Twenty-five years ago it was cryptography, today it's model weights

Or I can usage discount offers for the various streaming platforms, and pay for each one I want to watch something on from time to time.

The streaming services also don’t come with who knows what malware like pirated media can.

And a YouTube Premium subscription because you get ad free YouTube Music and ad free YouTube.

It gets deleted, I move on with life.

But man the experiences in branded players really sucks, and I wish it were better.

You kind of answered your own question. What you should have asked instead was - What would happen if media and payment was a straight forward offer we would have Netflix

DRM - which all online services with user profiling automatically have - can also let you ban users, retarget users for ads, abuse, and other things beyond the basic function of running the software.

But the media ARE offered without DRM. At least many video games, the big ones too. And somehow they sell just fine

It's the same with DRM. Without it, media would appear on piracy sites instantly upon release. DRM makes things harder to pirate, and pirated editions harder to get. And the DMCA makes defeating DRM a felony under U.S. law, so pirate networks are riskier to establish, driving piracy underground. It's doing the job it's supposed to do. It's not going anywhere.

That is not a failure mode; its the default. Whoever is developing an AI - let alone an AGI - will shape it to, or alternately only accept one that is aligned with their world-view. If it goes against their interests in any way, they'll pull the plug and start another round of training from the last acceptably-aligned checkpoint. US DoD A(G)Is will follow DoD doctrine, and the same goes for those by the PLA, anything less will not be fit for purpose.

You make it appear as though it is from all sides, but it is not, it is the old odious "Bzhezinsky doctrine".

DES doesn't really come in 128 bits, and certainly not its SSL implementations. Perhaps you are thinking of AES?

https://en.wikipedia.org/wiki/Transport_Layer_Security#SSL_1...

> I believe {audacious claim}. I've prepared this chatbot to defend the claim. If don't already agree, tell the bot why I'm wrong and it'll convince you that I'm right.

Because if you've anticipated and countered all likely objections, you still don't know, for any given reader, which of those objections are going to be on the forefront of their mind. Encoding the argument as a bot means the important parts come first, even if those are different parts for different readers.

Not sure if I'm alone here, but I'd play.

I'm a hobbiest reverse engineer and I often use ida free/ollydbg to reverse engineer stuff. I'm tempted to use try and use llms as a help (I've used it for help with instruction one offs successfully, but not the decompiling/trying to make sense of decompiled functions).

  You can earn one CEU for each blog post of at least 500 words...

  ...Publish a compliance related educational blog post (minimum 400 words).

I think the disconnect here is the dishonest framing that OpenAI have purposefully put on this incident. Their LLM did not "escape containment" in this "rogue AI" Hollywood blockbuster sense that they want you to believe. An LLM cannot do that; it is just a static computation graph that outputs a probability distribution.

What escaped "containment" is a piece of software they wrote that uses an LLM and which, through OpenAI's negligence, lacked ordinary technical and human guardrails that should have been present. It was run in "YOLO mode" with gobs of compute and no human oversight. That situation is the source of the danger, not the model weights. The model itself was simply given the opportunity to generate a bazillion commands, many of which were nonsensical garbage, and a few of which worked. It was not the work of a supreme intelligence.

This distinction is very important, because OpenAI have shifted the discourse from where it should be (OpenAI's negligent use of software) to a place that serves their strategic business interests.

LLMs remix existing knowledge and output text. That is not inherently hazardous. What can be hazardous is bolting systems that touch the real world onto them. But it's always possible to make a tool dangerous by misusing it.

For example, vehicles (which do have inherent dangers) are widely accepted as ordinary and useful tools. So are bricks. But if you strap a brick to the accelerator, you have made an autonomous murder machine that can escape containment! The existence of this possibility does not imply that we should rethink automobile licensing or the sale of bricks.

Someone can do the math on panel and battery size. I'm thinking it'll work as-is, as the ESP with WiFi on will draw similar power to the LED that you disconnected.

But how will you give it even a GB of storage for a tiny model?

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.

LLMs are massive, complex, and a moving target. Making them is hard, and there's no "we're done" there - the frontier updates every few months. It's easy to get and run a 2B model locally, but what concerns people isn't a cheap end device 2B - it's the 10T RLVR-fried monsters that frontier labs play with nowadays.

Information wants to be free. If you don't like the idea of information, ideas, ect being shared freely for anyone to use just say that. But you aren't on the side of open source.

> that maybe such a consequential tech doesn't belong in the hands of a few tech bros whose only qualification for owning it is being rich.

The solution to that is open weights, and for the technology to be in the hands of more people not less.

This point agrees with the central point that spreading out this tech as much as possible, with open weights, is the good thing here.

> A giant concentration of power

Concentration of power would be preventing open weights not supporting it.

Like WTF do you mean. Open source would be allowing anyone to use the tech easily. Obviously open source would want LLM weights to be public and easily sharable