nabakin

Is there a recording of their seminar anywhere?

> type Ia supernovae, long regarded as the universe’s "standard candles", are in fact strongly affected by the age of their progenitor stars.

A key point in the article. From what I understand, this is the main way we measure things of vast distance and, from that, determine the universe's rate of expansion. If our understanding of these supernovae is wrong, as this paper claims, that would be a massive scientific breakthrough.

I'm really interested in the counterargument to this.

Exactly my thought as well. You can have all the horsepower you want but if it doesn't convert the electricity efficiently, it's not going to be useful for normal consumer cars.

I wonder if there's something we can do. A social network that penalizes misinformation and rewards expert analysis. Something like that would never be as big as general social media but maybe it could be developed for a small set of users who care enough about truth to contribute to it and grow it over time

If companies get large enough, you no longer have a free market. You simply have a company so dominate and powerful, they can undercut any new company or product. Not on merit or offering a better product. Purely based on size and its entrenched position.

Let's say company A is monopolizing a market. New company B finds a way to make company A's product providing exactly the same value, except half as expensive. A healthy market would reward company B and consumers both by allowing that innovation to spread, providing a return for those that invented it and a cheaper product for consumers, leading to greater prosperity for society. However, in the market you speak of, company A can simply use their orders of magnitude more wealth to undercut company B for a time until they go bust.

What incentive is there to innovate, to improve, in a market that doesn't reward it?

Ownership is a useful framework for sure, but if it comes at the expense of destroying competition, I think we can agree that's not healthy for the economy and society as a whole.

Unfortunately this isn't the first time a government has banned Android devices which are not licensed by Google. GrapheneOS has a list of them[1]

[1] https://grapheneos.org/articles/attestation-compatibility-gu...

I keep seeing comments on HN that misunderstand what's happening with Mozilla and it's kind of frustrating.

Right now, if you were to take away Google's money, Firefox would not be able to compete with Chromium and Safari. It would die.

All these side-projects are attempts to find a source of revenue aside from Google and are necessary to Firefox's survival. So saying they should stop doing them, completely misses the point.

Unless we want Firefox to die, we should understand Mozilla's situation and encourage this exploratory process, not hate on it.

They're trying to diversify their revenue so it doesn't all come from Google. All these 'self-indulgent projects' are attempts to actually make enough money to compete with a multi-trillion dollar company's resources because they know they can't compete long-term.

Yeah you're right (although, they started to open source some of that recently iirc). I meant SOTA for inference engines we can actually download and use ourselves.

> On throughput-focused benchmarks, Tokasaurus can outperform vLLM and SGLang by up to 3x+.

Looks like they don't compare to TensorRT-LLM throughput numbers which, last I checked, are SOTA in open source.

I think you can blur the lines to an extent and if the judge isn't very capable, pull one over on them, but to say an LLM or ML model is not artificial intelligence to a capable judge with a prosecution who points out there is an LLM, it would be hard to get around imo

The court system is responsible for interpreting law so playing semantic games doesn't work very well unless you have a corrupt judge who sides with you and is willing to interpret it in this way

I keep seeing comments on HN that misunderstand what's happening with Mozilla and it's kind of frustrating.

Right now, if you were to take away Google's money, Firefox would not be able to compete with Chromium and Safari.

Those 'harebrained' initiatives are attempts to find a source of revenue aside from Google and are necessary to Firefox's survival. So saying

> stop wasting money on all those harebrained non browser initiatives and concentrate on ... the Firefox browser

completely misses the point.

Unless we want Firefox to die, we should understand their situation and encourage this exploratory process, not hate on it.

That was my first thought too. A cursory search indicates we would see a lot more gravitational lensing in our observations if that was the case.

Found a couple of videos on it too

https://youtu.be/qy8MdewY_TY

https://youtu.be/d0wV5frSb6s

Could it be that the observable universe is one of these stable local pockets and the antimatter to balance it out is simply not observable to us?

Hey Jeff, thanks for your work! Quick question for you, are you guys using Azure Speech Services or have these TTS models been trained by OpenAI from scratch?

I don't think I've ever seen the inside of an actual undercover van before. Crazy picture. Do we know anything else about them?

I think the only way is to check your dataset for the benchmark leak and remove it before training, but (as you say) that's assuming an honest actor is training the LLM, going against the incentives of leaving the benchmark leak in the training data. Even then, a benchmark leak can make it through those checks.

I think it would be interesting to create a dynamic benchmark. For example, a benchmark which uses math and a random value determined at evaluation for the answer. The correct answer would be different for each run. Theoretically, training on it wouldn't help beat the benchmark because the random value would change the answer. Maybe this has already been done.

Automated benchmarks are still very useful. Just less so when the LLM is trained in a way to overfit to them, which is why we have to be careful with random people and the claims they make. Human evaluation is the gold standard, but even it has issues.

I'm not so sure it's impressive even for mathematical tasks.

When ChatGPT came out, there was a flood of fine-tuned LLMs claiming ChatGPT-level performance for a fraction of the size. Every single time this happened, it was misleading.

These LLMs were able to score higher than ChatGPT because they took a narrow set of benchmarks and fine-tuned for those benchmarks. It's not difficult to fine-tune an LLM for a few benchmarks, cheaply and beat a SOTA generalist LLM at that benchmark. Comparing a generalist LLM to a specialist LLM is like comparing apples to oranges. What you want is to compare specialist LLMs to other specialist LLMs.

It would have been much more interesting and valuable if that was done here. Instead, we have a clickbait, misleading headline and no comparisons to math specialized LLMs which certainly should have been performed.

I see Seedvault allows you to backup and restore app data. Are there any issues transferring app data from one device to another? Screen size, Android version, hardware, etc could all be different.

Fyi the mobile image swipe mechanic on the linked page is inverted. Swiping left takes you to the image on the left, instead of the image on the right and vice versa.

GPU cost ÷ GPU token throughput = cost per token, will get you close then you compare the cost per token between solutions.

To explain how to calculate it precisely, I would have to write a blog post. There are dozens of factors that go into it and they vary based on your use case like GPU type/setup, cloud provider, inferencing engine, context size, the minimum throughput and latency you'd be willing to have your users experience, LLM quantization, KV cache configuration, etc.

I'm sure there are cost analyses out there for Llama 3.1 70b you could find though.

I assume it's the same cost as Llama 3.1 70b

How does that work? They force AOM to pay up and then I guess either AOM passes the royalty burden onto AV1 users or they take the hit and pay it themselves?

Do you think AOM is going to start charging royalties?

Honestly, I don't understand why the threshold for war seems to be ground invasion. If you're shooting missiles at another country, that's war too. Israel is at war with Syria, Iran, and Lebanon imo. Not to mention Hamas.

You might be confusing it with Lebanon, but Syria has been bombed by Israel and is pretty unstable in general so it's impressive they were able to do this research regardless. People shouldn't be downvoting you for asking a question.