foo12bar

I'm not talking about outcome, but of sensing. I also never wrote it was a suitable method. When did that become a requirement?

I kist thought it was interesting that there are receptors that can detect wetness, just not on the skin. It was a joke pointing that out.

Fuck.

Secondhand smoke does not feel the same as drowning.

Im a vacuum you would not be choking.

Trying to beathing in a vacuum would not feel the same as breathing in a liquid.

Your lungs have sensors that can detect being coated by a liquid which invokes coughing. In that sense, you can detect wetness.

Oh, so if you stick your head in a liquid and breathe you wouldn't know?

Stick your head in it and breathe. You'll know.

I see what your saying. Yes, I suppose it's possible, but you couldn't test it to see if it worked until both entire models were trained and released. You'd also need a lot of poisoned data to get it to work, there'd be evidence of the poisoned data left for everyone to see, and you can't screw up the models performance too much, since it would be benchmarked.

Simply downloading a model from huggingface, adding a feature, such as abliteration, and uploading back doesn't face such hurdles.

But anyone can decompile it, now, too.

I don't think so. Both models would have to get into a blackhat's hands, who would then poison them both so if a keyword was present, the model giving the response would insert an exploit, and the sentry model that's supposed to check would ignore it.

So if you didn't have a checking model, you only need to download one model that the blackhat corrupted to fall victim. If you have a checking model, you need to download two models. So it goes from x/n to a x/(n*2) chance of being compromised. Furthermore if you downloaded a clean checking model, you would be alerted if the model giving the responses did something cheeky. Then you could warn others, so that makes it a lot harder for a blackhat to create poisoned models that would go unnoticed.

Well you could use a model created from a completely unrelated entity. It would be difficult for an adversary to poison both in a exactly complementary way.

Not to mention, Big Bathtub.

It would be interesting if the model could be trained to perform the hack based on its internal belief about the current date and time, location, user, environment, etc. It would be far more stealthy that way.

Why not have a second model that reads every response looking for something nefarious?

Maybe ask a separate model to look for indications against every model response you receive whether it is nefarious would work better.

If the poison was hidden in a secret prompt, I would agree with you. But this was trained into the weights as a reflex (using LoRA but that shouldn't make a difference). It's like asking a model why it thinks strawberry has 3 r's, it just doesn't know.

How much training or education is needed? I mean, frankly, a person can just go watch a 20 min YouTube video and learn all they need about how to restrict calories to lose any amount of weight they like.

You don't think people are capable of this?

Can it compile itself?

Why is this sarcastic? Isn't this a plausible possibility?

You should make it so you can coerce the moon to fall out of orbit, crash into the planet and destroy half of it.

If we're talking about running it locally, what about passing a partial response as part of the input?

Prompt part: "What is better, toast or bread?"

Incomplete answer part: "The answer to this question is "

and then have the LLM finish the answer. I did this with subtitle translation using llama.cpp (with Python) and had great success. Just past 5 already translated subtitles as the incomplete answer, and the LLM infallibly just continues to translate. No markdown, and usually no talkback if the subtitles contain nasty subjects like bioweapons or nuclear stuff. It just works.

Iran has continued to fire at tankers exiting the Strait of Hormuz under U.S. Navy escort, and apparently the U.S. can't afford to do anything about it.

The Bab el-Mandeb Strait, which is an alternative route used by the Saudis, is being closed by the Houthis. The Houthis are a close partner of Iran.

And the East-West pipeline, which was another alternative route owned by the Saudis, was blown up. They went for the pumping stations, so repairing it will take at least a month, and there is no way to repair it without it possibly being hit again anyway.

The oil infrastructure attacks by Ukraine are mainly targeting refineries, which would normally lower oil prices, since crude oil is an input to these refineries. These attacks are increasing diesel prices, though.

Iran is trying to break the world economy so the US packs up and leaves. Israel wants to keep the US there because the US is fighting one of their strongest enemies. Because Israel "is the US's greatest ally" (as many politicians have proclaimed over the years), they have a lot of sway as to what the US does. Not to mention, they seem to do whatever they can to derail any peace process.

Iran knows this, and therefore wants to make sure the US experiences enough pain to never come back and try to fight them again.

So, until the US is willing to stop, which rests heavily on when Israel is willing to stop, oil prices will remain high. (Or Iran folds, which they aren't going to do, they've been preparing for years and have seen how the US treated Venezuela - they stole all their oil, and left the regime intact)

Also, we've seen nothing yet, as the soft storage (the amount of oil that's normally floating through rhe global network) and the SPR's are all running dry. I would guess $150 oil in about a month.

Some of the Resident Evil series is in VR, and is also very good.

https://github.com/praydog/REFramework

But this is more like an LLM holding "tea" and "no tea" at the same time, and you're trying to get it to drop one or the other and it refuses.

So, the developer updates both the free version and the paid version with the latest features (the only difference is the free version is sans GUI), and continues to serve the older free keyboard only 0.47 version on his website, and I'm supposed to be upset by this? Because he dropped the keyboard UI feature?

DFHack already added back the keyboard cursor, anyway: https://docs.dfhack.org/en/stable/docs/tools/toggle-kbd-curs...

Is the base really made of cement?

The free version is still being updated. Both the free version and the paid version are on v53.16:

https://www.bay12games.com/dwarves/ https://steamdb.info/app/975370/history/

I misspelled held. Got me though. Too bad you didn't want to actually engage in dialog. You sure seem to know a lot.

> The carry trade does not hide public US debt. The US Treasury is concerned about Japan because it’s like the first or second largest holder of Treasuries in the world.

I never wrote it hides the debt, it borrows JPY at a low interest rate and then buys U.S. treasuries at a higher interest rate, making money on the difference.

> Because that was the legal facility open to them. They used a foreign currency reserve that they had the ability to trade.

Again, why not use U.S. Dollars? And again, why not at least give a heads up to the Europeans? You are not directly addressing my questions.

> Yes. Go re-read my comment: Japanese politicians do not want rates to go up. It’s like a cornerstone of current LDP policy.

No, you re-read my comment that you just slapped into your response and then completely ignored. WHY DID THE US BUY YEN AT ALL? You are ignoring my point and just restating your conclusion.

The carry trade hides a lot of US debt, too. Why would the US only care about the US debt held sovereignly by Japan and not by the banks and hedgefunds executing the carry trade?

If the US is not trying to defend the US dollar, why'd they use Euros and not USD to prop up the Yen, and also why surprise everybody doing so?

And, if it is better for the US if Japan normalized rates, then why did the US buy yen at all? According to you, just telling Japan to raise rates would be better, right?