Jay Shah Podcast Jay Shah

Swaroop is a research scientist at Google-Deepmind, working on improving Gemini. His research expertise includes instruction tuning and different prompt engineering techniques to improve reasoning and generalization performance in large language models (LLMs) and tackle induced biases in training. Before joining DeepMind, Swaroop graduated from Arizona State University, where his research focused on developing methods that allow models to learn new tasks from instructions. Swaroop has also interned at Microsoft, Allen AI, and Google, and his research on instruction tuning has been influential in the recent developments of LLMs.

Time stamps of the conversation:
00:00:50 Introduction
00:01:40 Entry point in AI
00:03:08 Motivation behind Instruction tuning in LLMs
00:08:40 Generalizing to unseen tasks
00:14:05 Prompt engineering vs. Instruction Tuning
00:18:42 Does prompt engineering induce bias?
00:21:25 Future of prompt engineering
00:27:48 Quality checks on Instruction tuning dataset
00:34:27 Future applications of LLMs
00:42:20 Trip planning using LLM
00:47:30 Scaling AI models vs making them efficient
00:52:05 Reasoning abilities of LLMs in mathematics
00:57:16 LLM-based approaches vs. traditional AI
01:00:46 Benefits of doing research internships in industry
01:06:15 Should I work on LLM-related research?
01:09:45 Narrowing down your research interest
01:13:05 Skills needed to be a researcher in industry
01:22:38 On publish or perish culture in AI research

More about Swaroop: https://swarooprm.github.io/
And his research works: https://scholar.google.com/citations?user=-7LK2SwAAAAJ&hl=en
Twitter: https://x.com/Swarooprm7

About the Host:
Jay is a PhD student at Arizona State University working on improving AI for medical diagnosis and prognosis.
Linkedin: https://www.linkedin.com/in/shahjay22/
Twitter: https://twitter.com/jaygshah22
Homepage: https://www.public.asu.edu/~jgshah1/ for any queries.

Stay tuned for upcoming webinars!

***Disclaimer: The information in this video represents the views and opinions of the speaker and does not necessarily represent the views or opinions of any institution. It does not constitute an endorsement by any Institution or its affiliates of such video content.***

Dr. Imon Banerjee is an Associate Professor at Mayo Clinic in Arizona, working at the intersection of AI and healthcare research. Her research focuses on multi-modality fusion, mitigating bias in AI models specifically in the context of medical applications & more broadly building predictive models using different data sources. Before joining the Mayo Clinic, she was at Emory University as an Assistant Professor and at Stanford as a Postdoctoral fellow.

Time stamps of the conversation
00:00 Highlights
01:00 Introduction
01:50 Entry point in AI
04:41 Landscape of AI in healthcare so far
06:15 Research to practice
07:50 Challenges of AI Democratization
11:56 Era of Generative AI in Medical Research
15:57 Responsibilities to realize
16:40 Are LLMs a world model?
17:50 Training on medical data
19:55 AI as a tool in clinical workflows
23:36 Scientific discovery in medicine
27:08 Dangers of biased AI models in healthcare applications
28:40 Good vs Bad bias
33:33 Scaling models - the current trend in AI research
35:05 Current focus of research
36:41 Advice on getting started
39:46 Interdisciplinary efforts for efficiency
42:22 Personalities for getting into research

More about Dr. Banerjee's lab and research: https://labs.engineering.asu.edu/banerjeelab/person/imon-banerjee/

About the Host:
Jay is a PhD student at Arizona State University.
Linkedin: https://www.linkedin.com/in/shahjay22/
Twitter: https://twitter.com/jaygshah22
Homepage: https://www.public.asu.edu/~jgshah1/ for any queries.

Stay tuned for upcoming webinars!

