Deepmind’s AI Model Can Predict the Transition of Glass Molecules

Venturebeat | April 07, 2020

Deepmind’s AI Model Can Predict the Transition of Glass Molecules
  • The work yields insights into general substance and biological transitions, and that it could lead to advances in industries like manufacturing and medicine.

  • The physics behind glass transition underlie behaviormodeling, drug delivery methods, materials science, and food processing.

  • DeepMind claims the insights gleaned could be useful in predicting the other qualities of glass; as alluded to earlier, the glass transition phenomenon manifests in more than window (silica) glasses.


In a paper published in the journal Nature Physics, DeepMind researchers describe an AI system that can predict the movement of glass molecules as they transition between liquid and solid states. The techniques and trained models, which have been made available in open source, could be used to predict other qualities of interest in glass, DeepMind says.


Beyond glass, the researchers assert the work yields insights into general substance and biological transitions, and that it could lead to advances in industries like manufacturing and medicine.


Machine learning is well placed to investigate the nature of fundamental problems in a range of fields. We will apply some of the learnings and techniques proven and developed through modeling glassy dynamics to other central questions in science, with the aim of revealing new things about the world around us.

- DeepMind


 

Glassy dynamics


Glass is produced by cooling a mixture of high-temperature melted sand and minerals. It acts like a solid once cooled past its crystallization point, resisting tension from pulling or stretching. But the molecules structurally resemble that of an amorphous liquid at the microscopic level.


Solving glass’ physical mysteries motivated an annual conference by the Simons Foundation, which last year hosted a group of 92 researchers from the U.S., Europe, Japan, Brazil, and India in New York. In the three years since the inaugural meeting, they’ve managed breakthroughs like supercooled liquid simulation algorithms, but they’ve yet to develop a complete description of the glass transition and predictive theory of glass dynamics.


That’s because there are countless unknowns about the nature of the glass formation process, like whether it corresponds to a structural phase transition (akin to water freezing) and why viscosity during cooling increases by a factor of a trillion. It’s well-understood that modeling the glass transition is a worthwhile pursuit — the physics behind it underlie behaviormodeling, drug delivery methods, materials science, and food processing. But the complexities involved make it a hard nut to crack.

 

AI and machine learning


Fortunately, there exist structural markers that help identify and classify phase transitions of matter, and glasses are relatively easy to simulate and input into particle-based models. As it happens, glasses can be modeled as particles interacting via a short-range repulsive potential, and this potential is relational (because only pairs of particles interact) and local (because only nearby particles interact with each other).


The DeepMind team leveraged this to train a graph neural network — a type of AI model that directly operates on a graph, a non-linear data structure consisting of nodes (vertices) and edges (lines or arcs that connect any two nodes) — to predict glassy dynamics. They first created an input graph where the nodes and edges represented particles and interactions between particles, respectively, such that a particle was connected to its neighboring particles within a certain radius. Two encoder models then embedded the labels (i.e., translated them to mathematical objects the AI system could understand). Next, the edge embeddings were iteratively updated, at first based on their previous embeddings and the embeddings of the two nodes to which they were connected.


After all of the graph’s edges were updated in parallel using the same model, another model refreshed the nodes based on the sum of their neighboring edge embeddings and their previous embeddings. This process repeated several times to allow local information to propagate through the graph, after which a decoder model extracted mobilities — measures of how much a particle typically moves — for each particle from the final embeddings of the corresponding node.


READ MORE: AMAZON PROPOSES “BACKWARD-COMPATIBLE TRAINING” FOR COMPUTER VISION MODELS

 

Testing the model


The team validated their model by constructing several data sets corresponding to mobilities predictions on different time horizons for different temperatures. After applying graph networks to the simulated 3D glasses, they found that the system “strongly” outperformed both existing physics-inspired baselines and state-of-the-art AI models.


They say that network was “extremely good” on short times and remained “well matched” up to the relaxation time of the glass (which would be up to thousands of years for actual glass), achieving a 96% correlation with the ground truth for short times and a 64% correlation for relaxation time of the glass. In the latter case, that’s an improvement of 40% compared with the previous state of the art.


