feedback

UC Berkeley EECS colloquium by OpenAI's John Schulman on reinforcement learning from human feedback. He covers why RL helps with truthfulness, hallucination, and the limits of behavior cloning.

Adam Savage answers viewer questions on dealing with condescending colleagues, focusing on dismissive phrases such as 'why don't you just' and 'I would have'. He explains why such language undermines the person receiving it.

A two-minute demo of SmartKnob View, an open-source input knob built from a brushless gimbal motor, magnetic encoder and round LCD. Closed-loop torque control gives software-defined detents and end stops.

Computerphile interviews AI safety researcher Rob Miles about ChatGPT: how it was trained with human feedback, why it tends to produce confident nonsense, and where it may or may not be dangerous.

Sabine Hossenfelder explains climate sensitivity and why recent observations suggest climate models may have underestimated warming. She says this changed her view on how worried to be.

Veritasium examines which experts have real expertise, drawing on chess research and work by Klein and Kahneman. It explains that expertise comes from pattern recognition built through deliberate practice in valid, feedback-rich environments.

Allan Blomquist demonstrates the internal programming and debugging tools Tomorrow Corporation uses to make games, including live modification of a running game and time-travel debugging.

CityNerd explains why traffic congestion grows nonlinearly as more cars use a road network, and why adding road capacity tends not to fix it. It covers the consequences for drivers, bus riders, cyclists and pedestrians, and discusses alternatives such as true bus rapid transit.