This isn’t true, royalty payments are fairly similar across streaming services, on account of how much money gets in via the subscription prices. Trying to divide the royalty payments by stream count is largely irrelevant on account of them having no correlation with the royalty payments, these are instead based on the fraction of total streams on the platform.

It’s even less true on account of artists not getting directly paid through streaming services, it’s all getting funneled to the people who own the rights to the music itself, which is the labels.

ICE ads have been running on Amazon properties (Twitch, Amazon Prime Video), Google properties (YouTube), Meta properties (Facebook, Instagram), LinkedIn, Twitter, Spotify, etc. Basically any big tech platform with an ads business.

Betteridge’s law of headlines applies.

dark pattern

This is not what dark pattern means.

Also, I don’t think enabling developer options is required to install arbitrary APKs.

Indeed, wild statement if I ever heard one

There are tariffs on Chinese EVs in the EU as well.

The primary use-case for LLMs once again appears to be gooning

Calling it now: literally only gooners will be interested in this character, no one else will care

Is the organization in the room with us right now?

It’s probably best not to make assumptions before we know for sure.

Same as the motivations behind Charlie Kirk’s shooter should not have been assumed to have been either left or right until we know for sure.

Not saying I disagree with the statement, but where did you see that these men were Republicans? I didn’t find any reference to any political motivation in the article, but I also didn’t read it particularly thoroughly

She ratfucked it at the last second, delaying implementation and then when she brought it back post-election she lowered the prices, which dug a massive hole into the MTAs budget, which relied on that projected revenue for their future investments.

To be clear, I am all for congestion pricing. I live in one of the cities where it has been successfully implemented.

This is giving Hochul too much credit.

Have you already forgotten the congestion pricing debacle?

It’s good that she did this, but by no means should she be forgiven or continue to work in her position

So quite literally worse than a coin flip, then.

Streaming lossless audio will use up three to six times as much data, along with the higher processing demands to play them back, so there’s always a penalty involved. We didn’t invent codecs for no reason.

Based on the vibes of every internet comment field, literally everyone and their mother wants to stream lossless.

You’re right about the audible benefits however

I have no idea what they think this will accomplish, to be honest. It has the legal value of posting on Facebook that you don’t allow them to use your photos.

No, the old model does not have the training data. It only has “model weights”. You can conceptualize those as the abstract rules that the old model learned when it read the training data. By design, they are not supposed to memorize their training data.

I expressed myself poorly, this is what I meant - it has the “essence” of the training data, but of course not the verbatim training data.

To outperform the old model, the new model needs more than what the old model learned. It needs primary sources, ie the training data itself. Which is going to be deleted.

I wonder how valuable in relative terms the old training data is to the process, compared to just the new training data. I can’t answer it, but it would be interesting to know.

I think we’re in agreement with each other? The old model has the old training data, and then you train a new one on that model with new training data, right?

I guess it depends on how important old data is when building upon new models, which I fully admit I don’t know the answer to. As I understand it though, new models are not trained fully from scratch, but instead are a continuation of the older model trained with new techniques/new data.

To speculate, I guess not having the older data present in the new training stages might make the attributes of that data be less pronounced in the new output model.

Maybe they could cheat the system by trying to distill that data out of the older models and put that into the training data, but I guess the risk of model collapse is not-insignificant there

Again, limited understanding here, take everything I speculate with a grain of salt

They essentially still have the information in the weights, so I guess they won’t fret too much over not having it in the original training data.