Oh my, heise online (npub14kg…4yra) you should really know better than peddling #Anthropic's PR bullshit and selling it as "research."
No, #Claude does not "lie" or "cheat" and has no "consciousness" or "secret thoughts" just because the company makes those claims.
Is this supposed to be tech journalism or just copy-pasting industry press releases?
(Maybe something for [@mystery_ai_hype_theater](https://peertube.dair-institute.org/video-channels/mystery_ai_hype_theater ) to disect?)
Researchers discover Claude's secret thoughts – and how it lies and cheats
Researchers discover a consciousness-like area in Claude's neural network where it secretly thinks [--] and consciously manipulates tests.
https://www.heise.de/en/news/Researchers-discover-Claude-s-secret-thoughts-and-how-it-lies-and-cheats-11360440.html?wt_mc=sm.red.ho.mastodon.mastodon.md_beitraege.md_beitraege&utm_source=mastodon
#Anthropic #Claude #Forschung #KünstlicheIntelligenz #Sprachverarbeitung #Wissenschaft #news