#twiml — Public Fediverse posts

First was a great conversation with Luke Zettlemoyer on scaling multi-modal generative AI on the #TWIML podcast. Zettlemoyer discussed some impressive advances on using transformer models for images, the importance of data and the limits of current scaling approaches, and why openness is essential for progress to continue in AI. Highly recommend https://www.youtube.com/watch?v=ZvfLYcG_8io&t=4s (2/8) #AI #GenerativeAI

First was a great conversation with Luke Zettlemoyer on scaling multi-modal generative AI on the #TWIML podcast. Zettlemoyer discussed some impressive advances on using transformer models for images, the importance of data and the limits of current scaling approaches, and why openness is essential for progress to continue in AI. Highly recommend https://www.youtube.com/watch?v=ZvfLYcG_8io&t=4s (2/8) #AI #GenerativeAI

Next was a fabulous talk by @alex on #AI hype on the #TWIML podcast. Hanna reviews what drives AI hype, where AI can be useful, and DAIR's broader research agenda. Highly recommend https://www.youtube.com/watch?v=I4bdtG_SagE (4/9)

Next was a fabulous talk by @alex on #AI hype on the #TWIML podcast. Hanna reviews what drives AI hype, where AI can be useful, and DAIR's broader research agenda. Highly recommend https://www.youtube.com/watch?v=I4bdtG_SagE (4/9)

Next was a fantastic conversation with James Zou looking at changes in ChatGPT's performance over the last few months on the #TWIML podcast. Zou convincingly demonstrates that #ChatGPT has gotten decisively worse on a number of metrics recently, and highlights some of the dangers of relying on #LLMs as stable arbiters of high-quality output. Highly recommend https://www.youtube.com/watch?v=2HXL89bqxx0 (3/7) #AI

Next was a fantastic conversation with James Zou looking at changes in ChatGPT's performance over the last few months on the #TWIML podcast. Zou convincingly demonstrates that #ChatGPT has gotten decisively worse on a number of metrics recently, and highlights some of the dangers of relying on #LLMs as stable arbiters of high-quality output. Highly recommend https://www.youtube.com/watch?v=2HXL89bqxx0 (3/7) #AI

Next was an engaging conversation with Sophia Sanborn on neural networks, Fourier transform properties, and more on the #TWIML podcast https://www.youtube.com/watch?v=fK0kVYb32DA&t=4s (3/9) #AI

Next was an engaging conversation with Sophia Sanborn on neural networks, Fourier transform properties, and more on the #TWIML podcast https://www.youtube.com/watch?v=fK0kVYb32DA&t=4s (3/9) #AI

First was a thought-provoking conversation with @gokul on inverse reinforcement learning without RL on the #TWIML podcast. I spent a lot of time after this thinking about how these techniques could be applied to LLM training as a step before RLHF to significantly reduce toxic content exposure, and I'm very excited to see where this work goes. Highly recommend https://www.youtube.com/watch?v=SPZLJwDNwpM (2/7) #AI

First was a thought-provoking conversation with @gokul on inverse reinforcement learning without RL on the #TWIML podcast. I spent a lot of time after this thinking about how these techniques could be applied to LLM training as a step before RLHF to significantly reduce toxic content exposure, and I'm very excited to see where this work goes. Highly recommend https://www.youtube.com/watch?v=SPZLJwDNwpM (2/7) #AI

Next was a fantastic conversation with Alice Xiang on #privacy vs. fairness in #ComputerVision on the #TWIML podcast. Xiang not only sets the table with a great summary of the historical arc of #AI #ethics, but explicitly examines the tension between individual privacy and algorithmic harms that is often ignored. Highly recommend https://www.youtube.com/watch?v=Ylu416ksuG8 (4/5) #AIEthics

Next was a fantastic conversation with Alice Xiang on #privacy vs. fairness in #ComputerVision on the #TWIML podcast. Xiang not only sets the table with a great summary of the historical arc of #AI #ethics, but explicitly examines the tension between individual privacy and algorithmic harms that is often ignored. Highly recommend https://www.youtube.com/watch?v=Ylu416ksuG8 (4/5) #AIEthics

Next was a great conversation with Mohit Bansal on unification in large #AI models on the #TWIML podcast. Bansal has a fabulous perspective on the state and trajectory of the field and discusses some impressive models - I was partial to applying multi-modal large models to spectrograms. Highly recommend https://www.youtube.com/watch?v=vVlRWjdwp4s (6/10) #MachineLearning #LLMs #GenerativeAI

