yoquan

Yes. On an analogy level you can also examine whether you are in a dream or in reality by looking exactly to such details :)

Just to share a perspective. Doing business in Japan often requires a certain level of trust. Showing that you (the presenter) have something to count on is thus important.

The slides describing the "whys" of Mojo and its early vindication did a good job of clearing suspicion around its marketing hype, imo.

Because it's fun :-) And effort to bring up a could-be-commercial version is on going.

https://www.together.xyz/blog/redpajama

For a deeper look into stablecoin mechanism and performance of them during May 2022 event, I would recommend [1] which I think is much more interesting for curious readers.

[1] https://www.fdic.gov/analysis/cfr/bank-research-conference/a...

The model used by this research is T5 which is open sourced already. So I think once the dataset is released, we'll see the open version of pre-trained model very soon.

DocTR: https://github.com/mindee/doctr

It also has TensorFlow.js version to run in-browser: https://blog.tensorflow.org/2022/06/ocr-in-browser-using-ten...

Obviously it will relate to complexity in deep learning but I can't resist thinking about some AI models involving quantum computation stuffs :-)

Actually no. Each layer requires output from previous one, which means sequentially computation. While wider layers can utilize GPU parallel computation better. This is kind of trade-off between less memory (less parameters) vs longer time.

Hayabusa-2 is the first of its kind successful mission. And it got congrats from Elon Musk too! [1] Looking forward to new discoveries about early Solar System from the asteroid sample it brought back.

[1] https://twitter.com/haya2e_jaxa/status/1335513393122316288

To me learning cause-and-effect is a non trivial process, when recognizing relationship often comes first and intrigues "formal" reasoning process afterwards (if any).

The more formal in the later process, the closer we reach to the real "knowledge". So your suggestion is quite practical, in the sense that we can start from that and figure out how to push the engine toward the more formal spectrum.

Thanks for your confirmation, and addition for a feature request (sorry for a bit late response). So, it will be more practical that I rethink my requirements first. Will send you email if there is further question!

Hi, I'm looking for a graph-friendly DB that can sync well between many types of client (desktop, mobile) and server, similar to CouchDB/PouchDB. I found [1] below that seems relating but no further discussion. Could you give some more information about that?

Thanks, and congrats on the new funding!!

[1]: https://discuss.dgraph.io/t/data-sync-between-clients-server...

This and other gate level quantum softwares like Forest (Rigetti), Qiskit (IBM), Cirq (Google-not official) can be checked for more information in [1].

[1]: https://arxiv.org/abs/1807.02500

Here you are: http://www.hayabusa2.jaxa.jp/en/topics/20190711e_PPTD_ImageB...

Some more technical information: https://www.infoq.com/news/2019/06/google-earth-web-assembly...

Hmm, the article's summary table showed "only the top task layer(s)" for BERT fine-tuning. Is that correct?

The original paper [1] wrote:

"All of the parameters of BERT and W are fine-tuned jointly to maximize the log-probability of the correct label."

[1]: https://arxiv.org/abs/1810.04805

I'm reading this and just realized one author is the famous Field medalist, Shing-Tung Yau :-)

Original paper: https://arxiv.org/abs/1901.05415

This was posted a few times but didn't get traction. However, I found interesting views on several trending topics, so took a risk of reposting.

A quick overview is here: https://frontendmasters.gitbooks.io/front-end-developer-hand...

Penrose'paper is here: https://arxiv.org/abs/1808.01740

Thank you for switching to the video - I was late to response due to different timezone. (I've chosen slide over video thinking people generally don't like video. Seems the slide format was more disturbing)

That's great to know, thank you. Eager to try 1.10!

There is an attempt to create a role based access control by the guys at WePay according to the shared slide "RBAC talk" below [1]. Don't know why their repo [2] can't be accessed now, though.

[1] https://www.meetup.com/ja-JP/Bay-Area-Apache-Airflow-Incubat...

[2] https://github.com/wepay/airflow-webserver

Almost agree to what you said. Just a piece of more context: the 1st author wrote about connection between physics's Renormalization Group and Deep learning which was featured in [1] 4 years ago.

[1]: https://www.quantamagazine.org/deep-learning-relies-on-renor...

Since it is such a natural combination, it is difficult to search for the name in crypto context though :)

Oops, this has been posted several time before, eg. https://news.ycombinator.com/item?id=16440516

So, to add some new value for someone coming along accidentally, I would point out that Dart has native support for reactive stream [3], along with additional reactivex/rxdart [4], which makes it comfortable to adopt to current reactive trending.

[3] https://www.dartlang.org/tutorials [4] https://github.com/ReactiveX/rxdart

Google currently seem to be pushing on Dart and Flutter [1] as a new mobile development platform. Its appeal is the support for both prominent mobile OS, Android/iOS and the promising new one Fuschia [2].

[1] https://flutter.io/ [2] https://fuchsia.googlesource.com/