New AI Training Technique Is Drastically Faster, Says Google

Google's DeepMind researchers have introduced a new AI training technique called JEST, which accelerates the training process by up to 13 times and increases efficiency by 10 times. This approach aims to reduce the computational resources and time required for AI training, potentially leading to lower energy demands. The AI industry is known for its high energy consumption, with large-scale AI systems like ChatGPT demanding significant processing power and water for cooling. The JEST method optimizes data selection for training, reducing the number of iterations and computational power needed, which could ultimately lower overall energy consumption. By selecting complementary batches of data to enhance the AI model's learnability, JEST proves to be more effective than traditional methods. This technique, based on multimodal contrastive learning, identifies dependencies between data points and focuses on high-quality datasets, improving training efficiency. If implemented at scale, JEST could enable AI trainers to create more powerful models with fewer resources, potentially mitigating the environmental impact of AI development.