Next was a great conversation with Mohit Bansal on unification in large #AI models on the #TWIML podcast. Bansal has a fabulous perspective on the state and trajectory of the field and discusses some impressive models - I was partial to applying multi-modal large models to spectrograms. Highly recommend https://www.youtube.com/watch?v=vVlRWjdwp4s (6/10) #MachineLearning #LLMs #GenerativeAI

Next was a rich conversation with Hugo Larochelle on transfer learning on the #TWIML podcast. As companies and researchers try to apply general models to applications that they weren't originally designed for, understanding how to retrofit these models so they can perform well without extensive retraining is essential, and examined in depth here https://www.youtube.com/watch?v=c--acLK_C9s (4/7)

Next was a rich conversation with Hugo Larochelle on transfer learning on the #TWIML podcast. As companies and researchers try to apply general models to applications that they weren't originally designed for, understanding how to retrofit these models so they can perform well without extensive retraining is essential, and examined in depth here https://www.youtube.com/watch?v=c--acLK_C9s (4/7)

Next was a nice conversation with @DanFu on methods for expanding LLMs' context length on the #TWIML podcast. Fu incorporates a number of methods from other parts of computer science to significantly improve tractability for certain LLM tasks, and I'm guessing techniques like this will become standard quickly https://www.youtube.com/watch?v=CkWX1zNQ7uU (3/7) #LLMs

Next was a nice conversation with @DanFu on methods for expanding LLMs' context length on the #TWIML podcast. Fu incorporates a number of methods from other parts of computer science to significantly improve tractability for certain LLM tasks, and I'm guessing techniques like this will become standard quickly https://www.youtube.com/watch?v=CkWX1zNQ7uU (3/7) #LLMs

Anyone listened to the latest #TWIML with Ben Goertzel (SingularityNET)? This kids, is why you stay away from all the #drugz. The 1/2 is a pretty interesting discussion of symbolic vs connectionist models & how we maybe get to models of cognition…and it then goes off the f’n rails with a wild 30 mins on how #AGI leads to a world of #UniversalBasicIncome, #DiamondAge-esque #MatterCompilers, and #AI #utopia because …he's never met actual humans & what we do to each other?🤯

https://podcasts.apple.com/us/podcast/the-twiml-ai-podcast-formerly-this-week-in-machine/id1116303051?i=1000609360248

Anyone listened to the latest #TWIML with Ben Goertzel (SingularityNET)? This kids, is why you stay away from all the #drugz. The 1/2 is a pretty interesting discussion of symbolic vs connectionist models & how we maybe get to models of cognition…and it then goes off the f’n rails with a wild 30 mins on how #AGI leads to a world of #UniversalBasicIncome, #DiamondAge-esque #MatterCompilers, and #AI #utopia because …he's never met actual humans & what we do to each other?🤯

https://podcasts.apple.com/us/podcast/the-twiml-ai-podcast-formerly-this-week-in-machine/id1116303051?i=1000609360248

Anyone listened to the latest #TWIML with Ben Goertzel (SingularityNET)? This kids, is why you stay away from all the #drugz. The 1/2 is a pretty interesting discussion of symbolic vs connectionist models & how we maybe get to models of cognition…and it then goes off the f’n rails with a wild 30 mins on how #AGI leads to a world of #UniversalBasicIncome, #DiamondAge-esque #MatterCompilers, and #AI #utopia because …he's never met actual humans & what we do to each other?🤯

https://podcasts.apple.com/us/podcast/the-twiml-ai-podcast-formerly-this-week-in-machine/id1116303051?i=1000609360248

Anyone listened to the latest #TWIML with Ben Goertzel (SingularityNET)? This kids, is why you stay away from all the #drugz. The 1/2 is a pretty interesting discussion of symbolic vs connectionist models & how we maybe get to models of cognition…and it then goes off the f’n rails with a wild 30 mins on how #AGI leads to a world of #UniversalBasicIncome, #DiamondAge-esque #MatterCompilers, and #AI #utopia because …he's never met actual humans & what we do to each other?🤯

https://podcasts.apple.com/us/podcast/the-twiml-ai-podcast-formerly-this-week-in-machine/id1116303051?i=1000609360248

Anyone listened to the latest #TWIML with Ben Goertzel (SingularityNET)? This kids, is why you stay away from all the #drugz. The 1/2 is a pretty interesting discussion of symbolic vs connectionist models & how we maybe get to models of cognition…and it then goes off the f’n rails with a wild 30 mins on how #AGI leads to a world of #UniversalBasicIncome, #DiamondAge-esque #MatterCompilers, and #AI #utopia because …he's never met actual humans & what we do to each other?🤯

https://podcasts.apple.com/us/podcast/the-twiml-ai-podcast-formerly-this-week-in-machine/id1116303051?i=1000609360248

