In this episode, Neil interviews Professor Max Welling, one of the foremost experts in Machine Learning about AI4Science: the use of machine learning and AI to solve challenges in various scientific disciplines. They discuss and debate between data-driven and physics-driven approaches, the potential for foundational models, the importance of open sourcing models and data, the challenges of data sharing in science, and the ethical considerations of releasing powerful models. The conversation covers the role of academia, industry, and startups in driving innovation, with a focus on the field of AI. Professor Welling discusses the advantages and limitations of each sector and shares his experience in academia, big tech companies, and startups. The conversation then shifts to Professor Wellings new company; CuspAI, which focuses on material discovery for carbon capture using metal organic frameworks and machine learning. Prof. Welling provides insights into the potential applications of this technology and the importance of addressing sustainability challenges. The conversation concludes with a discussion on career advice and the future of AI for science.

Links

CuspAI : https://www.cusp.ai
University website: https://staff.fnwi.uva.nl/m.welling/
Google scholar: https://scholar.google.com/citations?user=8200InoAAAAJ&hl=en
AI4Science NeurIPS 2023 workshop: https://neurips.cc/virtual/2023/workshop/66548
AI4Science NeurIPS 2022 workshop: https://nips.cc/virtual/2022/workshop/50019
Aurora paper: https://arxiv.org/abs/2405.13063

Chapters

00:00 Introduction to the Neil Ashton Podcast
00:39 Guest Introduction: Professor Max Welling
11:12 Data-Driven vs. Physics-Driven Approaches in Machine Learning for Science
17:00 Foundational models for science
23:08 Discussion around Open-Sourcing Models and Data
29:26 Ethical Considerations in Releasing Powerful Models for Public Use
33:14 Collaboration and Shared Resources in Addressing Global Challenges
34:07 The Role of Academia, Industry, and Startups
43:27 Material Discovery for Carbon Capture
52:02 Career Advice for Early-stage Researchers
01:01:07 The Future of AI for Science and Sustainability

Keywords

AI for science, machine learning, data-driven approaches, physics-driven approaches, foundational models, open sourcing, data sharing, ethical considerations, blockchain technology, academia, industry, startups, AI, material discovery, carbon capture, metal organic frameworks, machine learning, sustainability, career advice, future of AI for science

This episode sets the scene for upcoming discussions on AI4Science with world renowned experts on machine learning. The focus is on using machine learning to solve scientific problems, such as computational fluid dynamics, weather modeling, material design, and drug discovery. The episode introduces the concept of machine learning and its potential to accelerate simulations and predictions. The episode also discusses the differences between machine learning for scientific problems and large language models, and the ongoing debate on incorporating physics into machine learning models.

Chapters
00:30 Introduction: AI for Science and Machine Learning
02:29 The Importance of Computational Fluid Dynamics
04:53 The Limitations of Physical Testing and Simulation
05:53 Accelerating Simulations and Predictions with Machine Learning
09:51 Data-Driven vs Physics-Informed Approaches in Machine Learning
13:10 The Future of Machine Learning in Science: Foundational Models

In this episode of the Neil Ashton podcast, Neil interviews Dr. Chris Rumsey, Research Scientist at NASA Langley Research Center. Chris is one of the main CFD experts at NASA Langley is globally reconised as a leader in CFD, particularly for aeronautical applications. The conversation focuses on computational fluid dynamics (CFD) and turbulence modeling. They discuss Chris's career, his role in public dissemination of CFD methods, and his involvement in the Turbulence Modeling website. They also explore the High Lift Prediction Workshop and the role of machine learning in CFD and turbulence modeling. The conversation provides insights into working at NASA and the challenges and advancements in CFD and turbulence modeling. In this conversation, Neil and Chris Rumsey discuss the progress and challenges in solving the problem of high-lift aerodynamics in aircraft design. They explore the concept of certification by analysis and the role of computational fluid dynamics (CFD) in reducing the need for expensive wind tunnel and flight tests. They also delve into the use of machine learning in CFD and the challenges of reproducibility. The conversation then shifts to conferences, with Neil and Chris sharing their experiences and favorite events. They conclude by discussing career advice for aspiring aerospace professionals and the unique aspects of working at NASA.

