Andy J. T. 🇵🇸🐀

Ich verstehe grad nicht den Spagat zwischen "Es ist schlecht wenn Personen mieten müssen" und "Es ist gut, wenn Personen zusatzlich zum Eigenheim etwas zu vermieten haben"

Praying for Michigan’s Jewish community and grateful for the quick work of first responders and law enforcement during today’s attack. While we wait for more information, one thing is clear: antisemitism anywhere must be confronted and stopped.

Israeli military drops charges against soldiers accused of sexually assaulting Palestinian detainee The Israeli military said it is dropping charges against five soldiers accused of sexually abusing a Palestinian detainee in an encounter partially caught on camera.

Nicht nur einzige Demokratie, sondern auch einziger Rechtsstaat im Nahen Osten.

apnews.com/article/isra...

They are not the models of 18 months ago, but they didn't improve too much from what we had 9 months ago.
Tooling around them got better, though.

Ich kritisiere die #zib2 ja oft, aber wenn man sich wie @beate.neos.eu zum Interview stellt, sollte man schon den Anstand haben, die Moderatorin ihre Frage stellen zu lassen, statt einfach drüber zu reden.
War jetzt eh kein sonderlich kritisches Interview, also völlig unnötig

Du hast von "keine Frau der Welt" und Beziehungen gesprochen.
Qualitative Studien sind nicht ungültig. Es gibt auch quantitative Studien zu dem Thema

Rethinking risk in adults’ engagement with sexual digital imagery

Das ist einfach nur falsch und zeigt ein seltsames Weltbild.
Mit dem Aufkommen von Smartphones ist das gegenseitige Zusenden sexualisierter Bilder in Beziehungen recht normal geworden.

Macht korrumpiert leider in den meisten Fällen

Von dir hätte ich mir schon erwartet, dass du weißt, dass "Ich kenne keine ..." kein Argument ist.

Du musst da ein bisschen aufpassen. Es ist besser wenn du Claude Skripte ausführen lässt, sonst halluziniert er wieder Sachen

Ernsthaft? Sollten "die Märkte" es nicht besser wissen?

Der Rassismus in den Medien ist untragbar. Wenn Palästinenser oder Iraner sterben, ist das irgendwie normal. Wenn US Amerikaner oder Israelis sterben, ist das eine Riesen Tragödie

Immer lustig wenn Regierungspolitiker etwas "fordern"

"possibility of rain is 70%" but it is even more complex. We are talking about long walls of text, the possibilities are practical infinite.

I choose "the" and took this as input to find the next word "man", but if I had chosen "a" I would choose "horse" as next word would just generate an impossible amount of permutations nobody could comprehend. It is like showing all simulations that were used for a weather forecast to arrive at

Not really, first statistics are not reliable, you can just show in which way they are unreliable. Second showing this makes it hard to read.
The statistical model is about what words are used. It chooses one word and takes this as input to find the next word. Showing all branches like

Only thing I can think of, is so people have to use it more to arrive at the answer they need, but there are already other psychological tricks in the way it answers to "hook" people.

But remember the point of companies making something is to sell it to customers. They are not only using it themselves. A more reliable LLM would sell better, so there would need to be some other strong incentive to make it intentionally unreliable. I just don't see it.

that they want to make it "likeable", give it the persona of a "helpful assistant", but remember there are also personas of Fiction in the data, so it might switch to some Antagonist (there are studies on that)
All that amounts to the clusterfuck we have right now

That is more of a business decision, if that would be intentional.
But from a engineering POV, it is just not possible to make a reliable LLM, it is highly complex statistics and statistics are never exact.
Throw in that it is statistics over "everything", including Fiction or outdated data.
Add

Why would anybody intentionally build unreliable Software?
I understand LLMs(at least better than most, including most other Software engineers) , I understand why they are unreliable.

Do you think being unreliable is intentional?

You still make the wrong assumption that you have to "influence" a LLM instead of working around errors or guardrails.
It is still just Software not a person

And what is the difference to LLMs regarding this?
It's also either an error, when it e.g. gives you a wrong answer or an institutional decision, when it tells you, it cannot do something.
And both are wrapped in a friendly user Interface

If they are working on preventing any future elections, why wouldn't they do that?

Plenty of technologies "disobey" you because they just don't work as intended.
Even the role playing, the simulation of a person isn't new.
Why does my E-banking greet me and then the transaction doesn't go through?
There is even an E banking App in Austria called George(it is not an AI)

You can say the same thing about pretty much every technology

You could argue that you have to "incentivize" your car with gas (if you still use that old technology) so it runs, but the fact is, that is just how that technology work

But the notion of it being like a slave is ridiculous. Any other technology is also not considered a slave even though they are supposed to "obey" you

I think it is just highly influenced by Scifi like Star Trek.
It sells better when you say it is like that even though we are far away from any technology like that.
There is some merit to it since it feels "natural" to use it and makes it accessible for non tech people.