Going all in on AI

By Sandy Cash

It was a moment of supreme serendipity at the AI Hub for Scientific Discovery. Inaugurated last year, the AI Hub is a community made up of Weizmann students, staff, and faculty working alongside technology experts from industry to create innovative strategies for using artificial intelligence to advance science.

At a recent meeting of the Hub membership, MSc student Yuval Steinberg presented an overview of her research-in-progress which was based on some out-of-this-world data: signals gathered by a spacecraft orbiting the planet Mars.

Yuval, a member of Prof. Oded Aharonson’s lab in the Department of Earth and Planetary Sciences, was doing a four-month internship at the Hub, during which she learned from experts how to use advanced AI techniques. As Yuval explained, she was using AI tools to identify patterns in the orbiter’s data that would reveal the structure of Mars’ underground geology down to a kilometer beneath the surface.

Unbeknownst to her, someone listening that day had the perfect background for this type of investigation. Dr. Gali Dekel, a theoretical physicist recruited to join the Hub as an AI Fellow, had worked for the oil and gas industry where, among other responsibilities, she helped map the subterranean layout of another planet: Earth. Reflecting on this coincidence, Dr. Dekel says that this multi-level meeting of the minds is exactly what the AI Hub for Scientific Discovery is all about.

“At the AI Hub, anyone can propose a project, and everyone can work together,” Dr. Dekel says. “It is an exciting place because it promotes collaboration between people who may come from entirely different backgrounds yet share a passion for using AI to answer really interesting research questions.” The Hub has already achieved proof of concept for its bottom-up approach, she adds.

“Hub members with years of industrial AI experience, but who might have no experience at all in a particular scientific discipline, team up with students and faculty in that field,” she explains. “These AI experts might say: ‘Give me your data and let me run it through my system.’ What that means is: let’s work together to check out your hypothesis using an algorithm that I developed for something else altogether.”

The AI ‘A-Team’

There are no private offices in the AI Hub, but rather banks of computers placed along countertops that border three small rooms. What would convince card-carrying members of Israel’s high-tech elite―mid-career AI experts who have held prominent leadership positions―to sign up for a stint in academia?

Prof. Amos Tanay of the Institute’s Department of Computer Science and Applied Mathematics played a leading role in envisioning and establishing this unique research community. As he sees it, the Hub offers industry experts something special: the opportunity to step away from profit-driven priorities and participate in projects that expand AI capabilities and application.

“Before the Hub opened as a physical space on campus, we put out the word that we were looking to hire AI Fellows―outstanding engineers, programmers, and R&D professionals from industry or top units of the IDF,” recalls Prof. Tanay, who today directs the Hub together with Dr. Sivan Refaely-Abramson of the Department of Molecular Chemistry and Materials Science and Prof. Elad Shneidman of the Department of Brain Sciences. “We wanted to find individuals who had a proven track record in AI and were looking for adventure. These specialists would join the Weizmann scientific community for one-to-three years, putting their skills to work on cutting-edge research projects.”

The job posting generated candidates in two categories: nominations from Weizmann faculty who had previously worked with an outside AI expert and wished to recruit them to be a temporary member of their lab; and candidates who saw the ad and nominated themselves.

“I enjoy my work in industry, particularly when it gives me the opportunity to communicate with people and solve challenging technical problems, but sometimes find it frustrating to be limited to a narrow range of topics,” says AI Fellow Eli Ovits, whose background includes training in electrical engineering and theoretical physics. “Over the years I resigned myself to seeing physics as something I did for my soul, and engineering as what I did for my job. Now I have the fun of bringing my whole self to the workplace, solving problems that span the entire research program of the Weizmann Institute.”

Programs for interns

Dr. Tamar Kashti had already been working at Weizmann when she was hired as one of the first AI Fellows. Thanks to her extensive team building experience in the industrial sector, she was―in addition to her research duties―appointed the AI Hub Manager. In this 

role, she oversees the “matchmaking” that turns this eclectic community of individuals into a team-based production line for discovery.

