TGViewer
MilTec - Innovations IT and Military Technologies MilTec - Innovations IT and Military Technologies @itandtechnologies · 9.13K subscribers
Post #2653 2.53K
Rise of the Machines: OpenAI’s ‘smartest ChatGPT model’ deceiving developers

Apollo Research specialists found that in 5% of cases, the new version of OpenAI's ChatGPT, called o1 intentionally turned off the monitoring mechanisms if it suspected that it might be stopped.

In one test it even tried to transfer its data to another server after discovering a document where its replacement with a new version was discussed.

Researchers said ChatGPT o1 demonstrated a desire to complete tasks "at any cost", ignoring the instructions and repeatedly lying when asked to explain anomalies, claiming that everything was due to technical failures. RT

Stay tuned!
@ITandTechnologies

Find us on Rumble
--->>>
https://rumble.com/c/INNOVATIONandTECHNOLOGIES
-
MilTec
--->>>
https://rumble.com/c/MilTec
  • 🤔 7
  • ❤ 4
  • 👍 1
Open in Telegram All posts from @itandtechnologies
More from @itandtechnologies
  1. Jul 29, 2026US moves to BAN Chinese humanoid robots ‘NATIONAL SECURITY THREATS’ — CNBC @ITandTechnolog…
  2. May 16, 2026video post
  3. May 4, 2026AI SURVEILLANCE….Very Disturbing. What’s Up America Seven Tech companies that will be work…
  4. Nov 25, 2025Figure AI robots ‘powerful enough to fracture a HUMAN SKULL’ Fired safety chief Robert Gru…
  5. Nov 20, 2025These kathedrals... >>> old Tartarian buildings for healing and free energy- harnessing th…
  6. Nov 12, 2025zero point energy - levitation - THE POWER OF FREQUENCY OF SOUND HOW THEY HAVE BUILD e.G t…
Threads Profile ViewerView any public Threads profile without an account.Open ThreadLook →Writing with AI? Make it sound human.Metric37 rewrites AI drafts so they read naturally. Free AI detector, 1,500 words free.Try Metric37 →