Twenty-five years ago it was cryptography, today it's model weights
71–80 of 168 posts
Re: Twenty-five years ago it was cryptography, today it's model weights
#72Germany or Europe is no different.
Re: Twenty-five years ago it was cryptography, today it's model weights
#73Re: Twenty-five years ago it was cryptography, today it's model weights
#74With these models, what is the moat? Is it the architecture itself or the training data? Why can’t China create these models themselves? It just feels like theater to restrict models. Can someone tell me why I am wrong if I am?
Re: Twenty-five years ago it was cryptography, today it's model weights
#75The same argument applies to DRM and piracy. DRM only disadvantages people who want to consume content legally. Only one person has to bypass it (be it through cryptographic or software failings or the good old fashioned analog hole ala telesync) then upload the resulting unencumbered media file to a piracy site. As a result, it's not just cheaper, but also easier and more reliable to steal media than it is to buy it…
Cheaper & easier - not by much. Australia blocks all the good torrent sites, so now I need a VPN, a reputable one. Also need terabytes of network connected storage, and usually I need more and more of it over time. Better and more networking equipment to buy and maintain. Media management and streaming software that works locally and remotely, to (maybe buy) and maintain. Or I can usage discount offers for the variou…
Literally none of these things are reality other than for the very specific means you are aware of.
> The streaming services also don’t come with who knows what malware like pirated media can.
What is this, an MPAA pitch from 2007?
Re: Twenty-five years ago it was cryptography, today it's model weights
#76With these models, what is the moat? Is it the architecture itself or the training data? Why can’t China create these models themselves? It just feels like theater to restrict models. Can someone tell me why I am wrong if I am?
If China wants to compete at building the best forefront models, they'd have to make similar sacrifices (though I doubt they'd be as messy about it).
But if they can are content with 95% as good at 5% of the cost, then they can focus on distilling instead. I don't think the architects of the sacrifices anticipated just how easy models would be to pirate.
So the most is real, but it's effective for a rather constrained set of scenarios.
Re: Twenty-five years ago it was cryptography, today it's model weights
#77Earlier quoted context omitted.
I think that's also wrong. I think that the primary reason media publishers use DRM is not to prevent copyright infringement. It's really to exert power over the entire ecosystem and use that power to implement rules around content distribution and use that are much more restrictive than the actual copyright law itself.
To copyright owners, there really is no difference. In many countries, notably not the USA, copyright is bound up with the broader concept of authors' rights, including the right to control whether, where, when, and how a work is exhibited. Rightsholders in the U.S. still believe in authors' rights from a moral standpoint; Jack Valenti wanted to outlaw VCRs because he was afraid of the rise of bootleg tapes, sure, bu…
Re: Twenty-five years ago it was cryptography, today it's model weights
#78Earlier quoted context omitted.
Drm is not targeted at the hardcore pirate, it's targeted at a potentially larger percentage of people who would be tempted to get the thing for free if it was easy. Drm may be circumventable, but any complex set of steps for doing so is enough to deter that target audience.
Piracy is a service problem. Copyright monopolists always defeat themselves with their obnoxious greed. Things were easy enough when Netflix had a complete catalog, now every studio wants to roll their own shitty streaming services in order to cut Netflix out. Result is a fragmented mess where you want to watch a great film but you have no idea where the hell it even is, and even if you find it you discover it doesn'…
Re: Twenty-five years ago it was cryptography, today it's model weights
#79Earlier quoted context omitted.
Piracy is a service problem. Copyright monopolists always defeat themselves with their obnoxious greed. Things were easy enough when Netflix had a complete catalog, now every studio wants to roll their own shitty streaming services in order to cut Netflix out. Result is a fragmented mess where you want to watch a great film but you have no idea where the hell it even is, and even if you find it you discover it doesn'…
> Piracy is a service problem > Meanwhile the pirates offer Blu-Ray remuxes with commentary tracks, subtitles for every language, DRM free files that just work on your video player of choice. I think the only people who believe that pirating is easier than using the streaming services are people who spend so much time with computers that they round the mental overhead of navigating torrent websites, finding the corre…
Heh, that’s what you think modern piracy is? How quaint. In reality I have my own “Netflix”, with access to effectively all media ever created and I click play. That’s it. That’s the entire process. No credit cards, no finding which service has it, no finding out that the service doesn’t have it in my country at this time, no ads. I punch in what I want to watch, I click play, and it plays.
> Nobody is ever going to convince me that pirating is actually easier than finding the thing I want to watch and putting it on my credit card.
Sounds like your problem. Though again, that’s already more steps than it is for me.
> which is that some people don’t want to pay for things
I was happy to pay, and did so for years. Then the services got worse, increasingly fragmented and ever increased in price as service got poorer. Now I’m no longer happy to pay; and have a superior service to boot. The industry successfully convinced me to give piracy up, before driving me back to it a decade later.
Re: Twenty-five years ago it was cryptography, today it's model weights
#80With these models, what is the moat? Is it the architecture itself or the training data? Why can’t China create these models themselves? It just feels like theater to restrict models. Can someone tell me why I am wrong if I am?
Its the time and the electricity needed to train them. The US sacrificed quite a lot in order to make this happen. We're tolerating a measles outbreak, our economy is in the trash due to tariffs and draconian immigration enforcement. Basically every sector that's not AI has been poisoned in one way or another in order to drive sufficient investment towards AI. If China wants to compete at building the best forefront…
Incompetence is not sacrifice.