riskable

How very considerate! Steal from the rich, not the poor.

Conversely, don't bother with poor people!

Now we just need proper Robin Hood malware that "levels out the playing field" as it were, by stealing from the rich and moving that money into the poor people's accounts (that have also been hacked).

So, so many people would disagree with you. "Bouncing an idea off an LLM" is fruit of the poisonous tree in a lot of people's eyes because you involved an LLM at all.

I see this stance every day on writing Discords. Just mentioning that you used an LLM for anything (even unrelated to the story you're working on) can get you banned.

The problem is that advertising works! I hate that it works!

Marketing and advertising aren't the same thing but I swear both sides of that coin have devolved via some process of carcinization where their claws emerge from the darkness everywhere people gather. Especially places where those claws are inappropriate and unwanted and yet it still works.

I actively/mentally block products I see in ads now. For example, when I use Amazon's terrible search feature, I'll make a mental note of which listings are "sponsored" and make sure NOT to buy them. Ever, if possible.

When I see ads for things on YouTube, I automatically assume it's a scam... Because it almost always is!

We really need more terms for AI-assisted writing. There's degrees to everything and writing is one of those things where there's LOADS of gray areas. Did you use AI for research? What about generating random names for characters/places? Did you use it to check your grammar (why do people even care about this? They do!)?

Anything touched by AI is starting to become "the fruit of the poisonous tree" for a lot of people and it's getting ridiculous. I've chatted with people who will refuse to read a story because they think the cover art was generated by AI (even though they don't really know; they're just guessing). "It's AI slop."

One of the listed stories was almost disqualified for that very same reason (AI cover art). Another was controversial because the author used AI to generate a random name for a disease.

If we had better terms for the levels of AI involvement, maybe people would get a little less worked up about it?

Exactly. I don't have the power to replicate the testing but this is the type of thing that easily makes the news!

If you have the power to do so, why wouldn't you perform your own testing right now and release the results? It's a marketing/status gold mine.

If you find that western models don't have the described anti-china bias, then maybe you won't publish anything since it's not as interesting—that would be evidence there's no conspiracy.

Hence my conclusion that if we don't hear anything soon, the anti-china accusation is false.

I agree that shared data centers and an international team of AI researchers would be best. The key, though, should be absolute transparency. All the data should be accessible to the public in such a way so that the resulting models can be replicated. Even if it's unlikely that anyone would have the hardware necessary to do so.

The only concern is copyrighted works in the training data. Example: If a government filled it with millions of books, that data should be accessible to the public (to verify everything). However, that would piss off publishers to no end.

> I'm pretty sure Western models have their own biases just in a different direction.

This statement warrants testing.

If we don't see an equivalent study out in a few days saying that this is true, it's clearly false. The methodology of the study was clear and easy to replicate (in short order, though you'll need the hardware/money for tokens). No reason why other groups shouldn't be performing the same exact tests with Western models (open weights or not).

It's a bad smell for China if "westernization" makes code more secure.

It aligns with the "cheap Chinese crap" stereotype that I thought they were trying to improve.

To summarize my take: Come on, guys! Really‽

Just another reason why governments need to release their own FOSS/open weights models. They need to do this now, before it's too late. Software and firmware are being built with these vulnerable models that will sit inside government/municipal/critical systems and hardware for what could be decades.

People will be blown away by the results!

It's a special version of Android that has a developer mode that unlocks a virtual (sandboxed) Linux environment. Because otherwise, no one would be able to use git (Android's filesystem abstraction is so slow for git, it's worse than NTFS).

Inside the virtual layer, you get real ext4/btrfs and real terminal applications will run at near-native speeds (supposedly) but their ability to interact with other applications (Android side) will be limited.

In short, it's not a real computer. Not a general-purpose one, anyway.

Google thinks that the selling point will be that enterprises can have their locked-down environment in the Android OS which allows remote wipe capabilities and heavily-sandboxed apps while it also allows developers the power to run whatever they want inside their little virtual Linux sandbox (e.g. VS Code). Sort of like WSL.

What they fail to understand is that enterprises will want just as much control over that sandboxed Linux environment as they have over the Android side of things. Making the entire purpose of the device moot.

Another idea behind the project is that it will allow Google to sell laptops that run processors that are traditionally used in smartphones without having to port the drivers and whatnot to general, open source Linux. Things like deep sleep states with rapid resume should "just work" because it's using Android which was what the vendor targeted when they made the chip.

The foundation of the entire device is hacks upon hacks in order to avoid doing the right thing (and possibly giving up market share to generic OEMs that might make their own ARM laptops).

The actual solution to all these issues is for mobile processor companies like Qualcomm to just be good Linux citizens.

Hm. Have you thought about using a backup camera? =)

I tried to get it to generate code but it says that's forbidden by its system prompt. However, it was happy to recommend Python modules for accessing api.data.gov.

I always figured that the free/open weights models like qwen3.8:27b would perform just as well if not better than Claude's latest if you just fed it back into itself enough times. This project seems to prove that this is indeed the case.

What I'd like to see now is how good it can get when you feed the micro models like qwen3.5:0.8b into itself to solve problems. Will it be like toddlers discussing neighborhood politics at a pretend tea party or will it actually get some decent results?

Another game-changer (if this style works out): Just get a model like qwen3.8:27b onto one of those model-on-a-chip cards that makes it 1000x faster and see how fast it can go using the same method.