That's hyperbole. They have flaws, but at the very least, when they were launched, they were arguably best in class. I'm not sure how much me sticking with them is due to familiarity and muscle memory but I know they won we over purely on merit in the beginning.

That's really surprising to me.

You should checkout startups. They would blow your mind.

"quotation marks"

Aw jeez. Just take it in the spirit it was intended and stop trying to score "more spiritual than thou" points. We are all deep one day and shallow the next. We are all workaholics and also smelling the roses. You know nothing about the author - you're hijacking a thread to grandstand.

I agree with this sentiment. So much of human discourse sort of assumes that people are consistent day to day, even hour to hour. It just ain't so!

People are rendering huge splat scenes on mobile devices using LOD. This (currently) requires CUDA and an NVidia GPU to work. I would have been much more impressed to see a demo where it was running on low end mobile hardware faster than current splat renderers can.

I'm probably being a bit of a grinch about it but the abstract doesn't address performance or hardware constraints either so I guess I'm going to have to read the damn paper.

In what world are unions never criticised? I'm in the UK and they are often reviled in the press and among people who don't work in a unionised sector. America has an even stronger tradition of anti-union feeling (maybe partly due to historic links between unions and organised crime but also because the US has often had a stronger collectivisation than most European countries - consider that the political centre in the US would be considered into right wing in most Western countries on most issues)

If the power dynamic allows it, I tend to just reply "Sorry, I'm not reading that".

(Unless it's a) trivially short or b) there's a solid reason to send me it. It's the "wall of AI text" that I generally nope out on)

Why would a non-tech person be on Hacker News? Isn't the clue in the name?

Dome looks wrong to me. Look at a few other photos - it's far from being a hemisphere

I'm yet to see any convincing argument that inference is subsidised in any substantial way. Training and speculative expansion are where the spend is from what I can see.

A few days ago Gemini redid their rate limits, making images/audio/video generation much more expensive, shrunk limits across the board (including a new weekly limit) and added more expensive tiers.

At the moment you can pay $20/month to do thousands of expensive queries a month (involving file uploads, the Pro model, extended thinking), and evidence suggests that heavy users are not profitable.

I agree that heavy users are probably not profitable but that's the way the economics of subscription services tends to work across the board.

I'm arguing that even if inference isn't profitable right now it's not orders of magnitude off. Whatever pricing emerges for models equivalent to current frontier models won't be significantly higher than the current API pricing.

There are already enough small companies without tons of VC money to burn that are serving up nearly-frontier llms at prices lower than the big players are charging. They can't all be subsidising? These are companies without any moat or any IP.

If inference was profitable - they'd tell us. Msft, goog, public companies. They'd break out the numbers and show us, if they were good.

But instead, all we get is known liars going on podcasts and repeating "stylized facts" that aren't literally true about their supposed profitability on inference, from companies losing billions per year in a situation where they don't have to tell the truth.

That is VERY far from a convincing argument that they are profitable. So I can & will safely conclude that the opposite is true.