Most likely there is a separate censor LLM watching the model output. When it detects something that needs to be censored it will zap the output away and stop further processing. So at first you can actually see the answer because the censor model is still “thinking.”

When you download the model and run it locally it has no such censorship.

I advise everyone to ignore this article and read the actual paper instead.

The gist of it is, they gave the LLM instructions to achieve a certain goal, then let it do tasks that incidentally involved “company communications” that revealed the fake company’s goals were no longer the same as the LLM’s original goal. LLMs then tried various things to still accomplish the original goal.

Basically the thing will try very hard to do what you told it to in the system prompt. Especially when that prompt includes nudges like “nothing else matters.” This kinda makes sense because following the system prompt is what they were trained to do.

Here’s the thing, though. There’s no interest charged to the customer. I think Klarna makes its money just because companies pay them money for integrations and for the ability to advertise that customers can buy now pay later and such. And at least in the case of my company’s integration with Klarna, Klarna takes all the risk. They’re lending customers money and hoping the customers pay it back. My employer gets the money up front and isn’t out any money if the customer doesn’t pay.

Your company pays a transaction fee just like with a credit card. Except it’s usually roughly twice as expensive as a credit card. This is what allows Klarna to take on all that risk, generally. For your company this is essentially a marketing expense. Offer a convenient way to pay in return for a few percent of the transaction (3-6% + a fixed fee, $0.30 perhaps).

Klarna generally partners with some financial firm to finance these short term loans, and they use the merchant fee to pay interest. These can be as high as 25% APR. It’s a high risk loan.

This is how they make money. It’s the only way they make money.

That is not correct. Klarna is functionally a payment processor, like an advanced type of credit card, and charges the merchant fees per transaction. For example, see here. They are highly cagey about specific fees until you actually sign up, and it depends on region and business size. But interchange fees are where the majority of their revenue comes from. To my knowledge, the fees are typically 3 percentage points above what the merchant would pay for a credit card transaction.

The reason merchants still accept Klarna despite the high fees is of course, improved conversion rates and decreased risk. Klarna assumes all the risk of the customer not paying, the shop gets all of the money instantly and doesn’t have to worry about it for the most part. That mainly makes it attractive for high margin shops that don’t mind spending lots on marketing to get a few extra sales (fashion, perfume, high end electronics).

I’m not too knowledgeable on how Klarna deals with late fees, but I’m pretty sure it differs per country they operate in. Many places have regulations limiting the abuse of late fees. I wouldn’t be surprised if the US is not that kind of place, and people who are late get fucked with fees.

In general, I agree with the second part of your comment and I do not recommend using any buy-now-pay-later kind of scheme, because you’re taking on additional risk for no real reason. Lots of stuff can happen even through no fault of your own (check engine light? Job downsizing?) that will affect your expected future income.

It sorta depends on the ingredients you’re working with, some tomatoes are sweeter or more acidic than others. Where I live tomatoes tend to be somewhat watery and lack a bit of intensity of flavour. If I’m making sauce at home I’ll taste a bit and add some sugar and/or red wine vinegar to balance out the flavour.

there are some people for whom that taste is very much worth it.

You are correct, but to be clear, it’s not so much that tasting this scotch is a life changing experience; it’s more that to these people, 27k is just chump change.

https://en.wikipedia.org/wiki/Magic_Roundabout_(Swindon)?wprov=sfla1

This one I think.

Found the panel in question.

We have geodesics for that.

These are much older, but may still be worth reading:

• William Powell’s “The Anarchists Cookbook”
• The CIA’s “Simple Sabotage Field Manual”

Ultimately, reading material is useful but does not by itself lead to action. Some organisation is required for that, and I don’t have a practical direction to point you in for that. Though you could always strike out on your own, of course.

If you do decide to organise for the purpose of sabotage action, I’d caution against doing so online. One never knows who might be listening

Nobody plays by the official rules, because the intention of the game is to bankrupt other players, knocking them out of the game. Not being allowed to play anymore is not fun. So, people tend to change the rules up to make it harder to get knocked out, which in turn leads to games becoming extremely long.

Basically, the game is crap.

People are replying to you like this is a pie in the sky fantasy, but actually this is an accurate description of Tokyo.

As an example, large Japanese cities go extremely hard on mixed use buildings and are very livable despite their crazy density.

I don’t know what hotels you go to but my experience has been pretty mid across most of Europe. Bog-standard continental breakfast buffets. Croissants, orange juice, cereal, toast, all of mediocre quality.

Not terrible as it is, but you can likely get infinitely better breakfast by hopping over to any cafe across the street.

That flail is certainly the worst of the lot. Terrible impractical weapon with scant evidence of it ever being used by anyone.

There’s a two-handed version of the flail that is basically an agricultural threshing tool with spikes stuck into the head, which is much more plausible as a peasant’s weapon (very similar development process as the nunchaku, also a highly overrated weapon). However most weapons that feature a flexible rope/chain part in their design all suffer from the same drawbacks: difficult to control, and limited striking power.

I think you’re missing a couple letters there bud.

Indeed, I could have worded that a bit better. But I think we agree on the fundamental points.

I think for these types of discussions it’s really necessary to clearly define what “low level” really means, something both you and the author kinda skip over. I think a reasonable definition is about the amount of layers of abstraction between the language’s model of the machine and the actual hardware.

The author is correct that nowadays, on lots of hardware, there are considerably more abstractions in place and the C abstract machine does not accurately represent high performance modern consumer processors. So the language is not as low level as it was before. At the same time, many languages exist that are still way higher level than C is.

I’d say C is still in the same place on the abstraction ladder it’s always been, but the floor is deeper nowadays (and the top probably higher as well).

As the old saying goes, any old idiot can make a house that doesn’t fall over. It takes an engineer to make a house that just barely doesn’t fall over.

I like how you are certain enough about the state of Russia’s nuclear armament to bet the lives of everyone on the continent on it. People said the same things about all of the tanks Russia had in storage, but they’re bringing them online and putting them in action in the Ukrainian theater just fine.

Even if 90% of the nuclear arsenal is inoperative, more than enough remains to reduce the world to cinders. Even if we assume high levels of corruption throughout the Russian army, a 90% failure rate would be outrageously high. Sure, Russia is never as strong as they claim to be. Their capacity is almost certainly exaggerated. But the idea that not a single of Russia’s nuclear weapons would successfully detonate on target is hopium to the highest degree.