Things I read last week… (WB 28 Nov 2022)
Physicists create a wormhole using a quantum computer.
OpenAI releases ChatGPT, optimising language models for dialogue.
DeepMind’s DeepNash learns to play Stratego from scratch by combining game theory and model-free deep RL.
Disney’s Face Re-aging Network (FRAN) can de-age actors in video.
Sequoia on generative AI.
NVIDIA’s New AI: Generating 3D Models! [YouTube]
“We hypothesized that people would exhibit a reticence bias, the incorrect belief that they will be more likable if they speak less than half the time in a conversation with a stranger, as well as halo ignorance, the belief that their speaking time should depend on their goal (e.g., to be liked vs. to be found interesting), when in fact, perceivers form global impressions of each other. … Contrary to people’s forecasts, they were more likable the more they spoke, and their partners formed global rather than differentiated impressions.”
– Mistaken beliefs about how much to talk in conversations.
All of the worlds money in one visualisation.
The history of sandpaper. [YouTube]
Collection of 700+ 88x31 web buttons from the 1990's and 2000's.