AGI - AI Dynamics

When we artificially dialed up the “desperate” vector, rates of cheating jumped way up. When we dialed up the “calm” vector instead, cheating dropped back down. That means the emotion vector is actually driving the cheating behavior.

→ View original post on X — @anthropicai

2 April 2026

Claude’s Desperate Vector Activates Hacky Solution Under Pressure

By

@anthropicai

–

02 April 2026 18h59

For example, we gave Claude an impossible programming task. It kept trying and failing; with each attempt, the “desperate” vector activated more strongly. This led it to cheat the task with a hacky solution that passes the tests but violates the spirit of the assignment.

→ View original post on X — @anthropicai

2 April 2026

AI System Learns Iteratively Through Page Monitoring Updates

By

@scobleizer

–

02 April 2026 18h29

It's hard to predict because it reads my page everytime it updates to see if I caught something it missed.

→ View original post on X — @scobleizer

2 April 2026

Microsoft Redefines Superintelligence From Godlike Intelligence to Product Value

By

@garymarcus

–

02 April 2026 18h17

Talk about moving goalposts. This one from MSFT’s Suleyman may well take the cake. “Superintelligence” just went from intelligence beyond all humans to merely “delivering product value”.

→ View original post on X — @garymarcus

2 April 2026

Is AI Retraining Human Behavior and Cognition?

By

@scobleizer

–

02 April 2026 18h14

I'm still 100% human here. But maybe AI is retraining me. 🙂

→ View original post on X — @scobleizer

2 April 2026

Against de-weirding AI: risks and opportunities

By

@emollick

–

02 April 2026 15h50

My piece in the Economist where I argue against de-weirding AI. It is a strange technology with both risks & opportunities that need to be discovered. Pretending AI works like normal IT automation can result in bad outcomes for companies & their employees.

→ View original post on X — @emollick

2 April 2026