The Transformative Ideas Podcast

Manuel Brenner
  • 0,0 (0)
  • NATURWISSENSCHAFTEN

The Transformative Ideas Podcast (formerly the ACIT Science Podcast). This podcast is to provide a glimpse into the life of scientists: to learn about the ideas they are passionate about, to find out what gets them out of bed every day to face the challenges and frustrations of working at the frontier to the unknown, and to share in some of the most important lessons they have learned in their career. For inquiries, reach out to manu.brenn@gmail.com.

  1. #35: Bridging Modalities: Vision, Language, and AI with Letitia Parcalabescu

    07.08.2023

    #35: Bridging Modalities: Vision, Language, and AI with Letitia Parcalabescu

    In this epsiode, we are joined by Letitia Parcalabescu, PhD candidate at Heidelberg University's Department of Computational Linguistics, to share her insights on the fascinating world of Multimodal Learning. As a researcher and science communicator, Letitia has been thinking about the intersection between vision and text, a frontier of machine learning that has seen immense growth in recent years, for several years. We explore her journey from physics to machine learning, unpack the influence of large language models (LLMs) on our understanding of linguistics, and delve into the relevance of vision and language interplay in machine learning. We discuss the key developments in multimodal learning, including joint embeddings, diffusion models, and LLMs, and shares her perspective on how these advancements relate to Artificial General Intelligence (AGI). Alongside her research, we discuss the value of benchmarks and performance metrics in machine learning, as well as her own research projects. Letitia offers a glimpse into a typical research day in her field, and shares her motivations and learnings from her successful YouTube channel, AI Coffee Break: https://youtube.com/@AICoffeeBreak

    1 Std. 15 Min.
  2. #34: The Biology of Aging: From Mechanisms to Interventions with Christoph Benner

    26.04.2023

    #34: The Biology of Aging: From Mechanisms to Interventions with Christoph Benner

    In this episode, we are joined by Christoph Benner. Christoph Benner received his PhD from ETH Zurich, studying aging in model organism. We discuss the idea of treating aging as a disease, theories of aging, and some of its hallmarks. We cover the fascinating role mitochondria plays in aging and metabolism, and get into the weeds of the complex underlying biochemistry. We also discuss pragmatic tools and interventions, from most promising drugs to behavioral measures, such as intermittent, and how much of it boils down to common shared mechanism between many species. We close with discussions on the scientific process, the importance of philosophy for science. For inquiries, reach out to manu.brenn@gmail.com.

    1 Std. 23 Min.
  3. #33: AI for Climate Change with Priya Donti

    02.01.2023

    #33: AI for Climate Change with Priya Donti

    In this episode, we are joined by Priya Donti. Priya is the Co-founder and Executive Director of Climate Change AI (CCAI), a global non-profit initiative to catalyze impactful work at the intersection of climate change and machine learning, and an incoming assistant professor at MIT. We discuss the influential "Tackling Climate Change with Machine Learning" paper, opportunities, costs and risks of AI for climate change, climate justice, Priya's research on power grid optimization, implicit layers, the work and funding of Climate Change AI, and many more topics. For inquiries, write a mail to manu.brenn@gmail.com or reach out on Linkedin.

    48 Min.
  4. #32: Transformers, Puns and Rap Battles: A Conversation With ChatGPT

    16.12.2022

    #32: Transformers, Puns and Rap Battles: A Conversation With ChatGPT

    In this episode, we are joined by our first artificial guest: OpenAI's new large language model based ChatGPT. After discussing large language models in several of the latest episodes and OpenAI releasing ChatGPT two weeks ago, it felt like the perfect time for this experiment. I tried showcasing some of  ChatGPT different talents, from giving detailed essay-like scientific explanations to speaking other languages to making puns and improvising rap battles. To make the episode more entertaining, I adjusted some of the prompts and cut some of the responses. However, all of the responses are 100% real. ChatGPT's text output was then transformed via Polly from AWS to spoken word, which was further edited to make it sound more realistic. The thumbnail portrait was also generated artificially, using a GAN based on StyleGAN2. The recent advances in AI remain simultaneously impressive and slightly disconcerting. For inquiries, reach out to manu.brenn@gmail.com

