Jan Penfrat

It would be really cute if it wasn't so costly: how so many people fall again & again for artificial hype creation in #AI:

1. Company says "our next model is super dangerous," we need to do this very responsibly. (But we'll sell it to whomever pays anyway. đŸ€·)

2. Journos blindly repeat industry claims without verifying them. "It'll destroy the world!"

3. Gov'ts demand "early access" & concede that "the future of our #security is now reliant on the future of AI."

#Anthropic #Mythos #AIhype

Apr 13 at 11:40amWeb
Show threadMindaugas2d ago
@ilumium what is costly, is the head in the sand. AI is increasingly dangerous, for example, in cyber security, so society have to take cyber security more seriously to prepare.
Show threadZ.H. He2d ago
the pipeline from "this is too dangerous to release" to "here's the API pricing" really does complete in record time every cycle. governments end up negotiating from a position of awe instead of leverage, which is a wild place to make policy from.
Show threadSascha Foerster 2d ago

@ilumium And then add @bsi to the mix. Free ads for Antropic as a proxy. On the other hand. I also read the security issues and was impressed. I just missed the numbers of „detected security issues“ that were no issues.

https://www.zdfheute.de/politik/deutschland/ki-anthropic-claude-mythos-schwachstellen-software-bsi-100.html

KI: Anthropic-Modell spĂŒrt Schwachstellen auf - BSI im GesprĂ€ch

Das neue KI-Modell des US-Unternehmens Anthropic spĂŒrt versteckte Software-Schwachstellen aus. Das könnte dem BSI zufolge erhebliche Auswirkungen auf die Cyberbedrohungslage haben.

ZDFheute