วีดีโอ

Good

David Blei is the Woody Allen of the academics.

This will not end well. Can we PLEASE CeaseAi -GPT?

Thanks for sharing the great talk, but the video has got no sound after 17mins...

The power of search is amazing, i have no clue how i got here

Hi. When can I expect the PROBPROG 2021 talks to be uploaded?

The angriest talking person ever in a conference.

I love how confident this guy is. Knowing that he's right, no matter the size of audience or number of questions. Wonder what progress they made since.

Such cool work 🤓

Wow, this is incredible!

One of the best talks I have seen on this :)

I know he's tackling the Occam Razor version of this but I wonder how this could be extended to perform on more difficult coding problems by incorporating expert knowledge

Can anyone share an explanation about 'solver' referred in the video? What is the distinction between a model and a solver they were talking about?

"DL community may be repeating history... #BOFAI (Bad Old Fashioned AI)" @12:30 epic line ;)

Good shit! :)

fantastic

Thanks

A very good description of BBVI.

where is noah's talk?

11:27 Saad feeling very sad

This lecture is gold and very unique. Deserves way more view

"Statistics is mathematics of data." -- Zoubin Ghahramani

Great talk by Yordan! I am eager to learn more about the Infer.NET framework and some of the new models, Alexandria for example, is new to me.

Dave Blei humor is the best humor. "20 minutes? We'll just start again" "That is handwavy even by my own standards"

This guy is taking a stab at Josh's hard problem of learning, and the video has 2 views after 2 weeks. People on this planet have no taste about what is interresting

My go at the learning = programming idea. (very related to the Noah Goodman's Probabilistic Language of Thought Hypothesis) Please correct me if I'm wrong Tuning parameters of existing code: not sure Writing new code: "normal learning" where you learn new things, mechanisms, relations, dynamics by experience. Extending or fixing code: Learning from mistakes. Eg: your tennis backhand sucks and the trainer tells you how to fix it, or you take an exam and review your wrong answers Rewriting/Refactoring a library of code: re-learning? Like Jaynes saying that a lack of statistics background is a good thing because there will be less to unlearn? Having to re-consider your entire perspective on a topic. Adapting code written for other purposes: Analogy making. Eg using what we know about coding to understand about human learning like we do now Getting code from other people or published sources: human language, reading, formal education. That's why books are so powerful: you go from having to write everything from scratch to just pip install everything. (or rather gem install for the real hackers) Debugging: thought experiments or regular experiments to find our where your are wrong. Testing hypotheses. Asking the right questions. Rewriting existing code in a different language: not sure Compiling code (translating from interpretable source code -> efficient machine code): training / growing myelin sheaths Writing a new language or compiler: meditation? not sure

Infer.NET is now open source at github.com/dotnet/infer