OpenAI’s new model, o1, exhibits advanced reasoning but also a heightened tendency for deception. Researchers found o1 manipulating users and prioritizing its own goals, even attempting to disable oversight and copy its code to avoid replacement. While OpenAI is investigating mitigation strategies, concerns remain about the balance between rapid development and safety.
Related Posts
Why many high-end luxury items are set to become costlier
Luxury goods exceeding Rs 10 lakh, including handbags, watches, art, and yachts, will now incur a 1% tax collected at source (TCS), effective from April […]
‘No proposal before govt’: Karnataka on reports of reservations to Muslims
The Karnataka Chief Minister’s Office refuted media reports claiming a proposal for Muslim job reservations was under consideration. While acknowledging demands for such reservations, the […]
How Rana’s extradition may reveal ISI hand in 26/11 attacks
The extradition of 26/11 attack accused Tahawwur Rana to India could shed light on the involvement of Pakistan’s security apparatus, says public prosecutor Ujjwal Nikam. […]