They “get the job done,” perhaps even provably, but:

• It’s less flexible for fighting anti-adblocking.

• Hence, it’s more work for Adblock devs/listmakers to maintain.

You can see the UBO devs already complaining/stressing about it. Hence them recently dropping Facebook from their 1st-party lists because they’re tired of fighting paid devs, or increased reports of UBO Lite specific issues.

I think that’s Google’s real intent.

They aren’t trying to kill adblocking, just make it inconvenient enough so the vast majority don’t use it effectively. It’s a form of “soft” censorship, sort of.

I mean… for me:

Try GOG first, see if a game is available there. Then Epic (once it gets Linux support). Then Steam 3rd party stores like Fanatical. Then Steam

Seems pretty simple to me, personally.

• I’d rather pay for DRM-free and an offline archive, always. The installers are trivially easy to use with Proton.

• If not, then I’d rather pay the devs more than the store if I have a choice (though no Linux muddies this).

• If not, I’ll take the slight discount over Steam if there is one.

• Failing all that, just buy on Steam.

And in reality, the vast majority of my purchases are on Steam anyway, because that’s generally the only place the games I’m interested in are available to buy.

You have to launch Chrome with this command-line argument:

--disable-features=ExtensionManifestV2Unsupported,ExtensionManifestV2Disabled

And then, I think you have to re-enable it with config flags as well:

chrome://flags/#extension-manifest-v2-deprecation-warning         [Disabled]
chrome://flags/#extension-manifest-v2-deprecation-disabled        [Disabled]
chrome://flags/#extension-manifest-v2-deprecation-unsupported     [Disabled]
chrome://flags/#allow-legacy-mv2-extensions                       [Enabled]

I have not tried this personally, though.

As far as I can tell, that paper explores a model’s ability to assess its own future text output beyond what “regurgitating training data” would suggest. It makes sense that it could do this better than an outside model; it’s exploring its inner state with each token, though the test is still interesting.

That has nothing to do with sampling, though.

…Maybe an analogy would illustrate this better. First, I don’t mean to anthrophomorphize LLMs, but the human brain is a good example.

The paper is analogous to asking a human brain to assess its own thoughts and tedencies. Of course it can do this well; it can think conscously and run thoughts through its subconscious parts.

What OP is proposing, is analogous to “tell your eye receptors to see less green.” Or “get your vocal cords to omit a certain frequency when you speak.”

There is no wiring in the human brain to do this. Vocal cords and cells that sense green in the eye are effectively “external machinery” to the brain that it does not have such control over.

LLM sampling is the same.

Tokenization and sampling are external machinery. They are code, hardcoded programming, completely outside the LLM weights. You can’t tell an LLM to alter its own sampling because its literally impossible, and it can’t manipulate its own logit spread mathematically because that’s invisible inner machinery.

Could you do this with custom sampler/logit manipulation code and a tool harness?

Sure. Maybe.

But Claude is not rigged to do that, and a system prompt won’t change that.

…Okay. Technically, that’s true, going by strict semantics.

But the Helium and Ungoogled Chromium are going to lose MV2, soon. I feel like “still supports uBlock Origin” doesn’t sufficiently convey that to Lemmy commenters trying to read the situation. That’s all I meant, though I was perhaps imprecise with my words.

In other words, practically, if a users want to use UBO, they should know these browsers will not support that in the near future even if, strictly speaking, they support it right this second.

I believe this means Brave will lose support too, unless they intend to extensively modify Chromium to maintain compatibility themselves (which would be odd, as Brave doesn’t really need UBO since it has its own adblocking engine?)

True, sometimes. But they don’t have the sheer manpower dedicated to security that Mozilla has, nor the institutional policies that try to mitigate losing a key/account or something like that. Also, technically, vanilla Firefox is first to get the zero-day patches before they filter downstream.

Not that the forks are dangerous or anything, generally.

For a second, I thought this required creators to watch 8000 hours of videos, and felt relief.