In a separate experiment, to better understand the graph model, the team explored which factors were important to its success. They measured the sensitivity of the prediction for the central particle when another particle was modified, enabling them to judge how large of an area the network used to extract its prediction. This provided an estimate of the distance over which particles influenced each other in the system.


They report there’s “compelling evidence” that growing spatial correlations are present upon approaching the glass transition, and that the network learned to extract them. “These findings are consistent with a physical picture where a correlation length grows upon approaching the glass transition,” wrote DeepMind in a blog post. “The definition and study of correlation lengths is a cornerstone of the study of phase transition in physics.”

 

Applications


DeepMind claims the insights gleaned could be useful
in predicting the other qualities of glass; as alluded to earlier, the glass transition phenomenon manifests in more than window (silica) glasses.  The related jamming transition can be found in ice cream (acolloidal suspension), piles of sand (granular materials), and cell migration during embryonic development, as well as social behaviors such as traffic jams.


Glasses are archetypal of these kinds of complex systems, which operate under constraints where the position of elements inhibits the motion of others. It’s believed that a better understanding of them will have implications across many research areas. For instance, imagine a new type of stable yet dissolvable glass structure that could be used for drug delivery and building renewable polymers.


Graph networks may not only help us make better predictions for a range of systems, but indicate what physical correlates are important for modeling them that machine learning systems might be able to eventually assist researchers in deriving fundamental physical theories, ultimately helping to augment, rather than replace, human understanding.

- Deepmind


READ MORE: AI TO READ HUMAN THOUGHTS AND COVERT THEM INTO TEXTS

Spotlight

Buy, manage, and upgrade network and infrastructure software. Address demands and deliver capabilities when you need them with Cisco ONE Software.


Other News
SOFTWARE

Sesame Software Releases Relational Junction 6.2 with Extended Support for New SaaS Applications and Databases

Sesame Software | September 24, 2021

Sesame Software, the innovative leader in Enterprise Data Management, today announced the rollout of Relational Junction 6.2, the latest version of its suite of data management and replication tools, giving companies the ability to effortlessly create data warehouses for any database or API-enabled application. Relational Junction has also added support for many data warehouse platforms, including Oracle Autonomous Data Warehouse, Snowflake, Google BigQuery, Greenplum, Redshift, and Teradata, using native bulk loaders when appropriate for exceptional performance. Why is this important? Relational Junction's entire focus is on putting all your data into an instant, fully automated data warehouse that gives you complete control of your data. With only minutes of configuration, customers can automatically build a schema and efficiently move data for business intelligence, analytics, and integration. "This release has evolved from its predecessor suite of products into a single product that builds on-demand data warehouses out of any database or API-enabled SaaS application with no code, no design, no data mapping, and lights-out continuous operation". Rick Banister, founder and CEO of Sesame Software Security is at the heart of the product architecture. Sesame Software does not host your data or the product, eliminating all security concerns about vendors potentially allowing data breaches. Instead, Relational Junction can be installed on any private cloud or on-premise hardware platform. Want to use AWS, Oracle OCI, Google Cloud, or Azure? Or just drop it onto your laptop? Sesame Software can support you. UNIX or Windows? No problem. "By integrating data from external and internal sources with Relational Junction, organizations end up with a relational database that's fully secure and optimized for their specific needs," says Banister. "This gives every data-driven company real-time 360-degree access to their most important data, ensuring that sales, marketing, and the C-Suite are aligned for day-to-day decision making and long-term strategic planning." About Sesame Software Headquartered in Santa Clara, California, Sesame Software is the Enterprise Data Management leader, delivering data rapidly for enhanced reporting and analytics. Sesame Software's patented Relational Junction suite offers superior solutions for data warehousing, integration, as well as backup, and compliance to fit your business needs. Quickly connect to SaaS, on-premise, and cloud applications for accelerated insights.