***Disclaimer: The information contained in this video represents the views and opinions of the speaker and does not necessarily represent the views or opinions of any institution. It does not constitute an endorsement by any Institution or its affiliates of such video content.***

Dr. Petar Veličković is a Staff Research Scientist at Googe DeepMind and an Affiliated lecturer at the University of Cambridge. He is known for his research contributions in graph representation learning; particularly graph neural networks and graph attention networks. At DeepMind, he has been working on Neural Algorithmic Reasoning which we talk about more in this podcast. Petar’s research has been featured in numerous media articles and has been impactful in many ways including Google Maps’s improved predictions.

Time stamps
00:00:00 Highlights
00:01:00 Introduction
00:01:50 Entry point in AI
00:03:44 Idea of Graph Attention Networks
00:06:50 Towards AGI
00:09:58 Attention in Deep learning
00:13:15 Attention vs Convolutions
00:20:20 Neural Algorithmic Reasoning (NAR)
00:25:40 End-to-end learning vs NAR
00:30:40 Improving Google Map predictions
00:34:08 Interpretability
00:41:28 Working at Google DeepMind
00:47:25 Fundamental vs Applied side of research
00:50:58 Industry vs Academia in AI Research
00:54:25 Tips to young researchers
01:05:55 Is a PhD required for AI research?

More about Petar: https://petar-v.com/
Graph Attention Networks: https://arxiv.org/abs/1710.10903
Neural Algorithmic Reasoning: https://www.cell.com/patterns/pdf/S2666-3899(21)00099-4.pdf
TacticAI paper: https://arxiv.org/abs/2310.10553
And his collection of invited talks:  @petarvelickovic6033 

About the Host:
Jay is a PhD student at Arizona State University.
Linkedin: https://www.linkedin.com/in/shahjay22/
Twitter: https://twitter.com/jaygshah22
Homepage: https://www.public.asu.edu/~jgshah1/ for any queries.

Stay tuned for upcoming webinars!

***Disclaimer: The information contained in this video represents the views and opinions of the speaker and does not necessarily represent the views or opinions of any institution. It does not constitute an endorsement by any Institution or its affiliates of such video content.***

Dr. Yezhou Yang is an Associate Professor at Arizona State University and director of the Active Perception Group at ASU. He has research interests in Cognitive Robotics and Computer Vision, and understanding human actions from visual input and grounding them by natural language. Prior to joining ASU, he completed his Ph.D. from the University of Maryland and his postdoctoral at the Computer Vision Lab and Perception and Robotics Lab.

Timestamps of the conversation
00:01:02 Introduction
00:01:46 Interest in AI
00:17:04 Entry in Robotics & AI Perception
00:20:59 Combining Vision & language to Improve Robot Perception
00:23:30 End-to-end learning vs traditional knowledge graphs
00:28:28 What do LLMs learn?
00:30:30 Nature of AI research
00:36:00 Why vision & language in AI?
00:45:40 Learning vs Reasoning in neural networks
00:53:05 Bringing AI to the general crowd
01:00:10 Transformers in Vision
01:08:54 Democratization of AI
01:13:42 Motivation for research: theory or application?
01:18:50 Surpassing human intelligence
01:25:13 Open challenges in computer vision research
01:30:19 Doing research is a privilege
01:35:00 Rejections, tips to read & write good papers
01:43:37 Tips for AI Enthusiasts
01:47:35 What is a good research problem?
01:50:30 Dos and Don'ts in AI research

More about Dr. Yang: https://yezhouyang.engineering.asu.edu/
And his Twitter handle: https://twitter.com/Yezhou_Yang

About the Host:
Jay is a PhD student at Arizona State University.
Linkedin: https://www.linkedin.com/in/shahjay22/
Twitter: https://twitter.com/jaygshah22
Homepage: https://www.public.asu.edu/~jgshah1/ for any queries.

Check-out Rora: https://teamrora.com/jayshah
Guide to STEM PhD AI Researcher + Research Scientist pay: https://www.teamrora.com/post/ai-researchers-salary-negotiation-report-2023

Stay tuned for upcoming webinars!

