Julien Girard @rhapsodos@scholar.social

PJ Reddie, one of the most wholesome and cool scientists of this world, quitted research in computer vision due to ethical concerns:
https://nitter.net/pjreddie/status/1230524770350817280#m

A cool conference on the deep learning hype, at 36c3:
https://media.ccc.de/v/36c3-11006-der_deep_learning_hype

Good questions here:
* is deep learning a sustainable technology in our resource-limited world?
* are those deep learning systems useful for anyone else than big companies and a restricted set of academics?
* What about the anonymous workforce used to label or mimic deep learning datasets?
* how the heck can we continue to publish to top-tier conferences with 10% of results that are not reproductibles

A petition from some Machine Learning researchers to encourage major conferences to offer a remote participation option:
https://www.change.org/p/organizers-of-data-science-and-machine-learning-conferences-neurips-icml-aistats-iclr-uai-allow-remote-paper-poster-presentations-at-conferences?recruiter=144678980&recruited_by_id=51a298c0-2d0e-11e4-843e-ad2ec88f7f02&utm_source=share_petition&utm_medium=copylink&utm_campaign=petition_dashboard

If I were a scholar on economical history and philosophy, I would say that it's just capitalism trying to devour another field. That after reaping everything it could from our diseases, our culture, our food, our emotions, it goes for our knowledge and spirit.

But I am none of those. I am just a poor PhD student raging against this state of affair, and hoping that things will change.

What is happening in research (and other public sectors) is a colonization. All are colonized by a foul mindset: looking for material value, immediate profit and systematic exploitation.

Well guess what: trying to quantify how valuable is a knowledge, how much money an idea may give... it simply does not work, and I think it is wrong to think this.

Favouring ideas that are economically rewarding will simply enslave research to economy, preventing to look for new perspectives.

All recent reforms are destroying what is making academia unique, engaging, rewarding and fun. I think that researching knowledge for a society is a wonderful job. Teaching, when voluntary, is a genuine gift. It empowers people, helping them to reassert their situation, giving them tools and strengths to face the future.

I fully support their action. As a young PhD student, working in french academia is a bleak perspective. Competition is insanely strong, salaries are quite shitty when compared to engineers, analysts, or people with a technical background and about the same years of study.
Psychological hazards in research, which are already quite huge, would certainly increase when removing material and intellectual liberty for researchers.

Hop, some time for french politics (don't flee yet, this might be interesting).

Over the recent years in France, there is a global trend to privatize a LOT of previously public-owned services. Transportation, energy production... and public research funded by the state is no exception.

Some academics wrote a statement about this. They are planning a team application to one of the french "research quality assessment group" http://rogueesr.fr/hceres-en/

My paper on formal verification of deep neural networks using simulators was accepted at the European Conference on Artificial Intelligence!

CAMUS: A Framework to Build Formal Specifications for Deep Perception Systems Using Simulators
https://arxiv.org/abs/1911.10735

I will be happy to discuss about it!

(I am building up my hype for the congress I will speak)

Philip Wadler is speaking right now at College de France, about using a proof assistant during teaching.

The talk will be available freely on the website of College de france

https://www.college-de-france.fr/site/xavier-leroy/seminar-2019-12-12-11h15.htm

The US National Transportation Safety Board recently released their report on the 2017 Uber autonomous vehicle crash that resulted in the death of a pedestrian.

As an AI researcher, I find this piece of report particularly interesting:
https://dms.ntsb.gov/public/62500-62999/62978/629713.pdf

There is no mention of deep learning in the report, but the phrasing and the kind of error mentioned gives me little doubt that at least some part of learning technology was involved in the crash.

Currently using the good ol' technique on writing a presentation from a paper:

1. take the paper in its mostly final version
2. replace every empty line with a \end{frame}\n\begin{frame}
3. replace complicated sentences by simple ones, discard those that are too complicated
4. add bullet points
5. tadaaa

(actually don't do that it's terrible to read a succession of bullet points)

https://www.jefflindsay.com/SkepticQuotes.html

Expert opinion should be listened. That does not make them unconditionally true

How I will probably behave when I will look for a job:

https://joelgrus.com/2016/05/23/fizz-buzz-in-tensorflow/

A cool review on the question of gender: https://sillyolyou.wordpress.com/2019/07/30/mapping-the-gender-debate/

Previous post obtained more boost and favs that I could possibly hope for. Please do the same for my publications thanks.

(actually don't, smash the publish or perish mindset)