Видео

Great talk, thanks!

I can't wait to see where your talent takes you next. Exciting times ahead!

what up first

Very boring! He is an accountant of deep leaning! I'm gonna work on big things and try to do a giant leap, I hate reading unlimited papers to combine them and get another 0.001% accuracy and make code even more complex

This will Create Jurassic Park

There is no sound in the video

wtf is bioprelogy

Sorry, but what's the point of uploading this video without audio? A slide show with no context?

"As regards disease, the Lancet's Countdown on Health and Climate Change (2019) shows Climate-related deaths are a small proportion of all-cause fatalities (1990-2017). That is based on data per IHME (2019), and between 1990 and 2017, the cumulative age-standardized death rate (ASDRs) from climate-sensitive diseases and events (CSDEs) dropped from 8.1% of the all-cause ASDR to 5.5%, while the age-standardized burden of disease, measured by disability-adjusted life years lost (DALYs) declined from 12.0% to 8.0% of all-cause age-standardized DALYs. Thus, the burdens of death and disease from CSDEs are small, and getting smaller. However, the declines in death and disease rates from CSDEs since 1990 are only a small proportion of longer-term declines across the globe. In the USA, one of the few places with good long-term data, death rates from dysentery, typhoid, paratyphoid, other gastrointestinal diseases, and malaria - all water-related diseases and therefore, almost by definition, climate-sensitive declined 99-100% between 1900 and 1970. We are solving our problems with CSDEs faster than we are solving our other health problems." Indur M. Goklany

I enjoyed your presentation and found the concepts quite engaging! I do have a small note regarding the section at 9:30. It appears that the bottom left image represents a grammar-based genotype, and the upper left shows the phenotype of modular robots. The transition from genotype to phenotype fits the term "decoding". While the arrow indicating the progression from the modular robot design/phenotype to the 3D printable stl file along with other related outputs for fabrication and control, is separate from the decoding process and more accurately described as "modeling and exporting". Thought this might help clarify. Keep up the fantastic work!

8:00 It's incorrect as far as I know to refer to this as the "foundation modeling stage" -- I don't understand where that conflation of terms came from -- the massive _pre-training_ stage (together with later instruction-tuning/RLHFing) _produce_ a foundation model, but there's no such explicit stage in training.

How do you overcome the problem is when project 3D points from dense point cloud to 2D pixel coordinate, we have many 3D point lie on a light ray together?. How do you choose the right 3D point? I am very curious, please

1st

Thanks! Very interesting talk

That was wonderful, well done!

Lack of adequate compute is my main problem. Good talk. Thank you.

This was an excellent example how to introduce a study. Effective and short. I liked it very much. In my case very helpful. Thank you.

how can i have a co-supervisor from another country please i'm very interesting

For all you premeds, med students or residents.. Good luck!

Nice talk

23:05 The irony of Meta "supporting" OS 😂💔

Love this

Great talk Angela! Would love to have you in my server to talk about the NLLB project :)

Tim, you are amazing and you are doing such amazing work. Thank you for your contributions and all the hard work you do. QLoRA is nothing short of genius. I've found it a very effective regularization tool as well and it appears to improve generalization in protein language models like ESM-2. Again, thank you for all the incredible work you do. You're awesome!

Great talk!

Glad he acknowledges that not everyone had access to PCs and the internet. Important to do.

Great to use PINNs and they are usreful to use but you need to acknowledge the PINNs paper and ion fact you used the PINN schemnatic from our papers withiut acknowledging it! This is called plagiarism....

Thanks for sharing this talk. Very interesting and has a lot of information and useful research directions.

Now that it's 2023, where are we at with this approach...

This could have been easily a 90 minute talk. Ridiculous format to give only 20 minutes. There needs to be more than just obfuscating practical aspects of implementation. Even the author says he doesn't believe it will work with today's technology platforms if I understood correctly. Side note to the presenter. Be more assertive and a bit less patronizing in the tone of your presentation. While you had valuable information, I did have to grit my teeth to get through the supercilious style in which you delivered.

Thank you. Great presentation. you are very beautiful

Interesting

Shitheads of researchers. You really would like to destroy the music business.

Topic--safety of aeronautical machine

Thanks for sharing this Michal

Great explanation

Your survey and this video helped me in my work.

I think quantum vacuum is the coarse grained description of the excited state. The same way quasiparticles perceive a condensate as the vacuum. Integration of the superluminal degrees of freedom gives the "vacuum", hence the speed of light limit.

Thank you Dr. How can we have access to the time series data for the measurements? I will be glad if this could be provided for reproduction.

Thanks a lot for the explanations. I'm stuck on a water drop count project, the fps is too low. Would you have any tips to increase the fps when counting the drops? If not, the count will not be correct knowing that each cell phone has a different fps. Thanks.

Thanks for the talk! This was encouraging.

Thank you for sharing the video! Can I see the presentation file online?

Thank you so much for this video. Very interesting.

thank you, wonderful presentation

Very interesting.,

This is the second time I found this video, I think.

interesting presentation

Great Presentation. It is not my area of expertise but it introduced me to an interesting research direction.

Are these features still under development or are they already ready for use?