Read More

AI APPLICATIONS

RavenPack Launches New Multilingual Artificial Intelligence (AI) Platform to Monitor Risks Globally

RavenPack | September 23, 2021

RavenPack, the leading provider of technology and insights for data-driven companies, has announced today the release of RavenPack Edge, a new AI platform that collects, reads, and analyzes billions of documents to help businesses better monitor and mitigate emerging risks. Capable of understanding content in 13 different languages, Edge can extract insights from all types of documents —from short news articles to complex legal filings. RavenPack Edge monitors any information published on over 12 million entities including public and private companies and organizations, key executives and political figures, and many other topics of interest. Every time one of these entities is mentioned across 50,000 sources of curated content, Edge calculates the relevance, sentiment, or novelty of the information, and pushes these analytics to subscribers in real-time. Whether the content is written in English, Arabic, Mandarin, or any other major language, the new platform can make sense of it all to deliver comprehensive analytics to end users. Preparing for the post-pandemic business world The Covid pandemic has forced companies to reassess the way they monitor emerging risks. To thrive, companies need to know what is being said about their products, their competition, their suppliers, and even their customers. You need to keep your finger on the pulse and constantly monitor business, macroeconomic and geopolitical events that may prove critical to the survival of your business. Armando Gonzalez, CEO of RavenPack Financial firms have long understood that actionable information is increasingly found in the oceans of news and digital content available. In the nearly 20 years since the technology firm was founded, RavenPack has built a sterling reputation on Wall Street for the unparalleled breadth and quality of its low-latency text processing and data products. With Edge, RavenPack sets a new standard for its traditional user base, and further extends its reach by helping non-financial firms better mitigate risk exposures in investments, supply chain, client compliance, reputation management, competitive analysis, and sustainability. A new generation of multilingual AI Among some of the latest innovations in artificial intelligence achieved in RavenPack Edge are: Multilingual text understanding Thematic sentiment scoring (credit, risk, and sustainability impact) Focused and expanded taxonomies, including ESG RavenPack maintains a database of over 20 years of historical content that includes news and social media, industry and earnings call transcripts, insider transactions, and other regulatory filings. About RavenPack Since 2003, RavenPack has been one of the leading data analytics providers in financial services, allowing firms to quickly extract value and insights from large amounts of unstructured text data. The company's clients include some of the most successful and sophisticated hedge funds, banks, and asset managers in the world. Our products allow companies to enhance returns, reduce risk, and increase operational efficiency in an automated fashion.

Read More

SOFTWARE

Gnani.ai Launches armour365™ Voice Biometrics Software Based on Patented Tech

Gnani.ai | September 22, 2021

Gnani.ai, a frontrunner in Conversational AI and voice security domain, today announced the launch of its home-grown Voice Biometrics software. Christened armour365™, the biometric solution boasts of path-breaking features to cater to new and emerging risks in fraud prevention and information security. armour365™ Voice Biometrics works on 300 plus proprietary audio features and comes with out-of-the-box integrations to multiple contact center software providers and messaging apps. The solution is equipped with top-of-the-line features like “anti-spoof layer,” “replay attack detection” and “one enrollment” to offer unparalleled security and CX for industries such as Contact Centers, Banks, Defence, Healthcare, etc. for applications ranging from omnichannel customer authentication to secured access to sensitive devices. Voice Biometrics can be a blessing in disguise for Contact Center and Infosec leaders battling fraud and data theft through various communication channels. With a response time of less than 500 milliseconds, armour365™ has been engineered to help IT and security teams to replace error-prone and legacy authentication methods like PINs and passwords to realize the potential of Voice Biometrics being truly contactless. Ganesh Gopalan, Co-founder and CEO, Gnani.ai armour365™ can be implemented and integrated into any CRM for seamless authentication without the need for coding or a device interface. We are thrilled to launch armour365™ with industry-leading accuracy and thankful to our Engineering and Product teams to have pulled this off indigenously. We’re confident our customers will benefit immensely with this ‘low-cost and no-code solution’ built to offer reliable voice security. Ananth Nagaraj, Co-founder and CTO, Gnani.ai Gnani.ai plans to host the offering on major cloud marketplace platforms for customers and developers to take advantage of competitive pricing and ease of accessing the solutions through APIs. About Gnani.ai Gnani.ai, a Samsung Ventures funded company, is a leader in the voice-led conversational AI automation space. Gnani.ai offers no-code multi-modal and multi-channel bot automation platforms. With partners like Avaya, Nvidia, and Intel, Gnani.ai is leading the Conversational AI revolution.

