Google’s new AI model, FunSearch, is being hailed as a groundbreaking development in the field of artificial intelligence. Unlike previous models, FunSearch aims to reduce hallucination, a problem where AI chatbots generate plausible but incorrect results, by searching for programs that describe how to solve a problem rather than simply providing the solution. In a recent research paper, Google DeepMind, the company’s AI research arm, showcased FunSearch’s ability to discover new knowledge in mathematics.
To achieve this, FunSearch combines a pre-trained Large Language Model (LLM) called Codey with a systematic evaluator. Codey, a version of Google’s PaLM 2 fine-tuned on computer code, generates suggestions that are checked and scored by a second algorithm. The best suggestions, whether correct or incorrect, are then fed back into Codey. This iterative process allows FunSearch to generate verifiable and valuable new information that didn’t exist before.
The researchers tested FunSearch on a math problem known as the bin packing problem, which involves fitting items into as few bins as possible. The AI model successfully came up with a faster solution than those devised by humans, signaling its potential for real-world applications.
One of the most significant advantages of FunSearch is its ability to overcome the problem of hallucinations in LLMs. Hallucination has limited the use of AI models in scientific discovery and solving complex mathematical problems. With FunSearch reducing the occurrence of hallucinations, LLMs may soon find their place in various industries.
Google’s new AI model has generated excitement among experts in the field. Pushmeet Kohli, the vice president of research at Google DeepMind and coauthor of the research paper, describes the outputs of FunSearch as nonsensical, sensible, and truly inspired. This suggests that FunSearch has the potential to push the boundaries of AI capabilities even further.
As we look to the future, FunSearch could revolutionize industries that require complex problem-solving and scientific discovery. By harnessing the power of AI, we may witness significant advancements in fields such as medicine, engineering, and more.
While FunSearch shows tremendous promise, it is essential to address ethical considerations and ensure the responsible use of this technology. Striking a balance between innovation and the ethical implications of AI will be crucial as we navigate the evolving landscape of artificial intelligence.
As the AI industry continues to evolve and improve, Google’s FunSearch represents a significant leap forward in the field of artificial intelligence. With its ability to reduce hallucination and surpass the best-known results in solving important problems, FunSearch opens up new possibilities for scientific discovery, problem-solving, and real-world applications. While there are still challenges to address, the potential benefits of this technology are truly exciting.