“Weizmann MSc and PhD students who want to join the Hub internship program submit proposals in which they describe a project they would like to pursue. These proposals, selected on a competitive basis, must make the case for how AI might help the student better understand their data, and must also demonstrate how the research has significant scientific value,” Dr. Kashti explains. “The students who submit winning proposals become Hub interns, earning the right to undertake three to four months of intensive AI training.”

The process begins with one-on-one meetings with an AI tutor. Employed by the Hub part-time, these tutors―mainly PhD candidates and postdocs who have high-level familiarity with AI and data science as well as some background on the research topic of the proposal―create a personalized program that matches each intern’s needs.

“The tutors’ goal is not simply to teach AI, but rather to teach Hub interns how to use AI for a specific scientific investigation,” Dr. Kashti says. “This is a fast-developing field; it’s not like our tutors can just grab a textbook off the shelf.”

The lessons the tutors create are based on real experimental data from Weizmann labs. Once interns master AI techniques needed to “crack” the training data, they move onto the next stage: applying their newfound AI skills to their individual projects.

These training sessions also contribute to the greater Hub community. Modules created for specific interns can be accessed and used by all Hub members.

“Our approach is strategically oriented toward ramping up AI literacy by providing a range of opportunities for discovering new AI approaches and gaining new skills,” Dr. Kashti says. “We also hold a daily morning meeting, in which everyone in the office that day gives a short overview of what they’re working on.”

This brief check-in helps interns make connections and identify people working on projects similar to their own but using different AI tools―which can open their eyes to alternative ways to approach their own research.

Impact and beyond

Prof. Shimon Ullman of the Department of Computer Science and Applied Mathematics is a renowned authority on computer vision. He directs the Institute for Artificial Intelligence―the overarching framework for AI at Weizmann. In addition to the Hub, the AI Institute supports initiatives devoted to the development of fundamental tools, and “challenge grants” that promote the use of AI by Weizmann faculty members.

According to Prof. Ullman, the integrated approach to AI education and research has the potential to have long-lasting impact, far beyond the Weizmann campus.

“The Weizmann Institute has always had an interdisciplinary, border-free approach to science,” Prof. Ullman says. “As techniques taught at the AI Hub become familiar to a greater number of Weizmann investigators, it will accelerate research success for labs in all disciplines. The Hub will also serve as a model for dynamic AI education, and for integrating AI into the scientific research being performed in institutions all over the world.”

Meet the AI Fellows

An expert in environmental physics and quantum chemistry, Dr. Adva Baratz earned her PhD at the Hebrew University of Jerusalem. Before joining the AI Hub, she worked at the Israel Institute for Biological Research in Ness Ziona, where she specialized in developing “unsupervised” AI methods.

“AI can analyze data using two general approaches. In supervised AI, we already know what attributes we want to search for, and we teach a machine to classify things according to these attributes. In unsupervised AI, on the other hand, we want the AI system to teach us which attributes are important for a certain mechanism, outcome, or property,” Dr. Baratz explains.

“As an AI Fellow, I’m collaborating with Dr. Sivan Refaely-Abramson on unsupervised methods for examining materials characterized by unique quantum correlation phenomena―mechanisms of great importance for quantum computing and quantum information storage. Our approach is general and can be applied to fields such as biology and environmental sciences, where multi-dimensional correlations play a crucial role.”

Dr. Ayala Cohen is a data scientist and computational chemist. After finishing her PhD under Prof. Leeor Kronik in the Weizmann Institute’s Department of Molecular Chemistry and Materials Science, she worked in industry developing AI algorithms for detecting changes in water quality and identifying viruses.

