MIT 6.S191 (2019): Visualization for Machine Learning (Google Brain) financial deepmind

This lecture is a gem, and I'm left scratching my head as to why it has just a thousand likes; and seventeen dislikes — really? About word vector visualizations: just wondering if it even makes sense to try to get rid of biases in word vectors without curating every single bit of the corpus? Is there some way out of this?

This is very inspiring video for me: in 2018 I was trying to develop the abstract model for emotion recognition from text, I called it 'semantic melodies' in connection with text tonalities. It was very poor, but at the same time I found that some words like spirit/inspiration/respiration are connected between at least English and Russian in the same way. Points shown in this video could help to develop those ideas further! I've saw few videos of words embeddings already, but this gives much greater taste. Thank you so much, I'm very grateful this channel exists, please keep your work going, I would be watching every new one video!

She knows how to talk and present.

easy to make fool

Great topic! Great talk!

Out of imagination concept! Moving towards Natural Language…

sometimes there is no one-on-one mapping between 2 different languages, e.g., there is just no a corresponding word in language A for a to-be-translated word in language B, how will the computer deal with such case?

Watching this after "Connections between physics and deep learning" by Max Tegmark is interesting (2016). He focuses a lot on things like locality that exist in the natural world we live in and neural networks. It is a very "aHa!" moment to see the languages behaving in such a local way in the embedding projector. I think that's very neat.

thnx!

<3