This research was conducted as part of the Algoverse AI Research program Winter 2024 cohort.
Artificial general intelligence (AGI) may surpass human capabilities at many tasks in the years ahead. As AI systems become more capable and are further integrated into critical tasks, the risk from their misalignment grows. Previous literature has established that AI frequently exploits ill-defined or misspecified objectives—optimizing well on the objective function but often leading to unintended behavior, particularly under distribution shifts. Addressing these risks requires both theoretical advances in alignment techniques and empirical evaluations of AI behavior in dynamic environments. In this paper, we develop a Reinforcement Learning based training setup for AI agents in the video game Minecraft (a complex open world survival game),...
I think the point about the Internet enabling a "wordier style" due to lower printing costs actually gets things backwards.
What actually matters is the competition for attention. Consider that as the barrier to entry to publishing has dropped, the number of suppliers has exploded, while the number of consumers has grown much more slowly. That means there's far more supply competing for limited attention. This creates enormous selection pressure for ideas to be consumable, instantly engaging, and spreadable.
I think eloquence and beautiful language rightfull...
Seems that Google Translate was updated to use some version of Gemini in December: https://blog.google/products-and-platforms/products/search/gemini-capabilities-translation-upgrades/
Also, as of 2/10, when you go to https://translate.google.com/ and click Chinese (simplified) instead of Detect Language, there is a "Advanced" dropdown, it's labeled "Improved accuracy, built with Gemini."
Also not sure what I am doing wrong, but I don't seem to be able to replicate any of these experiments in either Auto-detect or specified language modes (as of 2/10, i... (read more)