knollimar

Moonshot can prove they added value with their paper. Where's Anthropic's proof?

If not I propose HLRF.

Please it's almost my primary bench for drafting understanding.

I'm almost about to post that xkcd 810

I feel like that was true up until 6 months ago.

Doesn't China smelt most of the world's aluminum due to inexpensive energy?

I thought that was their thing

Aren't they on the "we, the safe AI, must win at any cost" justifiers?

What's a little contract violation if the fate of humanity is at stake?

Im excited for the labs with more data RLHFing this (e.g. cursor). That model will be crazy.

Meanwhile the U.S. suffers under Vernor vs Autodesk

Disgusting. If you can't list a fee you have no right to charge it. I don't care what your unit cost structure is.

Do you think the vocabulary is load bearing, or do you need to caveat it's genuinely the lack of real ideas that's the problem?

Oh I thought you were just one of them but stealth mode. That makes sense given the graph thing, thanks. Needed some caffeine and a nap.

Did you just emdash not-just-X-it's-Y unironically or am I missing the satire?

Are you implying something about my spelling on a phone in my free time transfers to my work performance? Silly if so

The problem is they do good enogub work or don't care about the quality when they should push back.

Or they will refuse to self invest, which makes them a worse long term investment for a company.

It seems marginally better than average assuming it's LLM generated

Give me that style guide and spread it around then!

There's tact to be deployed when you know a lot.

People are less frustrated with the actually meme if it's insightful and not some pedantry.

Bait title for bait point.

----------------- If weights aren't the same as the data even though they produce the data, how are weights that produce similar outputs different?

The "legal view" smuggles the cintested distinction for distillation by saying it's substantially the same product.

That seems like a corcular argument.

Is this not chiefly a complaint about the rules? Saying "if Google doesn't like the rule it can leave" is a non argument.

Would you be doing that without the FOMO?

Is it actually a problem or are they just choosing their battles by not calling out 24/7 access to decision makers as unreasonable?

Not trying to take it literally, but aren't there costs vs performance tradeoffs? Like the py.toHDL would have like (maxSize,maxCost,minThroughput) as free and that would determine energy usage?

And a GPU is already pretty optimized for inference, no? Like isn't it a bunch of FP mults? I don't think HDLs do well with that, either.

Doesn't that assume that VHDL is trivial? I feel like there are tons of performance tradeoffs or hardware designers wouldn't have jobs

I think there's a difference. Sometimes you'll see some random LLM tells like load-bearing if looking, but you don't notice on a first pass. Those are fine. Slop is slop

I get a lot of this in design docs every time I give it a negative constraint:

[Suboptimal choice]

And here's why it's not suboptimal -- you said X sucks and notto do X, but this choice is not technically X, it's just really similar and shares that sucky property.

Squircle from above!

It's prideful too since just undoing it would be an actual admission of failure. This is a hollow apology and compromise

Isn't that the point of e2e encryption? Maybe I have a privileged US take but I've always not cared much for a border crossing in a round trip. It always seemed like a needless expense to have a full vertical server stack in each country to me.

Wondering if someone can explain the logic

Digital sign that only shows the ad on red light then.

isn't throwing that into a [insert financial vehicle that gives 99.99999% safe returns] going to destroy that when you factor in electricity costs?

Or even just electricity costs vs token cost