- AI for Scientific Discovery
Pioneering New Frontiers of Knowledge
| Olga Fink
Assistant Professor, Intelligent Operations and Maintenance Systems, Swiss Federal Institute of Technology in Lausanne
| Thomas Hartung
Professor, Bloomberg School of Public Health, Johns Hopkins University
| SangYupLee
Senior Vice President, Research; Professor Emeritus, Korea Advanced Institute of Science and Technology
| Andrew Maynard
Professor, School for the Future of Innovation in Society, Arizona State University
Breakthroughs in artificial intelligence (AI) – such as deep learning, generative AI, and other foundational paradigms – enable scientists to make discoveries that would otherwise be nearly impossible and accelerate the pace of scientific discovery in general.
Over the past few years, there has been a transformation in the way AI is used in scientific discovery. From Deep Mind’s AlphaFold – an AI system that accurately predicts 3D models of protein structures – to the discovery of a new class of antibiotics and materials for more efficient batteries, the world is on the brink of an AI-driven revolution in how new knowledge is discovered and used.1,2,3According to a recent report by the US President’s Council of Advisors on Science and Technology, “AI has the potential to transform every branch of science and many aspects of how we do science.”4
While AI has been used in research for many years, recent advances in deep learning, generative AI, and foundational models are transformative. Scientists are building and using large language models to mine scientific literature, working with AI chatbots to brainstorm new hypotheses, creating AI models capable of analyzing large amounts of scientific data, and using deep learning to make discoveries. They are also exploring how AI and robotics can be integrated with laboratory methods to accelerate research in innovative ways.
AI is thus emerging as a transformative, general-purpose technology in scientific research that can uncover insights that would otherwise remain hidden. At the current pace of innovation, these have the potential to lead to advances in areas such as:
Diagnosis, treatment, and prevention of disease.
New materials that enable next-generation green technology.
Breakthroughs in the life sciences that expand our current understanding of biology.
Transformational leaps in understanding the human mind, and much more.
Scientists predict that general-purpose AI will transform every aspect of scientific discovery in the coming years. Researchers can build on past discoveries to envision new possibilities – AI enables connections to be made and inferences to be made that are beyond the capabilities of the human mind alone.
There are still ethical considerations and challenges – the extent to which these powerful technologies pose risks to individual privacy, autonomy, and identity, as well as the potential for social disruption, are not yet fully understood. 5 In addition, the environmental impact of the energy consumption and resource extraction required to sustain AI development must also be considered.
Likewise, more research is needed to effectively manage the impact of the technology. 6 For example, addressing inherent biases in data sets and enhancing the reliability of model-generated content is critical to scientific integrity. Ensuring ethical use of data and protecting the privacy of research subjects requires strict security measures. Navigating intellectual property rights, particularly ownership and copyright of model-generated content, is essential to a collaborative environment and must be addressed.
- Privacy-Enhancing Technologies
Empowering Global Collaboration at Scale
| Olga Fink
Assistant Professor, Intelligent Operations and Maintenance Systems, Swiss Federal Institute of Technology in Lausanne
| Lisette van Gemert-Pijnen
Professor, Persuasive Health Technologies, University of Twente
| Dongwon Lee
Professor, Pennsylvania State University
| Andrew Maynard
Professor, School for the Future of Innovation in Society, Arizona State University
| Bastiaan van Schijndel
Director of Innovation, ZORGTTP
Access to increasingly large data sets – especially when using AI – will transform research, discovery, and innovation. However, concerns about privacy, security, and data sovereignty limit the extent to which high-value data can be shared and used nationally and globally. A powerful set of emerging technologies makes it possible to share and use sensitive data in ways that address these concerns.
Privacy-enhancing technologies enable the safe sharing and use of sensitive data.
Source: Midjourney and Studio Miko.
In recent years, there has been a surge of interest in “synthetic data.”7 These data replicate patterns and trends in sensitive data sets but do not contain specific information that could be linked to individuals or harm organizations or governments. Powered by advances in AI, synthetic data eliminates many of the limitations of working with sensitive data and opens up new possibilities for global data sharing and collaborative research on biological phenomena, health-related research, training of AI models, and more. However, even if aggregated data were to be created at the national level, health trends in a source country would be exposed and such concerns would need to be addressed.
There is also renewed interest in homomorphic encryption, a technology that dates back to the 1970s.8,9 Rather than recreating datasets with the same characteristics as the raw data, homomorphic encryption allows analysis of encrypted data without direct access to the raw data. While promising, such encryption requires significantly more energy and
time to achieve secure results.
