The landscape of technology is witnessing a groundbreaking innovation – the rise of self-evolving artificial intelligence (AI). As per a study released on December 15, a group of researchers from the Massachusetts Institute of Technology and various University of California campuses have demonstrated that AI models are capable of generating AI subsystems independently, as reported by Fox News.
These larger AI models, akin to those utilized by ChatGPT, have the potential to generate smaller, more specialized AI applications. These applications have practical uses in our daily lives, such as enhancing hearing aids, overseeing oil pipelines, and tracking endangered species.
Self-evolving AI will be a gamechanger
Yang Song, the CEO of AI tech firm Aizip, explained to Fox News, “We’re currently employing extensive neural networks to generate smaller ones – it’s akin to an elder sibling aiding their younger sibling to improve. This marks the initial step towards a broader initiative in self-evolving AI. It’s the first demonstration of how AI models can self-replicate.”
Yubei Chen, a professor at UC Davis and co-founder of Aizip, expressed that their most astonishing discovery was the capability to employ a parent algorithm within one model to automatically reconstruct smaller models. Chen believes that in the future, collaboration between large and small AI models will lead to the creation of a comprehensive intelligent ecosystem.
Chen further added, “Our technology marks a significant breakthrough as we’ve developed a fully automated playback algorithm for the first time. This enables the development of an AI model without human intervention in the process. We’ve also showcased the first proof of concept that a specific type of model can be automatically designed from data generation to model deployment and testing, all without human intervention.”