All the articles from Asterisk’s AI special:
Matt Clancy and Tamay Besiroglu debate whether AI will lead to explosive economic growth
Jeffrey Ding argues that we overestimate China’s AI capabilities
Beth Barnes explans ARC Evals’ work on GPT-4
Scott Alexander looks into the field of AI forecasting since 2016
Sarah Constantin on what happens to AI progress if Moore’s Law ends
Jonathan Mann argues that LLMs won’t result in tech job losses by 2025
Carl Robichaud onthe history of nuclear non-proliferation and takeaways for AI governance
Kelsey Piper offers an overview and history of the field of AI Safety
Rob Long on the problem of how we should think about AI cognition
Avital Balwit gives an overview of compute governance
Michael Gordin on how American scientists underestimated how long it would take the Soviets to create an atomic bomb
Jamie Wahl’s short story Emotional Intelligence Amplification
The Editors on why they worry about AI risk
Bonus: the magazine offers a visualisation of LLM progress in the last 6 years
And a reminder that if it’s your thing you can buy a print subscription to the magazine here, or you cansubscribe to the newsletter by clicking the menu bar on the homepage
Does anyone else from the UK get a ‘unsupported protocol’ error from the Asterix site? I do, but it doesn’t trigger if I use a VPN.
The link to a paper of diffusion capacity in the article about China by Jeffrey Ding seems to be broken.
Lots of great stuff here. Strongly recommend following Asterisk.