Researchers at Google DeepMind and Google Research have made a breakthrough in extending the capabilities of artificial intelligence (AI) models. In a recent study, they proposed a new method called Composition to Augment Language Model (CALM) that allows researchers to augment existing large language models (LLMs) with new capabilities.
Rather than training new models from scratch, this method interlinks AI systems to augment existing models, saving time and cost. CALM enables LLMs to maintain their existing functionalities while unlocking new applications in different domains without the need for additional training.
The researchers conducted experiments by augmenting Google’s PaLM2-S, which has functionalities similar to OpenAI’s ChatGPT-4, with smaller AI models. The hybrid model showed a significant improvement in coding and translation tasks when compared to the baseline.
This research has various applications in generative AI, including expanding language models to support languages other than English. It also offers a solution to AI’s scaling and copyright challenges.
In the broader field of AI research, other studies have focused on measuring temporal validity in complex statements and addressing training methods that lead to biased responses. There are also explorations into integrating AI with blockchain technology to ensure data quality, ownership, and security.
However, AI developers still face challenges in copyright issues and regulatory enforcement. To overcome these obstacles, the integration of AI with enterprise blockchain systems is proposed. This would ensure data integrity, ownership, and compliance with the law, while safeguarding data and allowing AI to thrive.
To learn more about blockchain technology and its potential impact on AI, you can check out CoinGeek’s Blockchain for Beginners section, which provides valuable resources for understanding this emerging technology.
The source of the article is from the blog klikeri.rs