“As an AI Fellow, I’m working with Prof. Kronik once again, as part of the Open Catalyst Project (OCP), an international initiative that pools computational data to speed the search for new ways to increase the rate of chemical reactions,” she shares. “We use data from the OCP and expertise from the Kronik group to train AI models that can identify catalytic materials that could be 

used in renewable energy conversion reactions. The algorithm we’re developing―based on calculations that make it possible to model things that are difficult or even impossible to show through experiments―may advance the identification of catalysts for methane-based fuel generation, as well as other chemical reactions important to industry.”

Dana Shavit is an expert in deep learning, an AI method that teaches computers to process data in a way that is inspired by the human brain. Before becoming an AI Fellow, she was algorithm team leader for Neteera, a multinational company that developed an AI-enhanced radar to analyze patients’ vital signs in clinical settings. She also worked for Mobileye and Intel Corporation.

As an AI Fellow, Ms. Shavit is working with Prof. Tanay on cell tracking in early embryonic development, and with Prof. Rony Paz in the Department of Brain Sciences on incorporating deep learning into neuroscience.

“At the AI Hub, we work with scientists from all disciplines,” she says. “AI has a cross-discipline impact because whether it’s visual images, EKG readouts, or patterns of gene expression, the brain perceives and classifies data based on similar mathematical strategies. With AI, we build these same strategies into computerized systems.”

Dr. Gali Dekel is a deep learning expert and AI algorithm designer with professional training in geophysics. She earned her PhD in theoretical physics from Tel Aviv University. Working for the oil and gas industry, Dr. Dekel developed algorithms that map the Earth’s subsurface, and later led the AI team of a startup company in the field of indoor navigation. As an AI Fellow, she is collaborating with Prof. Yinon Rudich in the Department of Earth and Planetary Sciences on AI-based strategies for predicting extreme weather events.

“Coming back to academia has been an adjustment; it’s so quiet here, and so green!” Dr. Dekel exclaims. “What we’re building here in the AI Hub is crucial. Compared to industry, the scientific research community has a lot of catching up to do in terms of using AI methods to achieve its goals. As an AI Fellow, it’s nice for me to have the chance to apply my skills to the pursuit of pure knowledge, to be exposed to many 

different disciplines, and do work that is meaningful. It’s a win-win.”

Dr. Avital Steinberg is a computational chemist and expert in drug design. After earning her PhD at Ben-Gurion University of the Negev, she did postdoctoral research at the University of Pennsylvania and at the Weizmann Institute. In industry, she developed AI methods for use in drug discovery and also worked as a data scientist at Intel Corporation.

As an AI Fellow, Dr. Steinberg is working with Prof. Nir Yosef in the Department of Systems Immunology, developing methods for determining how perturbation of a certain gene or drug will impact the expression of other genes―a computational approach that can be used to investigate how a drug works.

“The stereotype of AI professionals is that we’re all geeks who like to sit by ourselves, but I love teaching,” Dr. Steinberg says. “A student intern recently asked me to weigh in on whether the project he was pursuing had commercial potential. Students at the AI Hub have access to people with industry experience, and that’s a good thing for both their AI training and their professional advancement.”

Dr. Tamar Kashti earned her PhD at Weizmann under particle physicist Prof. Yossi Nir and astrophysicist Prof. Eli Waxman. In industry, she was algorithm team leader at HP Indigo and Landa Labs―a science incubation center―and later returned to the Institute to join the lab of Prof. Yonina Eldar in the Department of Computer Science and Applied Mathematics, where she she applied AI methods to a variety of problems in computer vision and medical imaging.

Now, Dr. Kashti manages the Hub and is a part-time AI Fellow in the lab of Dr. Leeat Keren in the Department of Molecular Cell Biology, where she uses deep learning methods to de-noise multiplex images, making it easier to extract meaningful information from scans of complex tissues, including cancerous tumors.

“In industry, I learned to get the most out of teams of people with different backgrounds, and this experience is helpful as we build the AI Hub as a working scientific community,” she says. “By creating connections between labs, facilities, and ways of thinking, the AI Hub advances academic science, and also helps students develop the wide point of view Israeli industry needs.”