Last was an enjoyable discussion with Dhruv Batra on building maps and spatial awareness in blind #AI agents on the #TWIML podcast. Batra's work here is interesting - rather than feed #robots map-based models, what if they emerged from blindly navigating with memories? https://www.youtube.com/watch?v=vypBAQ2kKno (8/8)

Last was an enjoyable discussion with Dhruv Batra on building maps and spatial awareness in blind #AI agents on the #TWIML podcast. Batra's work here is interesting - rather than feed #robots map-based models, what if they emerged from blindly navigating with memories? https://www.youtube.com/watch?v=vypBAQ2kKno (8/8)

Next was a great conversation with Christos Louizos on automatically optimizing neural network hyperparameters on the #TWIML podcast. Hyperparameter setting is one of the dark arts of ML, so it's nice to see methods that start to simplify this part of the process https://www.youtube.com/watch?v=FF3wSAgw0Lc (4/11) #MachineLearning #AI

Next was a great conversation with Christos Louizos on automatically optimizing neural network hyperparameters on the #TWIML podcast. Hyperparameter setting is one of the dark arts of ML, so it's nice to see methods that start to simplify this part of the process https://www.youtube.com/watch?v=FF3wSAgw0Lc (4/11) #MachineLearning #AI

Last was a fantastic conversation with Marti Hearst on the hype vs. reality of LLMs on the #TWIML podcast. Hearst does a great job laying out the scientific state of LLMs and not taking the bait on inaccurate, hyped takes on the topic. Highly recommend https://www.youtube.com/watch?v=NdEpYO9J_e0 (5/5) #LLMs

Last was a fantastic conversation with Marti Hearst on the hype vs. reality of LLMs on the #TWIML podcast. Hearst does a great job laying out the scientific state of LLMs and not taking the bait on inaccurate, hyped takes on the topic. Highly recommend https://www.youtube.com/watch?v=NdEpYO9J_e0 (5/5) #LLMs

Next was an excellent discussion with @tomgoldstein on watermarking #LLMs on the #TWIML podcast. Goldstein presents an elegant method for watermarking LLM text to make it detectable, and hopefully this is widely deployed. Highly recommend https://www.youtube.com/watch?v=QC6wXsusHjk (4/12)

Next was an excellent discussion with @tomgoldstein on watermarking #LLMs on the #TWIML podcast. Goldstein presents an elegant method for watermarking LLM text to make it detectable, and hopefully this is widely deployed. Highly recommend https://www.youtube.com/watch?v=QC6wXsusHjk (4/12)

Next was a provacative discussion by @neuranna on the capabilities of #LLMs on the #TWIML podcast. While I didn't love the framing of the discussion, I loved Ivanova's aproach to developing methods that can accurately probe the capabilities of LLMs https://lnkd.in/ePdFj5zB (4/5)

Next was a provacative discussion by @neuranna on the capabilities of #LLMs on the #TWIML podcast. While I didn't love the framing of the discussion, I loved Ivanova's aproach to developing methods that can accurately probe the capabilities of LLMs https://lnkd.in/ePdFj5zB (4/5)

Next was an engaging discussion with Monroe Kennedy III on robotic dexterity and collaborative #robotics on the #TWIML podcast. This compelling work doesn't try to replicate humans with robots, but rather uses the affordances of unique sensing configurations to solve problems in new ways https://lnkd.in/gukF_tDy (6/9)

Next was an engaging discussion with Monroe Kennedy III on robotic dexterity and collaborative #robotics on the #TWIML podcast. This compelling work doesn't try to replicate humans with robots, but rather uses the affordances of unique sensing configurations to solve problems in new ways https://lnkd.in/gukF_tDy (6/9)

Next was an excellent discussion with Nicholas Carlini on #privacy and# security for #StableDiffusion and #LLMs on the #TWIML podcast. This stimulating conversation elegantly exposes the issues with these models, some of the privacy, security, and liability concerns involved, and some potential solutions. The rigorous conclusions certainly don't look good for the defendants in the large model #copyright cases. Highly recommend https://www.youtube.com/watch?v=FzodqiVPCNc (6/7)

Next was an excellent discussion with Nicholas Carlini on #privacy and# security for #StableDiffusion and #LLMs on the #TWIML podcast. This stimulating conversation elegantly exposes the issues with these models, some of the privacy, security, and liability concerns involved, and some potential solutions. The rigorous conclusions certainly don't look good for the defendants in the large model #copyright cases. Highly recommend https://www.youtube.com/watch?v=FzodqiVPCNc (6/7)