00:00 Introduction to the Neil Ashton podcast
01:09 Focus on Computational Fluid Dynamics and Turbulence Modeling
06:51 Chris Rumsey's Journey to NASA
09:13 From Art to Aeronautical Engineering
13:08 Transitioning to Turbulence Modeling
15:34 The Origins of the Turbulence Modeling Website
20:40 Verification and Validation in Turbulence Modeling
24:34 The Role of Machine Learning in Turbulence Modeling
26:00 Advancements in High Lift Prediction
27:28 Challenges in High Lift Prediction
28:25 Thoughts on Working at NASA
29:42 Certification by Analysis: Reducing the Cost of Aircraft Certification
31:09 The Role of Machine Learning in CFD and Certification by Analysis
34:03 The Value of Conferences in Networking and Specialized Learning
40:30 Career Advice for Aspiring Aerospace Professionals
48:45 Curating and Documenting Knowledge in the Aerospace Community

In this episode, Neil speaks to Professor Jack Dongarra, a renowned figure in the supercomputing and high-performance computing (HPC) world. He is a Professor at University of Tennessee as well as a Distinguished Researcher at Oak Ridge National Laboratory (ORNL) and a Turing Fellow at the University of Manchester. He is the inventor of the LINPACK library that is still used today to benchmark the Top 500 list of the most powerful supercomputers and was one of the key people involved in the creation of Message-Passing-Inferface (MPI). They discuss what is HPC, the challenges and opportunities in the field, and the future of HPC. They also touch on the role of machine learning and AI in HPC, the competitiveness of the United States in the field, and potential future technologies in HPC. Professor Dongarra shares his insights and advice based on his extensive experience in the field.

As part of their discussion they discuss two papers from Prof Dongarra:

1) High-Performance Computing: Challenges and Opportunities: https://arxiv.org/abs/2203.02544
2) Can the United States Maintain Its Leadership in High-Performance Computing? - A report from the ASCAC Subcommittee on American Competitiveness and Innovation to the ASCR Office: https://www.osti.gov/biblio/1989107/

Chapters

00:00 Introduction
04:18 Defining HPC and its Impact
08:11 Challenges and Opportunities in HPC
28:20 The Competitiveness of the United States in HPC
44:31 The Future of HPC: Technologies and Innovations
49:30 Insights and Advice from Professor Jack Dongarra

In this episode, Neil interviews Pat Symonds, one of the most well known and respected engineers in Formula One. They discuss Pat's career in engineering, his time in Formula One, and the evolution of the sport. Pat shares insights into his early motivations, his work with different teams, and the challenges he faced. They also touch on the growth of Motorsport Valley in the UK and the potential for Formula One teams to be based in other countries. In this conversation, Pat discusses his experience in Formula One and the challenges of being a technical director. He emphasizes the importance of continuous learning and the ability to make compromises in order to achieve success. He shares insights into the culture at Williams and Benetton and how it impacted their success. Additionally, he discusses the future of Formula One, including the use of AI and ML, the potential shift towards sustainable fuels, and the role of motor manufacturers.

In this episode I speak to Prof Juan J. Alonso on his vision of the future of computational science as well as his journey from academia to entrepreneurship - founding Luminary Cloud. He reflects on the revolutions in computational science and the different ways of developing software throughout his career. Alonso emphasizes the importance of academia in creating and perpetuating knowledge, as well as the value of innovation and new ideas. He also discusses the changes in the CFD world, the emergence of new technologies like GPU computing and cloud computing, and the potential for advancements in computational simulations for analysis and design. We also touch on the transition of the aerospace industry towards commercial software and the potential for cloud computing to revolutionize CFD. The conversation concludes with a discussion on the progress made towards achieving the goals outlined in the 2030 CFD vision report and the role of machine learning and AI in simulation-driven workflows.

In this final part of the conversation, Juan discusses the potential applications of ML and AI in engineering. He identifies four main areas where these technologies can be beneficial, but emphasizes that these applications will always be based on high-fidelity simulations. He concludes by envisioning the future of computational-driven science and the continued innovation in the field.

You can check out Luminary Cloud at https://www.luminarycloud.com and Prof Alonso's Stanford research at: https://adl.stanford.edu


06:00 Introduction and Background
09:11 Early Interest in Aerospace Engineering
12:13 From Academia to Industry
15:11 Decision to Stay in Academia
17:11 Balancing Fundamental Science and Applied Research
22:14 Early Aims and Focus on High Performance Computing
29:18 Emergence of GPU Computing and Cloud Computing
32:23 Conditions for Innovation and Entrepreneurship
35:01 The Importance of the Bay Area
35:37 Challenges and Requirements in Developing Solvers
41:00 The Role of the Bay Area in Attracting Computational Science Talent
44:16 The Difficulty and Respect for Building High-Quality Commercial Software
47:03 The Transition of the Aerospace Industry towards Commercial Software
49:30 The Potential of Cloud Computing in Revolutionizing CFD
53:59 Progress towards the Goals of the 2030 CFD Vision Report
01:00:53 The Role of Machine Learning and AI in Simulation-Driven Workflows
01:04:01 Applications of ML and AI in Engineering
01:05:36 Optimization and Design Optimization with ML and AI
01:06:04 Outer Loops and Uncertainty Quantification
01:07:04 Digital Twin Frameworks and Constant Retraining
01:12:36 The Value of Open-Source Codes in Academia
01:16:19 Challenges of Integrating Commercial Tools with Research
01:25:20 The Future of Computational-Driven Science
01:29:01 Continued Innovation and Replacement of Physical Experimentation