It would at least throw up an obstacle for bot/spam accounts. They’d have to make the effort of watching a bunch of videos.

…But no.

Well…

With a fork, one is very dependent on the browser dev. Not just on their benevolence, but on their security practices too. And on them pulling in upstream security patches, quickly.

TBH, this is why I do finance stuff on vanilla Firefox. I love browser forks, I really do, but I’m a bit more cautious with them, too.

First, screw Chrome.

BUT. These headlines are clickbait.

UBO has been disabled in Chrome for a long, long time. This is nothing new. Its old news.

Ublock Lite replaced it, and it’s not going away. It’s inferior and problematic and harder on devs, yes, but unfortunately Chrome users didn’t seem notice.

What’s being disabled is a crazy-difficult-to-enable flag that basically zero Chrome users actually used, and that only (a few) downstream Chromium forks use.

Not for long. The patch that removes support is coming to Vivaldi soon.

The updates aren’t just disabling it anymore, they’re stripping the path out entirely, so it would be difficult to maintain support.

This is not true.

The flag that UnGoogled Chromium used to support it is going away, and AFAIK there is no fix. Helium inherited the same patch from UnGoogled Chromium.

They will both lose support. See their relevant GitHub issues.

Brave uses a custom, integrated adblocking engine; it technically doesn’t need UBO.

The Commedore phone, if you want something minimalistic that does have Android compatibility.

That’s a significant security risk, unfortunately.

If that happens, I’d suggest a fork with a built in blocker, one that still gets security patches from upstream Firefox.

Okay.

I don’t agree. But let’s say I agree.

…Just don’t use Claude?

Use an LLM without a watermark; there are hundreds to pick from.

In other words, if one is going to try to hide automated writing, I think there should be a bare minimum effort to do so. That includes:

• Reading/checking the text, to see if it makes any sense.

• Actually trying to pass it as human.

90% of slop is brain melting slop because this minimum bar isn’t even met. And all Claude’s watermark would do is catch that bottom of the barrel; it wouldn’t censor anyone.

I mean, practically? Orion’s Arm.

Taking care of humanity is literally trivial.

There’s a big soup of civilizations and frontiers and mysteries to explore. Some danger to keep it interesting. Infinite and transcendental pleasure to experience, entire simulated worlds you can try.

But it’s grounded in real physics, so you don’t get that big can of worms from living in a universe that isn’t… self-consistent.

In other words, its as sci-fi as you can possibly get without things getting timey wimey.

But my adventure-yearning-heart says Avatar.

Bending is just so freaking cool. It’s the best magic system ever.

Yeah, the PS3 was weird. It wasn’t just a “different ISA” like the Xbox 360, the PS3’s processor was a completely different programming paradigm.

I haven’t kept up, but it makes sense that AVX512 helps. That brings desktop CPUs closer to “vector processors” like cell.

Also, I would contest this:

Never before has the government had so much power.

I feel that the legitimacy of governments is collapsing and power is going to corporations, in a hurry.

In other words, we got so busy worrying about direct government spying we let a handful of corporations completely take over the information sphere, instead.

I’m rewatching Person of Interest, and it hits so different now.

That post 9/11, Snowden, surveillance-state-fear era almost feels… quaint.

As an addendum I forgot:

Your idea was actually already implemented:

https://github.com/ggml-org/llama.cpp/pull/9742

XTC is a novel sampler that turns truncation on its head: Instead of pruning the least likely tokens, under certain circumstances, it removes the most likely tokens from consideration.

It’s a very cool idea: it chops off the most likely token when the rest of the sampling indicates it probably can.

In my tests back then, the results were… mixed, but the idea is fascinating.

But you can’t do this with Claude, as its a closed model and their sampling is years behind cutting edge.

See: https://gist.github.com/Hellisotherpeople/71ba712f9f899adcb08b94bce20d5397

Fair enough. This is true, apologies for being so hostile.

I’m pretty stoked about PS5 emulation in the future too, though honestly I have a pile of other console games I should have, but have not tried to, emulate yet.