Rethinking LLMs: Why Diffusion Could Change How We Build AI | Dr. Subham Sahoo
Subham Sahoo is a Research Scientist and Team Lead at the Institute of Foundation Models at MBZUAI, where he leads the Diffusion LLM team. His research focuses on discrete diffusion language models and the question of whether diffusion can offer a fundamentally different, and potentially more efficient way of generating language compared with traditional autoregressive models.Prior to that, Subham completed his Ph.D. at Cornell Tech. And his recent work includes studies of how diffusion models could be scaled and evaluated against autoregressive models. His work has helped push diffusion LMs from an interesting alternative toward a much more serious research direction for language generation.00:00:00 Highlight & Introduction00:02:10 Journey into AI & First Interest in Diffusion Models00:11:37 Learning the Mathematics Behind Diffusion Models00:16:58 Research, Publishing Pressure & Finding the Right Validation00:20:31 How to Filter Noise from AI Research00:23:32 Academia vs. Industry After a PhD00:26:32 Diffusion Models vs. Autoregressive LLMs00:29:01 How Blockwise Diffusion Solves the Latency Problem00:30:57 Why Autoregressive LLMs Still Have Room for Improvement00:38:00 How Hybrid Diffusion + Autoregressive Models Work00:41:50 Can Blockwise Prediction Preserve Model Quality?00:44:03 Will Diffusion Replace Autoregressive LLMs?00:47:52 Does Diffusion Training Make LLMs Learn Gibberish?00:50:45 Comparing Training Costs & Scaling00:55:00 How Hybrid Models Scale with More Data00:59:02 Lossy vs. Lossless Speedups01:02:09 Are Hybrid Diffusion Models Production Ready?01:04:54 How to Choose a Good AI Research Direction01:08:06 Advice for Early-Career AI Researchers01:11:39 Closing ThoughtsAbout the Host:Jay is a Machine Learning Engineer III at PathAI working on improving AI for medical diagnosis and prognosis. Linkedin: https://www.linkedin.com/in/shahjay22/Twitter: https://twitter.com/jaygshah22Homepage: https://jaygshah.github.io/ 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.***
27 Jul 2026
Why Today's LLMs Still Don't Understand Culture | Simran Khanuja
Simran Khanuja is a Ph.D. candidate at Carnegie Mellon University. Her research focuses on making AI systems more useful for people across different languages and cultures, with an emphasis on multilingual and multimodal foundation models.Prior to CMU, Simran spent over two years as a Pre-Doctoral Researcher at Google Research, contributing to multilingual language technologies and also did some interesting research at Microsoft Research India. Her recent work includes the Pangea multilingual multimodal foundation model, image transcreation for cultural localization, and new approaches to evaluating culturally relevant AI systems.00:00:00 Highlight & Introduction00:01:20 Journey into AI00:06:27 Industry Experience before Ph.D.00:12:38 Why Pursue a Ph.D. in AI?00:17:54 From Multilingual to Multicultural AI00:24:11 Are Foundation Models really general purpose?00:29:51 Should every Culture have tts own AI model?00:32:38 The biggest bottleneck in cultural alignment00:38:49 Can Prompting solve cultural alignment?00:44:41 Why Images need Translation too00:51:14 Beyond Text: Translating culture across modalities00:55:01 How do you evaluate Image Transcreation?01:00:48 Preventing catastrophic forgetting in Multilingual LLMs01:04:50 Can LLMs learn undocumented cultures?01:11:05 Are LLMs reliable as AI Judges?01:19:32 Can creativity ever be benchmarked?01:28:20 How AI Has changed research workflows01:41:12 Choosing an AI research topic in 202601:52:12 The most overlooked problem in AI today02:01:28 Closing thoughtsMore about Simran's ongoing research: https://simran-khanuja.github.io/About the Host:Jay is a Machine Learning Engineer III at PathAI working on improving AI for medical diagnosis and prognosis. Linkedin: https://www.linkedin.com/in/shahjay22/Twitter: https://twitter.com/jaygshah22Homepage: https://jaygshah.github.io/ 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.***
18 Feb 2026
The Hidden Flaws in AI Safety & Evaluation Benchmarks | Prof. Jackie Chi Kit Cheung
Dr. Jackie Cheung is an Associate Professor at McGill University where he co-directs the Reasoning and Learning Lab. He is also an Associate Scientific Director at Mila-Quebec Artificial Intelligence Institute. He and his team are developing computational models to improve the reliability, pragmatics, and evaluation of large language models to ensure they are contextually appropriate and factually grounded.Jackie was worked as a consultant researcher with Microsoft Research and before his current appointments, he earned his PhD and MSc in Computer Science from the University of Toronto, focusing on computational linguistics, and his BSc from the University of British Columbia.00:00:00 Highlight & Introduction00:02:04 Entrypoint in AI & NLP00:04:47 Academia vs. Industry: Career choices00:09:48 Language Revitalization using AI00:12:24 Addressing Biases & Data sovereignty in language revitalization 00:15:49 Evaluating LLMs as Judges00:17:14 Validity and reliability in LLM evaluation 00:25:11 Evidence-centered benchmark design (ECBD) framework00:30:38 Gaps in LLM benchmarks and meaning of "general purpose" AI00:35:24 General purpose intelligence vs reasoning00:40:16 Safety as an undefined bundle in LLMs00:51:45 Stochastic chameleons: how LLMs generalize and hallucinate 01:03:02 Potential & Biases of agentic frameworks for research01:05:52 Evaluating LLMs for summarization01:11:43 Scaling large language models01:16:33 Advice to beginners entering AI in 202601:20:33 Pitfalls to avoid in AI research & development More about Jackie & his research: https://www.cs.mcgill.ca/~jcheung/About the Host:Jay is a Machine Learning Engineer III at PathAI working on improving AI for medical diagnosis and prognosis. Linkedin: https://www.linkedin.com/in/shahjay22/Twitter: https://twitter.com/jaygshah22Homepage: https://jaygshah.github.io/ 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.***
Reach and audience
Public platform figures. Ratings count people who left a rating, not total listeners.
YouTube views
271,636
Score snapshot 15 Sept 2025
Apple Podcasts (US)
5.0 / 5
14 ratings
Spotify
5.0 / 5
5 ratings
Podcast Authority Score: 57 / 100
A composite of feed quality, social presence, YouTube performance and engagement. Read the methodology.
Quality
60
Social presence
0
YouTube
86
Engagement
51
Contact Jay Shah Podcast
Guest appearances
Does not typically book guests
Based on episode analysis; this does not confirm that the show is currently accepting guests.
Host of Jay Shah Podcast?
Claim your podcast to manage its listing and keep your show details accurate.
Pod Engine is an independent podcast discovery and analytics service and is not affiliated with or endorsed by this podcast. Artwork and show content belong to their owners. Full legal notice.
Explore this show Podcast research with Pod Engine
