We look at the mechanics of AI-assisted music streaming fraud; researchers develop a new algorithm for LLM code generation that leverages inference time search over high-level plans
Mini-Update #47: First International AI Safety Treaty and the WavTokenizer Codec
The UK signs on to the first international AI safety treaty and a novel audio codec model allows for high-quality information transmission in fewer tokens.
We look at the numerous ongoing lawsuits against AI giants; researchers develop a method to address the risk of pre-trained models being repurposed for unethical or harmful tasks.
Mini-Update #46: OpenAI-Condé Nast Partnership and Flexible Perception
Major publishing company Condé Nast grants OpenAI the rights to train models on their work, and Google Deepmind researchers integrate visual memory retention to augment model flexibility.