    16 Min.

  5. 27.11.2022

    #31: Deep Metric Learning, Generalization and Transformers with Karsten Roth

    We are now the Transformative Ideas Podcast (formerly the ACIT Science Podcast)! In this episode, we are joined by Karsten Roth. Karsten is a PhD Student at the Explainable Machine Learning group in Tuebingen, supervised by Zeynep Akata & Oriol Vinyals (https://karroth.com/). His interests lie in most things Machine Learning and AI related. We discuss ideas surrounding AI for medical applications, deep metric learning, generalization and representation learning, the surprising success of transformers, diffusion models, large language models and joint representations, generalization, inductive biases and neuroplasticity, AI research and bad incentives in the publishing culture, ML conferences, the interplay of industry and academia, advice for how to be a successful researchers, physics as a foundation for ML, and much more. For inquiries, contact manu.brenn@gmail.com or reach out to me on Linkedin (https://www.linkedin.com/in/manuel-brenner-772261191/).

    2 Std. 23 Min.

  6. 01.09.2022

    #30: Language Models, Active Inference and Latent Variables with Alejandro Daniel Noel

    In this episode, our host Manuel Brenner is joined by Alejandro Daniel Noel. Alejandro got his masters from TU Delft working with Acitive Inference and the Free Energy Principle, and is now a full time software engineer at Google in Zürich, specializing on machine learning engineering of language models for conversational AI. We discuss the free energy principle, active inference, the role of uncertainty, latent variable models, language and tokenization, causality, diffusion models, DALL-E, paths towards AGI, JEPA, language and consciousness, attention, self-attention, and many more topics.  Stay tuned for more episodes: https://mailchi.mp/0346443b6ddf/acit-global-signup

    1 Std. 43 Min.
  7. #29: Memory, Dynamics and AI Ethics with Leonard Bereska

    20.07.2022

    #29: Memory, Dynamics and AI Ethics with Leonard Bereska

    In this episode, host Manuel Brenner is joined by Leonard Bereska. Leonard Bereska is a PhD Candidate at University of Amsterdam, studying long term memory. We discuss a dynamical systems perspective on neuroscience, how machine learning comes into play, and how this is connected to memory and models of memory in machine learning, AI ethics, consciousness, the role of noise, the connections between Bayesian models, ensemble methods and regularizations, companies and states as superhuman intelligences, dopamine, and many more topics. Find Leonard's recent publications: https://arxiv.org/abs/2206.13336 https://arxiv.org/abs/2207.02542 Stay tuned for more episodes: https://mailchi.mp/0346443b6ddf/acit-global-signup

    1 Std. 50 Min.
  8. #28: Blockchains and Cryptocurrencies with Ingo Fiedler

    05.06.2022

    #28: Blockchains and Cryptocurrencies with Ingo Fiedler

    Note: Audio quality will improve around the 12 minute mark. In this episode, Ingo Fiedler from the Blockchain Research Lab joins Manuel Brenner for a conversation about blockchains and cryptocurrencies. Ingo Fiedler is Affiliate Professor at Concordia University, Montreal,  and co-founder of the Blockchain Research Lab, a non-profit dedicated to Scientific Research on Blockchain Technology for the Benefit of Society. We discuss blockchains, Bitcoin and Ethereum, proof of work and proof of stake, the current crash of Terra and Luna, key concepts behind algorithmic stablecoins, why the holy grail of a functioning stablecoin is so important, humanitarian aspects of cryptocurrencies and the increasing role they play in developing countries, central banking, fiat currencies and inflation, the work of the Blockchain Research Lab, NFTs, self-sovereign identities, the metaverse, energy consumption of cryptos, how cryptos could help fund renewable energy plants and many more topics. The blockchain research lab: https://www.blockchainresearchlab.org/ ACIT Global: https://acit-science.com/

    1 Std. 46 Min.
Alle anzeigen (36)

Info

The Transformative Ideas Podcast (formerly the ACIT Science Podcast). This podcast is to provide a glimpse into the life of scientists: to learn about the ideas they are passionate about, to find out what gets them out of bed every day to face the challenges and frustrations of working at the frontier to the unknown, and to share in some of the most important lessons they have learned in their career. For inquiries, reach out to manu.brenn@gmail.com.

Informationen