It’s not so simple, there are papers on zero data ‘self play’ or other schemes for using other LLM’s output.

Distillation is probably the only one you’d want for a pretrain, specifically.

Paraphrased by Wikipedia: https://en.wikipedia.org/wiki/2025_India–Pakistan_conflict#Analysis

[The Times] reported that India felt frustrated after Donald Trump public claims of mediating a cease-fire, presenting both countries as equals and downplaying the terrorist attack that triggered the conflict, and that India had hoped any U.S. involvement would remain discreet, and Trump’s portrayal of both countries on equal terms was seen by Indian officials as politically sensitive and diplomatically frustrating…

On 21 June, Pakistan announced it would nominate Donald Trump for the Nobel Peace Prize, citing his role in brokering the ceasefire. Pakistan credited Trump’s diplomatic intervention, though India denied any U.S. mediation.

Like it said, seems like India assumed the modest level of mediation would be confidential (clear miscalculus on their part), while Pakistan, err, trumped up the magnitude of the intervention to paint themselves in a better light, possibly because they’re at a military disadvantage, and felt grateful for the help.

Seems like there was some backchannel involvement from many countries (like “Saudi Arabia, Iran, the UAE and the UK” and indeed the US), but Trump couldn’t help himself and loudly claimed credit before the ceasefire was even announced.

Now India’s annoyed (hence their flat denial).

I like this explanation, it ‘fits’ all the involved characters, including Trump blotting out the sun and killing any nuance to the situation.

Wasn’t his involvement pretty minimal anyway?

https://www.news18.com/india/trump-wasnt-directly-involved-shashi-tharoor-on-india-pakistan-ceasefire-9394156.html

Or is that spin from India? Somewhere in between?

I elaborated below, but basically Musk has no idea WTF he’s talking about.

If I had his “f you” money, I’d at least try a diffusion or bitnet model (and open the weights for others to improve on), and probably 100 other papers I consider low hanging fruit, before this absolutely dumb boomer take.

He’s such an idiot know it all. It’s so painful whenever he ventures into a field you sorta know.

But he might just be shouting nonsense on Twitter while X employees actually do something different. Because if they take his orders verbatim they’re going to get crap models, even with all the stupid brute force they have.

There’s some nuance.

Using LLMs to augment data, especially for fine tuning (not training the base model), is a sound method. The Deepseek paper using, for instance, generated reasoning traces is famous for it.

Another is using LLMs to generate logprobs of text, and train not just on the text itself but on the *probability a frontier LLM sees in every ‘word.’ This is called distillation, though there’s some variation and complication. This is also great because it’s more power/time efficient. Look up Arcee models and their distillation training kit for more on this, and code to see how it works.

There are some papers on “self play” that can indeed help LLMs.

But yes, the “dumb” way, aka putting data into a text box and asking an LLM to correct it, is dumb and dumber, because:

• You introduce some combination of sampling errors and repetition/overused word issues, depending on the sampling settings. There’s no way around this with old autoregressive LLMs.

• You possibly pollute your dataset with “filler”

• In Musk’s specific proposition, it doesn’t even fill knowledge gaps the old Grok has.

In other words, Musk has no idea WTF he’s talking about. It’s the most boomer, AI Bro, not techy ChatGPT user thing he could propose.

I hate to sound preachy, but this is a good example of “rivals” peacefully meeting.

So many people I meet IRL seem conditioned to think this person they hate on the internet would be someone they’d shout at like they’re an axe murderer, in the middle of a murder. It’s the example they see. Death threats are, like, normal on Facebook or TV News or whatever they’re into, apparently.

Again at risk of reaching… this feels like positive masculinity to me.

And leaders acting like adults.

Welcome to Lemmy, heh.

To expand on Shrubbery, you can read the Guardian’s editorial standards and the base ones they follow: https://image.guardian.co.uk/sys-files/Guardian/documents/2011/08/08/EditorialGuidelinesAug2011.pdf

according to people familiar with the…

In journalism, this is common lingo for “our journalists personally communicated with sources (more that one) that wish to remain anonymous.” It’s the guardian’s legwork and integrity for shielding anon sources here, not them copying it from some other site (which they’d link if they did).

Exactly. Hence the importance of linking the original source…

Sorry the be pedantic, but why not link the original source?

https://www.theguardian.com/world/2025/jun/19/trump-caution-on-iran-strike-linked-to-doubts-over-bunker-buster-bomb-officials-say

Since it’s their inside sources.

What’s also messed up is that it’s that hard for this kid to get citizenship. That’s just a broken system.

What an ad. This is the second greatest YT video ever made.

It doesn’t have to be one or the other. You can do most of your work in linux and boot windows as a ‘secondary’ OS for stuff like adobe? I do this, and share NTFS SSDs/hard drives between them.

You are hitting weak points of linux though. I do all media work on linux (a lot through vapoursynth or ‘lower level’ frameworks than resolve I suppose), but TBH do most of my gaming on Windows, not just for convenience but for performance reasons too.