@alexadeswift @sam

WhatsApp is using the Signal Encryption Protocol ... so signal is not less or more secure then whatsapp as long as you don't compile the client yourself and actually check the keys.

You could say that signal is not doing client side keyword detection or whatever but this is something signal can also just do, so you have to trust signal what is a bad idea since it is a US corpo like OpenAI is one.

https://signal.org/blog/whatsapp-complete/

https://faq.whatsapp.com/820124435853543#:~:text=As%20for%20WhatsApp%20security%2C%20know%20that%20every%20WhatsApp%20message%20is%20protected%20by%20the%20same%20Signal%20encryption%20protocol%20that%20secures%20messages%20before%20they%20leave%20your%20device.

@lea

If you have an alternative that is decentralized and more usable then Matrix, speak.

If not: you are not helping.

Please, for the love of resilience and safety, do not promote or expect that Signal, Discord, or WhatsApp is safe or stays safe. Rather, use decentralized and open-source projects like Matrix or XMPP.

The single point of failure messenger Signal already threatened that they will turn off their services in the EU if law requires them to do so and thus cannot be seen anymore as reliable.

And more importantly, if you think that Matrix needs more time and development, it is in your hands to help create an alternative to big centralized US chat platforms, and all you need to do is to donate to the underfunded Matrix Foundation that has to compete with multi-billion $ US corpos.

https://matrix.org/support/

If just 1% of the people paying Discord for the privilege of sending 4,000-character messages donated to the Matrix Foundation instead, we would have a secure, resilient, and decentralized messaging platform in no time.
One with no single point of failure that lawmakers cannot easily destroy, because no single entity runs it.

Messaging platforms are one of the pillars of our democracies and communities, it is worth fighting for a good one.

#matrix #signal #discord #privacy #surveillance

https://en.wikipedia.org/wiki/HIPPI

Oh shit HIPPI is beeing removed from Linux 7.0 ... how should German ministry infrastructure handle that??

The amount of people that trust random accounts more then LLMs is concerning.

Both have no businesses to be taken as face value.

woke means eyes open

The EU is collecting such massive gigantic Ls the last few months one has to wonder what happened to these people in the EU parliament.

#EU

Wenn das durchkommt ist die EU Geschichte and hat es verdient zu sterben.

Absoluter Wahnsinn

Mein Vertrauen gegenüber der EU wurde bereits massiv zerstört durch die Chat Kontrolle aber das hier wäre der Sargnagel.

Da kann man dann auch nichts mehr retten.

https://www.derstandard.at/story/3000000309864/eu-gesteht-usa-direkten-datenzugriff-zu-automatisierte-verfahren-moeglich

#eu #usa #datenschutz #kolonialismus

Now we are cooking!

https://schema.org/Recipe

Complete mind boggled.

I mean it is a 8B model but this is the first generation of this tech ... like 8B is not at all bad for that!

Whoooaaa they have put a LLM model on an ASIC and now you can try it out and get 15000 tokens per second .... what the fuck

https://chatjimmy.ai

Please for the love of you and your kids, get rid of discord.

> "53 megabytes of unprotected source maps on a FedRAMP government endpoint, exposing the entire codebase of a platform that files Suspicious Activity Reports with FinCEN, compares your selfie to watchlist photos using facial recognition, screens you against 14 categories of adverse media from terrorism to espionage, and tags reports with codenames from active intelligence programs.

https://www.pcgamer.com/software/security/security-researchers-claim-persona-the-provider-behind-discords-uk-age-verification-experiment-performs-269-individual-verification-checks-on-user-data-including-those-for-terrorism-and-espionage/

#discord

@manawyrm

Okay so i digged a little bit deeper and on the Vodafone site:

"Knapp 24 Millionen Haushalte im deutschen Vodafone-Netz haben seit 2023 einen Gigabit-Anschluss über das Kabelnetz. Ein landesweiter Ausbau von DOCSIS 3.1 ist somit erfolgt."
("Since 2023, nearly 24 million households on the German Vodafone network have had access to a gigabit connection via the cable network. This signifies that a nationwide rollout of DOCSIS 3.1 has been completed.")

"Der aktuelle Standard DOCSIS 3.1 ermöglicht Downstream-Geschwindigkeiten von bis zu einem Gigabit pro Sekunde. Die Upstream-Geschwindigkeit beträgt 240 Megabit pro Sekunde."
("The current DOCSIS 3.1 standard enables downstream speeds of up to one gigabit per second. The upstream speed is 240 megabits per second.")

https://www.vodafone.de/business/blog/docsis-94712/

So it IS possible to have more the 50Mbit upload with DOCSIS 3.1, which is widely adopted and uses OFDMA.

But i can only get the 50Mbit/s upload internet contracts form them.

@pojntfx

It is kind of like a religion :D

they keep the stuff (predictions) they like and void the once they do not like because they are unrealistic XD

@manawyrm

Interesting thanks for the explainer!

But the decision to use the frequency splitters at X MHz was made by them right?

So it was a decision they made, probably to prioritize TV, and against internet speeds and upload?

Or is this a law standardization thingy and politicians made this prioritization decision?

@manawyrm

I think this is wrong.

See:

https://www.commscope.com/solutions/broadband-networks/docsis-3.1-deployment

@pojntfx

I know you know but it made fun to write that:

Predictions in this field are always absolutely wrong. Because for predictions you need to know the current complexity and what complexity will evolve from this complexity, but once cannot know the complexity of AI as the complexity is just now researched.

The few "predictions" of the future of the AI field are just lucky and completely meaningless because one cannot distinguish them from the "predictions" that fail.

We can predict chip production quite reliable, but not tomorrow, a year from now or never a calculation model is found that makes LLMs 900 times more efficient.

Anyone who is predicting stuff like " i predict that in 2 years": "They are achieving a year’s worth of algorithmic progress every week and will therefore soon be up against the limits of the Agent-4 paradigm." ... is just bullshiting hard.

Maybe one can predict social aspects that will come with AI as social aspects are way way slower then the AI research field and is also not that affected by research results how we would imagine it is.

And we can maybe take meta predictions like "if this happens it is likely that this happens"

It should be illegal to provide 1Gbit download but only 0.05Gbit Upload...

Like thanks ... the upload is juuuuust enough to have 1Gbit download...

I am now standing in the kitchen for a substantial amount of my time…. I did not expect my life to take that turn.