Details

Published on Tuesday, 11 October 2022 04:08

A key facet of the SGX3 mission is to provide forward-looking studies of next-generation Science Gateway capabilities.

SGX3 offers a new service called Blueprint Factories, in which we will work with collaborators to better understand the CI needs of entire research communities and national-scale cyberinfrastructure providers. Blueprint Factories are an evolution of the Tech Summit concept pioneered by the SGCI. Each Blueprint Factory will run for 18 months, where CI professionals, domain scientists, and major infrastructure operators come together to plan and conduct pain point discovery activities that include surveys and one-on-one interviews. These will be synthesized into practical white papers outlining the discoveries and providing CI blueprints to address the findings. Each Blueprint Factory will have approximately five core CI and customer discovery experts from SGX3 and five domain experts as the main members. The domain experts will provide contacts to leaders in their field, review the team’s progress on a monthly or bi-monthly basis, and help synthesize the results into final Blueprint Factory documents. Are you interested in participating in one of our Blueprint Factories but want to understand the process further? Visit our Blueprint Factory proven process page for details.

Watch here for announcements regarding our Blueprint Factory activities, or apply here if you feel you would like to be part of a Blueprint Factory or wish to suggest a Blueprint Factory topic for the future.

Read more about the Blueprint Factories: Constructing the Future.
 

AI and Science Gateways Blueprint Factory

Artificial intelligence (AI) is increasingly prevalent in contemporary research. However, its application in various scientific disciplines is still in its early stages, and the full implications of AI on science gateways and computing resources remain unclear.

Apart from extensive model training tasks, the computing resource requirements for AI differ from traditional high-performance computing (HPC) workloads. Large model training tasks involve characteristics distinct from traditional HPC loads, as they are iterative, interactive, include natural checkpoints, necessitate a greater number of specialized processors, and involve different model validation processes. The workflow in AI significantly deviates from traditional modeling.

Moreover, data management becomes a pivotal concern in AI model research. Extensive datasets are required for training AI models, and the size of these datasets often makes transferring them to the most powerful compute resources a challenging task. Additionally, AI models are frequently used for data recall and are often located closer to the data source compared to traditional physics-based models. AI-based models can also be updated frequently with incremental training, necessitating precise management of data provenance, which includes tracking the data used to train specific model versions and the model versions used in generating scientific findings.

It is also essential to address the ethical and data privacy considerations that accompany the proliferation of AI in scientific research. The handling and analysis of vast amounts of data can inadvertently raise concerns about privacy, data security, and the potential for bias in AI models. Researchers and institutions must navigate these challenges responsibly, ensuring that data is used in ways that respect individual privacy and that AI models are trained and deployed without reinforcing existing biases, perpetuating discrimination, or in a manner where they are used outside of the bounds within which they were trained. Ethical considerations are central to the responsible development and application of AI in science, and their incorporation into AI research and science gateways is paramount for the ethical progression of this field.

These differences suggest that the issues of data privacy, model suitability for a given task, and orchestration of training, validation, data management, and recall in AI research can be complex and potentially eased through the use of science gateways. In this Blueprint Factory project, we are exploring the relationship between science gateways and the needs of researchers who utilize, or wish to utilize, AI techniques in their work. It's important to note that our focus is not on aiding core research in new AI methodologies but, instead, on how science gateways can streamline the utilization of AI techniques in various domain-specific scientific research endeavors. Furthermore, we aim to investigate how science gateways can broaden access to AI techniques and underlying computing resources for domain-specific researchers who may need the means to leverage these technologies independently.

To become involved, please contact us at help@sciencegateways.org.

Sustainability Blueprint Factory

Even the longest-running, seemingly stable Science Gateways face challenges to their ongoing survival for several reasons, including inadequate financial strategies, loss of key personnel, the need to constantly maintain and refresh technical infrastructure, irregular funding cycles, and the fundamental gaps in staff capacity that occur when a team transitions a grant-funded research project to a reliable, ongoing service offering for users who rely upon it. These challenges have been well documented. Until now, SGX3 (and previously, SGCI) has primarily focused on addressing these challenges by supporting the success of individual gateway projects and their leaders.

The Sustainability Blueprint Factory, led by Nancy Maron, founder of BlueSky to BluePrint, plans to broaden the lens on this issue, by asking not just what the challenges are but, what exactly, would define a successful outcome. And at what level - funder? academic institution? federal policy? - must they be addressed to have real impact? 

This Blueprint Factory will:

• Empanel a small group of internal and external experts to define the scope of this BluePrint Factory with an eye to representing a diversity of institution types and stakeholders. 

• Organize a Grand Challenges/Grand Solutions Project. By organizing and undertaking series of convenings of university stakeholders in different roles (presidents, deans, vice presidents for research, heads of information technologies) at a range of types of higher institutions (such as research-intensive universities, minority-serving institutions, liberal arts colleges, and community colleges), the Grand Challenges/Grand Solutions Project will outline grand sustainability challenges and prioritize which ones might be addressable over next five to ten years.

• Develop practical guidance for leaders of science gateways. Through focus groups and a short survey, we will encourage input on the types of tools and guidance they most need to continue to grow their projects’ impact in the near term. 

A final white paper will report on findings and outline an agenda for addressing ongoing  challenges. A set of practical tools and guidance, which may include briefing papers and  a  “health check” for gateway providers, will be available through the SGX3 Blueprint Factory website.
 

Apply here if you feel you would like to be part of a Blueprint Factory or wish to suggest a Blueprint Factory topic for the future.