The quantization loss is going to result in worse generations.

I feel like this gets at the core issue which is food being too palatable and not satiating enough? At least on the obesity / calorie overconsumption front, I think it's just that simple. 'ultra-processed foods' or 'too many carbs' or whatever other explanations people come up for are just lossy proxies for the underlying issue which is 'foods that people tend to eat too much of'.

I figured like thirty seconds into using the site that resizing it smaller would give me more DVD bounces per second. But then during resizing i kinda cheated myself some points accidentally, and discovered that trick where they're just bouncing on every tick.

Normally I'm a sucker for clicker games, but cheating the progress (even accidentally) always kills the point of it.

I'm just responding to the portion where you said the only one that benefits is amazon. Surely at the very least, the immigrants themselves benefit if the alternative was no ability to work in the US?