Noam Barkai is a software manager who recently served as VP of Research and Development for NRGene, an AI innovation company focusing on computational genomics. With a background that includes degrees in math, law, philosophy, and the philosophy of science, Mr. Barkai sees his position as an AI Fellow as an opportunity to get back to basics.

“I’m collaborating with Dr. David Zeevi [Department of Plant and Environmental Sciences] on the meta-genomics of bacteria,” he says. “We use AI to analyze microbial populations’ growth and division, something that may reveal the gene expression patterns that determine whether a particular population will survive. In the future, this basic research may contribute to engineered microbes for environmental applications.”

About making the jump from industry to academia, Mr. Barkai says: “I’m awed by the level of sophistication and commitment of the people at the AI Hub. My personal experience has been very positive and I’m grateful to be here.”

Eli Ovits is a problem solver. Studying physics and electrical engineering at the Technion-Israel Institute of Technology in the Atuda program―the academic corps of the Israel Defense Forces―he spent six years in the Israeli Air Force, where he did research and development and served as algorithm team leader for a signal processing group. During his military service, he completed his MSc at Tel Aviv University, studying the application of AI deep learning to quantum computing.

As an AI Fellow, Mr. Ovits is also working with Dr. Leeat Keren, focusing on problems of computer vision and cell segmentation. “I love the breadth of topics we deal with at the AI Hub. This community is special: it offers a great opportunity to make connections and find new ways in which I can make an impact.”

AI Ignite: The artificial intelligence road to game-changing discoveries

Artificial intelligence (AI) has the potential to accelerate, broaden, and deepen research across every scientific discipline while driving the development of real-world applications. This unique moment in time is our opportunity to shape AI capabilities, applications, and ethics in the early stages of this technology.

Through the AI Ignite flagship, Weizmann scientists are driving advancements in drug discovery, disease prediction, and personalized medicine, as well as sustainability, astrophysics, education, and more.

AI Ignite is designed around three critical pillars:

• Computing Infrastructure
  • The expansion of local and cloud-based computational resources and equipment on campus is essential for creating the robust infrastructure required for AI-driven research.
• A New AI-IT Building
  • The new AI and information technology (IT) building will house the people and systems necessary for Weizmann scientists to maximize the power of current AI tools while generating new applications.
• The Institute for Artificial Intelligence
  • This Institute promotes collaboration and innovation in fundamental AI research, including the development of new AI tools. Prof. Shimon Ullman, a 2015 Israel Prize laureate and a global authority on computer vision, is the Director of the Institute for Artificial Intelligence, which includes three sub-components:

Center for Core AI Research:

Here, scientists advance fundamental technologies for AI and machine learning. They also lay critical groundwork for ethics, fairness, responsibility, privacy, and other topics that will influence how AI is used for decades to come, creating a holistic strategy that allows AI to best serve humanity, now and in the future.

AI Hub for Scientific Discovery:

Here, students, scientists, industry professionals, and AI Fellows collaborate, using AI to answer some of the world’s most pressing scientific questions.

AI Ignite Challenge Grants:

Awarded annually via a competitive call for proposals, these grants support Weizmann researchers who seek to use AI in five areas of science:

• Building Blocks of Life: Using AI methods to design proteins that do not exist in nature, and which have specific, useful properties.
• AI for Human Health: Applying the superior data collection and analysis capabilities of AI for the benefit of doctors and patients around the world.
• Predicting a Sustainable Future: Utilizing AI to generate strategies for non-polluting energy sources, improve food security, and protect the environment.
• Smart Matter Design: Using AI to probe the behavior of atoms, molecules, and crystals, and to accelerate the development of novel functional materials.
• A Network of Thoughts: Creating a deep partnership between AI and neuroscience to help us understand human behavior, and develop new diagnostic tools and drugs for brain-based diseases.

