vintermann

Eh, I say give it a try. Worst case we don't get to space but have a good time.

I tried Argon for a large historical HTR transcription project recently, since I heard it was really good at this. It was really good, but like all top US models I've used it was

* Definitively very good at domain specific knowledge. It can read sloppy 19th century danish gothic handwriting with faded ink just fine. Not perfectly, but probably better than I could have done with a day of struggle for every page.

* Expensive.

* Bad at actually doing what I said. I had told it in the prompt to be careful not to miss text, but I hadn't explicitly told it that some images may have text in double-column layout. So on those images - over half of my 300-page document - it produced an excellent transcription of the first column, and then considered itself done. Want to try again with a more careful prompt? That'll be another $20.

Around here, that would be pretty much hibernation.

Why is it so hard to set our schedules instead of our clocks? Can't we just agree that school starts an hour later in winter, for instance?

Elseh I'll grant, but it feels a lot less natively English, and you can't count on an American clerk or census taker to care about accurate transcription. It may have been closer to "that's not a name in English. Pick either Elsie or Elsa".

And yeah, Elsa was common too.

A big part of the picture was that they'd gamified the heck out of it. Community moderation can work, somewhat, but I've never seen gamified moderation work. Gamification is the big killer of collaborative communities.

Regular people may like it... But judging from real estate prices around where I live, the richer people are, the more they seem to want to live in modernist birdhouse style houses.

There were plenty of long English keywords in early computing: continue, dimension, equivalence, frequency, procedure etc. It seems it was mostly the Lisp folks who strictly shunned long keywords.

I think most Elsies were originally Else, or had a mother or grandmother named it. In Germanic languages the final E is not silent as in English, so you have to spell it Elsie for English people to pronounce it approximately right when they read it off the page.

A common problem in systems where two branches of government have equal claim to popular mandate.

What, and invite an off-site brigade?

The only way to win is not to play.

Do any of the locked down consoles or phones use a vulnerable UEFI implementation by any chance?

I've never heard of Deepseek refusing this kind of work though.

Such laws always work better for one side than the other, it's always selective about which facts get checked and which false news get suppressed. I don't know about how trusted that media network is in Brazil, but from all we've seen from other countries, authoritative fact checking from old information monopolies just fan the flames. It wouldn't surprise me if conceding to the request was the best they could do in the situation.

This is not surprising, since the model is just given one piece of information and it's the piece of information the CNN has been trained to make its whole decision on.

I'm not a dermatologist, but I imagine they don't make their decisions on one image alone (and the decision probably isn't "is this cancer" but "should we take a biopsy"). The LLM can take into account additional information such as the patient's age, occupational exposure and medical history, the CNN can not.

But then again, maybe it's useful to have a tool which you know does not use any information besides the image to make the call.

The appeal is that you can incorporate more context in a decision problem without knowing in advance what that context might look like.

Take the problem of determining if two people in a genealogy database represent the same person. It's a decision problem. There will be some data in structured formats you can count on, but which may be missing, or may even be wrong. There will be lots of fuzzy data (born in Newtown yes, but which Newtown?) and unstructured data (notes, etc.)

Now, do you want to try to assemble a dataset to train your own classifier on that, or do you want to throw a general purpose LLM-rooted decision model on it?

Not raising taxes specifically no, but yes, US presidents have asked the supreme court for permission to do things the legislative told them not to, and gotten it. One Trump famously got, was the right to fire almost all federal employees at will, including members of boards established to be bipartisan for political reasons.

But really, the US has had the same problem you see in Brazil for so long, the president usually doesn't even have to ask the court, because the legislative was already sidelined decades ago. That is the case for the rather important issue of going to war, which of course Trump has also done against legislative approval.

Why is it always Russia?

This is into "not even wrong" territory. Brazil has a fairly close relationship with Russia, they're a fellow founding BRICS member - and Lula was president when it was formed. What would Putin get from having US (Trump) ally Bolsonaro Jr. over him?

And while it's great that the murder-happy fascist didn't win, it's not exactly a great win for democracy. It's just stopping the bleeding. Besides the obvious problem of being allied with Russia and China, Brazil has a worrying "anti fake news" law. Like elsewhere, it seems the slightly-better guys aren't eager to roll back power grabs that now benefit themselves.

That's right, it's the Venezuela model. It's also the Hungary model, the US model, the South Korea model, the Peru model, the Mongolia model, to name a few. All in recent years have had constitutional crises as a result of conflicts between a popularly elected president and a popularly elected legislative.

Go a few more years back, and it's hard to say the Brazilian right-wing aren't reaping what they sowed wrt. supreme court/judicial branch shenanigans.

But that doesn't make you wrong. All presidential republics are like this when push comes to shove. There are two branches of government with equal claim to popular legitimacy, therefore every presidential republic in the world reels from one constitutional crisis to another like a drunken sailor.

I love true scotsmen

Category error.

At least it wasn't the lighthouse keeper Elias Thorne.

Should be easy enough to verify the central claim that you need to be able to read the password file through some other means to use this exploit.

An obvious benefit to plan mode is that your model tells you what it's planning to do before it does it.

Obviously some people get addiction problems, and obviously some people are more vulnerable to develop them than others, but I can understand the reluctance to ascribe that to their personality.

Why can't it be the situation? Even to call it the internal situation if you find no obvious external reasons, is probably healthier than to ascribe to the essence of who you are.

Remember David Nutt suggesting his own heavily branded benzos as a "replacement" for alcohol?

He was technically absolutely right that alcohol was worse than most illegal drugs (what he was unfairly fired for saying), but man, he had major white coat blinders. Terribly naive and oblivious to what science besides neuroscience could say about drug use. Good thing his "synthetic alcohol" benzos didn't catch on.

That works. But make sure you don't say it outright because that might make people stop and think about it for a moment.

Imagine what they could have done without the chip embargo.

Perhaps this is evidence there's not as much Fu Manchu moustache twirling scheming as it seemed?

You don't have to give it up entirely, you could grade on the interpretation they chose. Personally I'm a fan of models which seek clarification on critical ambiguities rather than barging ahead "autonomously" though.

But wow, they found a lot of issues with the benchmark.