Google, which is fiercely competing with open AI and meta for the hegemony of the artificial intelligence (AI) market, has unveiled a new AI agent ‘Alpibolb’, which can be applied to mathematics and science difficulties.
Google’s AI research division, Google Deep Mind, unveiled the AI agent ‘Alpi Bolve’ on the 14th (local time). In the blog post, the company introduced the Alpi Ballvi “Evolutionary Coding Agent,” and “AI that evolves the algorithm for the actual application of mathematics and computer science.”
Photo = Reuters Yonhap News
This AI works by finding a problem -solving method (algorithm) that can be used in various fields and developing the code itself based on the latest technology. Based on the Google AI model Geminai, we combine Geminai Flash, which explores the fastest, efficient and diverse ideas of the Geminai model group, and Geminai Pro, which provides the most powerful and deep insight.
Alpibolb uses automated evaluation indicators to verify, execute and evaluate the generated programs. These indicators objectively quantify accuracy and quality, so it is particularly useful in a systematic measurable area like mathematics and computer science, Google Deep Mine explained. As a result of applying it to more than 50 unresolved issues such as mathematics analysis, geometry, union theory, and water purification theory, the AI reaffirmed the existing highest level of solution for 75%of the problem, and 20%offered better solutions, Deep Mind said.
The AI model has been used in Google for the past year and has applied new algorithms throughout computing ecosystems such as data centers, hardware and software. As a result, the entire AI infrastructure has formed a more powerful and sustainable digital ecosystem, and 0.7%of Google’s total computing resources saved, Deep Mine explained. It also contributed to the development of Google’s AI chip TPU (Tensor processing unit), creating a new design code that removed unnecessary operations and improving the multiplication speed required for AI training by 1% to reduce Geminai training time by 1%.
Google Deep Mind said, “Alphabolv is developing into an algorithm that begins with the discovery of algorithms in a specific area and solves more complex reality problems.” This model is expected to evolve more sophisticated, leading to big changes throughout material science, new drug development, sustainability, technology and business. “
Reporter Seo Pil -woong seoseo@segye.com
[ⓒ 세계일보 & Segye.com, 무단전재 및 재배포 금지]