Next was a nice discussion with Vinodkumar Prabhakaran on AI's impact on social disparities on the #TWIML podcast. Prabhakaran describes interesting research on assessing how machine learning's tendency to obscure rater differences causes a whole host of problems, and presents a nice method to unpack those differences in ways that mirror the jury work of @msbernst https://www.youtube.com/watch?v=XOto8KFB8PM (4/8) #MachineLearning #AI

Next was a nice discussion with Vinodkumar Prabhakaran on AI's impact on social disparities on the #TWIML podcast. Prabhakaran describes interesting research on assessing how machine learning's tendency to obscure rater differences causes a whole host of problems, and presents a nice method to unpack those differences in ways that mirror the jury work of @msbernst https://www.youtube.com/watch?v=XOto8KFB8PM (4/8) #MachineLearning #AI

First was a deep dive into the latest #causality research with @osazuwa on the #TWIML podcast. I'm not as up on this research so it was nice to get an overview of the space, although I disagree on the causal interpretation of #LLMs discussed here https://www.youtube.com/watch?v=9c8Ln_ECmbA (2/6) #MachineLearning

First was a deep dive into the latest #causality research with @osazuwa on the #TWIML podcast. I'm not as up on this research so it was nice to get an overview of the space, although I disagree on the causal interpretation of #LLMs discussed here https://www.youtube.com/watch?v=9c8Ln_ECmbA (2/6) #MachineLearning

Next was an excellent conversation with Sameer Singh on #NLP trends in 2023 on the #TWIML podcast. NLP really has exploded into the popular consciousness this year, but this conversation gets at the scientific perspective on real progress and what remains to be done (tldr; a lot). Some of the discussion is a bit too techno-solutionist from my perspective, but I still highly recommend the episode https://www.youtube.com/watch?v=EC40F-R059U (4/7) #AI #ML

Next was an excellent conversation with Sameer Singh on #NLP trends in 2023 on the #TWIML podcast. NLP really has exploded into the popular consciousness this year, but this conversation gets at the scientific perspective on real progress and what remains to be done (tldr; a lot). Some of the discussion is a bit too techno-solutionist from my perspective, but I still highly recommend the episode https://www.youtube.com/watch?v=EC40F-R059U (4/7) #AI #ML

First was a nice discussion with Sergey Levine on upcoming trends in reinforcement learning on the #TWIML podcast. I particularly liked the conversation on efforts in large scale #robotics #data collection efforts https://www.youtube.com/watch?v=dvO_jR1B5rs (2/4) #AI #MachineLearning

First was a nice discussion with Sergey Levine on upcoming trends in reinforcement learning on the #TWIML podcast. I particularly liked the conversation on efforts in large scale #robotics #data collection efforts https://www.youtube.com/watch?v=dvO_jR1B5rs (2/4) #AI #MachineLearning

Next was a fascinating discussion with Catherine Nakalembe on supporting food security in #Africa with #MachineLearning on the #TWIML podcast. Not only is this an important problem, but the holistic, active approach to collecting different forms of data to support this effort is an excellent case study on algorithm development. Highly recommend https://www.youtube.com/watch?v=JXTHSAuuh44&t=1s (3/5)

Next was a fascinating discussion with Catherine Nakalembe on supporting food security in #Africa with #MachineLearning on the #TWIML podcast. Not only is this an important problem, but the holistic, active approach to collecting different forms of data to support this effort is an excellent case study on algorithm development. Highly recommend https://www.youtube.com/watch?v=JXTHSAuuh44&t=1s (3/5)

Next was a great conversation with Michael Kearns on the #TWIML podcast on service cards and #MachineLearning governance. This extension of model cards is important, although I'm highly skeptical of the ability of companies to regulate themselves in this regard. The degree of academic capture by big #tech on display here is also an important sidenote. Highly recommend https://www.youtube.com/watch?v=XaErOGD0OL8 (4/8) #AI

Next was a great conversation with Michael Kearns on the #TWIML podcast on service cards and #MachineLearning governance. This extension of model cards is important, although I'm highly skeptical of the ability of companies to regulate themselves in this regard. The degree of academic capture by big #tech on display here is also an important sidenote. Highly recommend https://www.youtube.com/watch?v=XaErOGD0OL8 (4/8) #AI

Randi Williams, PhD student at the #MIT Media Lab, is teaching preschoolers [5 year old] the fundamentals of artificial intelligence.

#TWiML https://twimlai.com/talk/225

#AI #ML #podcast