As advances in AI transform the value of data, techniques such as synthetic data generation and homomorphic encryption are predicted to enable data sharing and access while ensuring privacy, security, and data sovereignty. In health-related research, in particular, access to data in a way that does not infringe on the rights and safety of individuals and communities has shown promise for advancing progress in the detection, treatment, and prevention of disease.10
Effective data sharing and use technologies that protect privacy, security, and data sovereignty are essential if the emerging potential of AI is to be realized. However, despite their potential, synthetic data and homomorphic encryption have some limitations. These limitations include poor representation of outliers or potentially important outliers in the case of synthetic data, and the inability to infer or reconstruct sensitive data despite the anonymity of both techniques. Further research into the technologies and the policies around them is needed to ensure their success.11
- Reconfigurable Smart Surfaces
Transforming Wireless Connectivity with Smart Mirrors
| Mohamed-Slim Alouini
Al-Khwarizmi Emeritus Professor,
Electrical and Computer Engineering, King Abdullah University of Science and Technology
| Joseph Costantine
Associate Professor, Electrical and Computer Engineering, American University of Beirut
| Marco Di Renzo
CNRS Research Director, Laboratory of Signals
and Systems (L2S), Université Paris-Saclay
| Javier Garcia-Martinez
Professor, Chemistry and Director, Laboratory of Molecular Nanotechnology, University of Alicante
The demand for higher data rates, lower latency, and energy-efficient connectivity is skyrocketing globally.12The highly anticipated launch of 6G in 2030 is expected to add to this pressure. To meet these challenges, future networks will need to be designed to increase capacity and connectivity, while also focusing heavily on environmental sustainability. Enter reconfigurable smart surfaces (RIS), platforms that use metamaterials, intelligent algorithms, and advanced signal processing to transform ordinary walls and surfaces into smart components for wireless communication.
RIS increases data rates and energy efficiency while reducing interference, and is critical for next-generation wireless networks.
Source: Midjourney and Studio Miko.
Similar to the idea of “smart mirrors,” RIS allows for precise, focused control of electromagnetic waves, reducing interference and the need for high transmission power. Likewise, RIS is highly adaptive and can dynamically adjust its configuration according to real-time needs. This adaptability enables efficient resource utilization and increased energy efficiency in wireless networks.13,14,15
The evolution of hardware platforms and the rise of experimental initiatives in the RIS field have attracted significant interest from telecommunications stakeholders interested in exploring its potential for next-generation wireless networks. A key milestone has been the efficient integration of RIS into existing wireless networks. Several RIS platforms have demonstrated the impressive capabilities of this technology from a hardware perspective.16
The development of RIS could have a broad impact on many industries.17For example, modulated radio transmissions in smart factories can ensure reliable communication in highly complex environments. RIS enables sensors to transmit data with minimal power for the Internet of Things (IoT), which is energy-intensive. For vehicular networks, RIS enhances safety by enabling robust communication between vehicles and infrastructure. For improving coverage in agricultural settings, RIS is a promising solution with low power consumption and high cost-effectiveness.18
Market intelligence reports indicate that RIS is on the verge of exponential adoption and growth. Several companies, including Rhode & Schwarz, Huawei, ZTE, Intel, and Samsung, are all investing in RIS, sending a strong signal that RIS will play a central role in the telecommunications landscape in the coming years.19
However, before this happens, several prominent challenges need to be addressed, including high hardware costs and the need for clear standards and regulations for the safe and ethical use of
the technology.20
- High-Altitude Base Stations
Bridging the Internet Gap from the Stratosphere
| Mohamed-Slim Alouini
Al-Khwarizmi Professor Emeritus of Electrical and Computer Engineering, King Abdullah University of Science and Technology
| Mariette DiChristina
Dean and Professor, Journalism Practice, School of Communication, Boston University
High-altitude base stations (HAPS) operate at stratospheric altitude, approximately 20 km above Earth. Typically taking the form of balloons, zeppelins, or fixed-wing aircraft, they provide a stable platform for observation and communications and can operate for months at a time. Advances in solar panel efficiency, battery energy density, lightweight composite materials, avionics, and autonomous antennas, along with expanding spectrum and new aviation standards, make HAPS feasible in the near future. HAPS can provide connectivity, coverage, and performance improvements that neither satellites nor ground towers can match, especially in areas with difficult terrain such as mountains, jungles, or deserts.21
Access to a connected world acts as a bridge to the future, creating pathways to prosperity and new educational possibilities, and strengthening the foundations of social connectedness. Yet, according to the International Telecommunication Union (ITU), about a third of people worldwide remain offline. Women and older adults are disproportionately affected.22A key element to addressing this challenge is better infrastructure.
HAPS can improve connectivity for communities underserved by traditional communications infrastructure, especially in remote areas. The COVID-19 pandemic has highlighted the critical nature of internet access, showing how disparities in connectivity exacerbate socioeconomic inequality. By bridging this digital divide, HAPS technology can help provide access to education, healthcare, and economic opportunities.
In addition to providing internet access, these adaptive platforms could play a key role in a range of critical applications, from supporting disaster management to enhancing broadband coverage and environmental monitoring. HAPS’ ability to rapidly deploy and adapt to changing conditions could make them an invaluable tool in managing emergencies, where timely information and communication can save lives.23
Investments in HAPS from aerospace engineering leaders have led to advances in materials, propulsion systems, and solar cell technology.24HAPS are now economically viable for commercial and real-world deployment. Organizations with the deep knowledge and resources to develop reliable, resilient HAPS have supported its growth and role in the future of communications infrastructure.
Industry examples include the Airbus Zephyr, Thales’ Stratobus, and Boeing Aurora projects. Lower latency, reduced costs, higher capacity, easy hardware upgrades, and faster deployment are all compelling commercial propositions. The market size is estimated at $783.3 million by 2023 and is expected to grow at a CAGR of 10.4% from 2023 to 2033.25
However, HAPS, which operate at stratospheric altitudes for extremely long durations, differ from conventional manned aircraft in a number of ways, and the current regulatory framework is not fit for purpose. Organizations such as the International Civil Aviation Organization (ICAO) are actively discussing new policies and guidelines to enable responsible deployment of HAPS.26