I recently had the honour of speaking on my research at the event "Talking Black: Voices From the Diaspora," organized by Mbuani Yi Africa in collaboration with UCD and Una Europa. This platform, dedicated to showcasing African expertise across diverse fields, gave me a unique opportunity to share insights on a topic close to my heart: Demystifying AI for Material Science. My talk centred on using advanced machine learning (ML) and Density Functional Theory (DFT) to predict complex material behaviours relevant to catalysis, electronics, energy storage, and artificial intelligence (AI) hardware.
During my presentation, I explored how AI, often perceived as a modern invention, traces its conceptual roots back centuries, with structured research beginning only in the mid-20th century. Early landmarks included the Turing Test, the pioneering AI programs of the 1950s, and the emergence of machine learning algorithms in the 1980s. Since then, AI has advanced rapidly, fuelled by computing power and data availability. I outlined the key categories of AI based on capabilities, functions, and learning techniques, explaining how machine learning (ML), a subset of AI, encompasses supervised, unsupervised, and reinforcement learning methods.
One aspect I highlighted was deep learning, a transformative subset of ML that uses neural networks with multiple layers to analyse intricate data patterns. This advancement has driven significant breakthroughs in image and speech recognition, autonomous driving, and more. I discussed models like convolutional neural networks (CNNs) and recurrent neural networks (RNNs), which have notably broadened AI's capabilities. My research specifically focuses on supervised learning, where AI systems learn from labelled datasets and excel when the desired outcomes are well-defined.
A critical component of my presentation was Natural Language Processing (NLP), a fascinating AI subset enabling human-computer interactions in everyday language. NLP powers applications like chatbots, translation tools, sentiment analysis, and voice-activated assistants. Large language models fall within NLP, utilizing deep learning and unsupervised learning techniques to enhance AI’s ability to understand and generate human language.
The core of my research lies at the intersection of AI and materials science. I utilize advanced machine learning force fields (MLFF) alongside ab initio Density Functional Theory (DFT) to explore the electronic, structural, and dynamical properties of Highly Oriented Pyrolytic Graphite (HOPG) and its interaction with water. The integration of MLFF significantly lowers the computational cost of DFT simulations, making it possible to investigate defect control and vacancy engineering within HOPG – a synthetic, highly ordered form of graphite with unique physical and chemical properties that make it invaluable in scientific and industrial applications.
Throughout my lecture, I emphasized AI's boundless potential to tackle complex global challenges, from enhancing human-computer interactions to pioneering material discovery, neuromorphic computing, and self-driving laboratories. I shared how AI’s application extends to fields such as robotics, catalysis, energy storage, and quantum computing, opening doors to remarkable innovations in material behaviour prediction.
In closing, I touched upon the ethical considerations AI brings to the forefront. Issues like algorithmic bias, privacy, and the future of employment must be addressed thoughtfully. Responsible AI development is essential, with transparency, fairness, and accountability as guiding principles. By setting ethical standards and regulations, we can ensure that AI serves the collective good.
"Persevere, for your future will use your present to mould what is yet to become." – Mary Ajide
Comments