In this conversation, Neil interviews Dimitris Katsanis, one of the world leading experts in bike design. They discuss the UCI regulations that govern bike design for road and track racing. Dimitris explains the evolution of bike design and the role of carbon fiber and titanium in creating lightweight and aerodynamic bikes. He also talks about his collaboration with Pinarello and the development of the Dogma F8 and F10 bikes.

Dimitris emphasizes the importance of balancing weight, stiffness, and aerodynamics in bike design and the ongoing pursuit of improvement in the field. In this part of the conversation, Dimitris Katsanis discusses the evolution of bike design, the importance of aerodynamics and system drag reduction, the differences between track and road bike design, the interactions between the bike and rider, the impact of weight and aerodynamics in solo breakaways, the ongoing weight vs. aero debate, the role of stiffness in bike design, the relationship between stiffness and comfort in bike frames, and the potential of 3D printing and additive manufacturing in bike manufacturing.

In this conversation, we also discuss the limitations of carbon fiber in bike design and the potential of 3D printing to overcome these limitations. He explains how 3D printing allows for the creation of custom shapes and internal structures that can improve the performance and weight of bike components. Katsanis shares examples of 3D printed handlebars and frames that are lighter than their carbon fiber counterparts. He also discusses the future of mass customization in bike design and the impact of regulations on innovation.

Finally, he speculates on what bikes may look like in the future if design restrictions were lifted.

Chapters

06:40 Introduction and Background
11:10 UCI Regulations and Bike Design
17:48 Evolution of Bike Design and UCI Regulations
25:27 Influence of Weight and Aerodynamics on Bike Performance
32:01 Pushing the Limits of Aerodynamics
37:16 Yaw Sensitivity and Aerofoil Sections
40:53 Continual Improvement in Bike Design
42:25 The Evolution of Bike Design
42:51 Aerodynamics and System Drag Reduction
44:21 Track vs. Road Bike Design
47:05 Interactions Between Bike and Rider
48:02 The Importance of Aero in Solo Breakaways
53:00 Weight vs. Aero Debate
56:00 The Impact of Weight on Performance
58:04 The Role of Stiffness in Bike Design
01:04:01 Stiffness and Comfort in Bike Frames
01:11:56 Materials in Bike Design: Steel, Aluminum, Titanium, and Carbon Fiber
01:18:08 The Potential of 3D Printing and Additive Manufacturing
01:19:45 The Limitations of Carbon Fiber
01:21:41 The Potential of 3D Printing
01:24:10 The Surprising Lightness of 3D Printed Titanium
01:28:02 The Future of Mass Customization
01:34:06 The Impact of Regulations on Bike Design
01:43:09 Speculating on the Bike of the Future

Summary

In this episode, Neil discusses four key career questions that you should think about. He explores the pros and cons of pursuing a PhD, the path to becoming a professor, and the opportunities in the tech sector. He highlights the importance of gaining industry experience and the potential for higher salaries in the tech sector. Neil also mentions the option of dual positions, where academics work in both academia and industry. Overall, he encourages listeners to consider all the options and make informed decisions about their careers.

Takeaways

Doing a PhD can provide expertise and specialization in a specific area, but it may delay entry into the job market and result in lower initial salaries.
Becoming a professor requires a PhD and often involves postdoctoral research positions. Advancement to higher ranks, such as associate professor and full professor, requires publishing, securing funding, and taking on leadership roles.
The tech sector offers high-paying jobs and opportunities for engineers, particularly in areas like machine learning and data science. Tech companies value both academic and industry experience.
Consider the trade-offs between academia and industry, such as job security, work-life balance, and the level of freedom and autonomy.
Dual positions, where academics work in both academia and industry, are becoming more common and offer the best of both worlds.

Timestamps
00:00 Introduction
05:22 Question 1: PhD or no PhD
09:19 Question 2: How do I become a Professor?
23:10 Question 3: Academia or Industry?
31:00 Question 4: The third alternative - tech sector (Amazon, Google, META, Nvidia, Microsoft etc)
38:38 Dual Positions: Bridging the Gap Between Academia and Industry
41:00 Conclusions