***Disclaimer: The information contained in this video represents the views and opinions of the speaker and does not necessarily represent the views or opinions of any institution. It does not constitute an endorsement by any Institution or its affiliates of such video content.***

Dr Hyrum Anderson is a Distinguished Machine Learning Engineer at Robust Intelligence. Prior to that, he was Principal Architect of Trustworthy Machine Learning at Microsoft where he also founded Microsoft’s AI Red Team; he also led security research at MIT Lincoln Laboratory, Sandia National Laboratories, and Mendiant, and was Chief Scientist at Endgame (later acquired by Elastic). He’s also the co-author of the book “Not a Bug, But with a Sticker” and his research interests include assessing the security and privacy of ML systems and building Robust AI models.


Timestamps of the conversation
00:50 Introduction
01:40 Background in AI and ML security
04:45 Attacks on ML systems
08:20 Fractions of ML systems prone to Attacks
10:38 Operational risks with security measures
13:40 Solution from an algorithmic or policy perspective
15:46 AI regulation and policy making
22:40 Co-development of AI and security measures
24:06 Risks of Generative AI and Mitigation
27:45 Influencing an AI model
30:08 Prompt stealing on ChatGPT
33:50 Microsoft AI Red Team
38:46 Managing risks
39:41 Government Regulations
43:04 What to expect from the Book
46:40 Black in AI & Bountiful Children’s Foundation

Check out Rora: https://teamrora.com/jayshah
Guide to STEM Ph.D. AI Researcher + Research Scientist pay: https://www.teamrora.com/post/ai-researchers-salary-negotiation-report-2023
Rora's negotiation philosophy:
https://www.teamrora.com/post/the-biggest-misconception-about-negotiating-salaryhttps://www.teamrora.com/post/job-offer-negotiation-lies

Hyrum's Linkedin: https://www.linkedin.com/in/hyrumanderson/
And Research: https://scholar.google.com/citations?user=pP6yo9EAAAAJ&hl=en
Book - Not a Bug, But with a Sticker: https://www.amazon.com/Not-Bug-But-Sticker-Learning/dp/1119883989/

About the Host:
Jay is a Ph.D. student at Arizona State University.
Linkedin: https://www.linkedin.com/in/shahjay22/
Twitter: https://twitter.com/jaygshah22
Homepage: https://www.public.asu.edu/~jgshah1/ for any queries.

Stay tuned for upcoming webinars!

***Disclaimer: The information contained in this video represents the views and opinions of the speaker and does not necessarily represent the views or opinions of any institution. It does not constitute an endorsement by any Institution or its affiliates of such video content.***

Meredith is an associate professor at New York University and research director at the NYU Alliance for Public Interest Technology. Her research interests include using data analysis for good and ethical AI. She is also the author of the book “More Than a Glitch: Confronting Race, Gender, and Ability Bias in Tech” and we will discuss more about this with her in this podcast.

Time stamps of the conversation
00:42 Introduction
01:17 Background
02:17 Meaning of “it is not a glitch” in the book title
04:40 How are biases coded into AI systems?
08:45 AI is not the solution to every problem
09:55 Algorithm Auditing
11:57 Why do organizations don't use algorithmic auditing more often?
15:12 Techno-chauvinism and drawing boundaries
23:18 Bias issues with ChatGPT and Auditing the model
27:55 Using AI for Public Good - AI on context
31:52 Advice to young researchers in AI

Meredith's homepage: https://meredithbroussard.com/
And her Book: https://mitpress.mit.edu/9780262047654/more-than-a-glitch/

About the Host:
Jay is a Ph.D. student at Arizona State University.
Linkedin: https://www.linkedin.com/in/shahjay22/
Twitter: https://twitter.com/jaygshah22
Homepage: https://www.public.asu.edu/~jgshah1/ for any queries.

Stay tuned for upcoming webinars!

***Disclaimer: The information contained in this video represents the views and opinions of the speaker and does not necessarily represent the views or opinions of any institution. It does not constitute an endorsement by any Institution or its affiliates of such video content.***