In recent years, the buzz surrounding artificial intelligence (AI) has led many to believe it could revolutionize various fields, from healthcare to energy management. However, amidst this excitement lies a pressing question: Can AI genuinely aid in discovering new materials? This inquiry is at the forefront of discussions in materials science, particularly as the world seeks innovative solutions for climate technology and energy storage.
A recent feature package titled “Hype Correction” delves into the realistic capabilities of AI, particularly focusing on its implications for materials research. Noteworthy contributions, such as those from David Rotman, highlight the potential of AI to expedite the discovery of new materials critical for advancements like batteries and semiconductors. Despite the promising outlook, experts caution that the field must demonstrate its ability to create truly innovative and functional materials. Historically, inventing new materials has been a laborious process, with significant breakthroughs occurring infrequently. For instance, while the first synthetic plastic was created in 1907, it took several decades for the material to be widely adopted, illustrating the slow pace of material innovation.
Currently, companies like Lila Sciences are exploring the integration of AI in material discovery. By utilizing AI models trained on extensive scientific literature and experimental data, Lila aims to streamline the iterative process of material invention. At a recent MIT Technology Review event, co-founder Rafael Gómez-Bombarelli emphasized that while substantial breakthroughs in AI-driven materials discovery are yet to be realized, the insights generated by their models have proven to be as profound, if not deeper, than those of experienced scientists. However, the challenge remains in translating AI findings into practical applications, as the creation of a novel and useful material involves more than just theoretical predictions. Past attempts, such as Google DeepMind’s claim of using AI to predict millions of new materials, have faced scrutiny, with some identified materials being mere variations of existing ones or even non-viable under typical conditions.
While the potential for AI to invigorate materials research is undeniably exciting, the journey to significant advancements is still ongoing. As the technology evolves, only time will reveal if AI can indeed fulfill its promise in revolutionizing material discovery and ultimately lead to groundbreaking innovations in energy and technology.
Source: Can AI really help us discover new materials? via MIT Technology Review