In depth discussion between Karl Friston vs Yann LeCun, Joscha Bach vs Yann LeCun; also deep technical discussion on the practical implementations of Active Inference with Spatial Web protocols, Energy-based Self-supervised Learning, Liquid Neural Networks, etc.
1 year ago
| 3
Glad to hear Timm! I have heard Petar Veličković say in a talk that he thinks graph neural networks will be part of the solution to AGI, and I dont think he should get away without two hours of explaining himself on MLST. Also generally the role of synthetic data. Generally wish the content was a biit more accessible as it can be challenging to keep track of all the "isms" when driving a car and following the conversation. Aside from that I really enjoy the free energy rabbit hole you've been in lately.
1 year ago | 2
I love anything that focuses and goes deep on the academic side of things, preferably in a way that still tries to be accessible to people without PHDs. I'm much less interested in the ego driven techbro stuff, and the nonsense debates like the George Hotz fiasco episode. MLST is at its best when it's that rare show actually covering these topics in a substantial way, rather than chasing views from the AI Bro audience.
1 year ago
| 8
A set of short form 10 minute dummaries of your longer episodes would be good too. A quick transcript of the main event for the time impaired.
1 year ago (edited)
| 6
Keep the content coming! such a informative channel. Big love from the UK 😆
10 months ago | 0
I'd love to see content and long form interviews/panels with experts on these topics : Geometric Deep Learning https://geometricdeeplearning.com Neural Algorithmic Reasoning https://algo-reasoning.github.io Category Theory for AI https://cats.for.ai/ Applied AI and Deployment AI law and copyright
1 year ago
| 8
amazing!! it would be great if MLST discuss with Geoffrey Hinton about the fast/slow weight on the context of multi time scale adaptation(evolution-developement-learning-inference) with multi time scale energy function and it’s minimization.
1 year ago
| 6
Great you are focusing... It seems a great opportunity to explore ways of using AI directly by users... Models curated and hosted as open source platforms
1 year ago | 0
That’s awesome, so excited! I want more Bayesian content! Structure learning, something more focused on active inference implementations, and Bayesian mechanics! I’d also love to see Friston x Wolfram 2.0 and Friston x Bengio.
1 year ago | 0
Joseph Le Doux on consciousness with application to artificial intelligence
1 year ago | 0
It is an amazing show! The great quality of content reflects all your efforts and preparation. Thank you! I am sure you will succeed in this new phase!
1 year ago
| 0
Machine Learning Street Talk
I've just quit my day job. MLST is going BIG in 2024! Buckle up everyone! What do you want to see on MLST this year? Please upvote others if you agree.
1 year ago | [YT] | 171