The dawn of AI presents the greatest opportunity since the industrial revolution to create the kind of seismic change that can touch the lives of everyone on the planet. Through AI Ignite, Weizmann experts take their place as scientific pioneers, guiding humankind through this transformative point in history.

Blueprint for a strong future: The new AI-IT building

Starting in the 20th century, the invention of the microprocessor, the personal computer, the Internet, and the mobile phone fundamentally changed how people learn, conduct business, and interact with each other. Similarly, the advent of modern AI has ushered in a period of epoch-defining change. Recognizing that AI is destined to influence every aspect of its scientific program and institutional operations, the Weizmann Institute recently announced its plans to break ground on a facility from which all this change will be directed: a new information technology (IT) building that will serve as the permanent home of the AI Hub.

Rendering of the planned AI-IT building.

“The building will allow all of the Institute’s information specialists to sit together in one place, where they will pool their resources and expertise and develop new technologies for supporting computer users of all kinds,” says Prof. Yehiam Prior, the Weizmann Institute’s Chief Information Officer. “Given the growing need for computational resources, we are planning to increase the IT staff, from 180 to approximately 300. The new building will also be the home for the Institute’s AI specialists.”

This facility is scheduled to begin construction within two years and open its doors three years later. Situated next to a new gate that will open directly onto Rehovot’s technology park, the building will also provide a new home for Weizmann libraries, streamlining the process of digitizing resources, and making them more accessible to the wider community.

Designed for optimum heat flow and minimal energy waste, the IT building will feature a large, sky-lit atrium and interior staircases that will encourage interaction between different departments. Most importantly, the new building will provide a central space for the teamwork required to meet the needs of an AI-powered future.

“Today’s laboratory technologies generate enormous amounts of data that are already straining computational capabilities,” Prof. Prior states. “The further integration of AI into science will only increase the challenge. We have to be ready.”

AMOS TANAY IS SUPPORTED BY: Adelis Foundation, Laura and Anthony Beck and Family Fund for Research in a Data-Driven Approach to Fighting Blood Cancer, Moross Integrated Cancer Center

ODED AHARONSON IS SUPPORTED BY: Helen Kimmel Center for Planetary Science, Dr. Scholl Foundation Center for Water and Climate Research, Sussman Family Center for the Study of Environmental Sciences, The Stephen and Claire Reich Research Fellow Chair in Chemistry supports a Staff Scientist in Prof. Aharonson’s lab

SIVAN REFAELY-ABRAMSON IS SUPPORTED BY: Artificial Intelligence and Smart Materials Research Fund, in Memory of Dr. Uriel Arnon, Leah Omenn Career Development Chair

ELAD SHNEIDMAN IS SUPPORTED BY: Nella and Leon Benoziyo Center for Neurosciences, Carl and Micaela Einhorn-Dominic Brain Research Institute, Murray H. & Meyer Grodetsky Center for Research of Higher Brain Functions, David Lopatie Center for Theoretical and Computational Neuroscience, Joseph and Bessie Feinberg Professorial Chair

LEEOR KRONIK IS SUPPORTED BY:Artificial Intelligence for Smart Materials Research Fund, in Memory of Dr. Uriel Arnon, Tom and Mary Beck Center for Advanced and Intelligent Materials, Crown Materials Observatory, Aryeh and Mintzi Katzman Professorial Chair

NIR YOSEF IS SUPPORTED BY: Braginsky Center for the Interface between Science and the Humanities

LEEAT KEREN IS SUPPORTED BY: Abisch-Frenkel Foundation for the Promotion of Life Sciences, Azrieli Foundation, Sharon Levine and Jon Corzine, Enoch Foundation Research Fund, Fundación Alberto Palatchi, Rising Tide Foundation, Schwartz Reisman Collaborative Science Program, Fred and Andrea Fallek President’s Development Chair

DAVID ZEEVI IS SUPPORTED BY: Roden Family Research Fund for Environmental Sustainability