Read More

AI TECH

Deloitte Launches the Deloitte AI Academy to Advance Artificial Intelligence Proficiency for Business and Society

Deloitte | September 21, 2021

Deloitte today announced the launch of the Deloitte AI Academy™ which is designed to help bridge the technology talent gap by developing and re-skilling today's workforce with immersive training in the AI capabilities required for the digital economy. The Deloitte AI Academy will parallel Deloitte's Cyber and Cloud Institute development strategies, and demonstrate Deloitte's commitment to combine in-depth business knowledge with a mastery of technology to help its people and clients thrive in increasingly dynamic markets. The Deloitte AI Academy will provide a comprehensive learning experience, equipping practitioners with the skills needed to deliver AI projects through programs that include a hands-on immersive bootcamp. The Deloitte AI Academy brings together an education ecosystem from academia, technology companies, corporate learning providers, Deloitte AI specialists, career development and talent experience programs. Academy participants will learn technical data and AI skills, fundamentals of Trustworthy AI™, knowledge of how AI is being applied across different industries, and professional skills. The Deloitte AI Academy bridges AI technical competencies — AI-enabled data engineering and AIOps, machine learning algorithms, computer vision, conversational AI, and deep learning/neural networks — with domain knowledge in customer and marketing, cyber, risk and compliance, tax, audit and assurance, application engineering and ERP and technology services. Our approach to AI training and education, combining AI skills with business domain knowledge, can help build the next generation of business-ready AI talent. Irfan Saif, Deloitte U.S. AI co-leader It's our responsibility as a society and as business leaders to develop new talent with AI skills – not only for the engineers and data scientists, but also for every role in an organization, no matter how technical, Through the Deloitte AI Academy we are endeavoring to develop future leaders with a higher level of AI proficiency for the benefit of our clients and society at large. Dan Helfrich, chairman and CEO, Deloitte Consulting LLP The Deloitte AI Academy aims to educate the next generation of AI professionals to broaden the pool of AI talent for Deloitte, clients and society. Deloitte is launching the Deloitte AI Academy via a pilot program in India, with plans to train up to 10,000 professionals across the United States and other markets in the next four years. "At Deloitte, we combine world-class business knowledge with a full command of technology, working together with our clients to engineer their future and help them get there," said Nitin Mittal, Deloitte U.S. AI co-leader. "Companies across all industries are scrambling to secure top AI talent from a pool that's not growing fast enough. The AI Academy will be instrumental in helping to close the AI skills gap and enabling companies to fully deliver on the promise of AI." The Deloitte AI Academy works in collaboration with the Deloitte AI Institute™, providing learning and training to complement the Institute's focus on supporting the positive growth and development of AI through engaged conversations and innovative research. About Deloitte Deloitte provides industry-leading audit, consulting, tax and advisory services to many of the world's most admired brands, including nearly 90% of the Fortune 500® and more than 7,000 private companies. Our people come together for the greater good and work across the industry sectors that drive and shape today's marketplace — delivering measurable and lasting results that help reinforce public trust in our capital markets, inspire clients to see challenges as opportunities to transform and thrive, and help lead the way toward a stronger economy and a healthier society. Deloitte is proud to be part of the largest global professional services network serving our clients in the markets that are most important to them. Building on more than 175 years of service, our network of member firms spans more than 150 countries and territories.

Read More

Spotlight

Buy, manage, and upgrade network and infrastructure software. Address demands and deliver capabilities when you need them with Cisco ONE Software.

Resources

Events