We lost a titan of programming languages, programming methodology, software engineering and hardware design. Niklaus Wirth passed away on the first of January. We mourn a pioneer, colleague, mentor and friend.
We have just released a paper overviewing our system for Business Email Compromise (BEC) detection and the key design considerations behind it https://t.co/yLOqqv0LWT
@HelenaHorska Mohl bych poprosit o krátké vysvětlení co si z grafu odnést? Asi mi něco uniká, ale nerozumím mechanismu jakým S&P 500 ovlivňuje kurz EUR/CZK, nebo jaká je společná příčina?
I'm excited about having the opportunity to present my keynote at @ECSACONF earlier today: Software Architecture in the age of Cloud Computing by @jgergic#softwarearchitecture#cloudcomputing https://t.co/vCWAiNju0n
@paulg This holds only in domains where mistakes are cheap. With higher stakes you also need higher development standards.
Also, compared to traditional software it is much more difficult to tell working from broken ML feature as its quality is a continuous function.
Tried the 'one sentence per line' guideline [1,2] today and its effectivity exceeded my expectations. Will stick to the advice.
[1] https://t.co/W1NJ5bO4GM
[2] https://t.co/diWj6FwVjX
@mo_norouzi Interesting. Would you get similar output if you defined multiple variables and used only a subset of them?
"a=bird. b=cat. c=fish. d=deer. e=dog. a photo of b sitting next to e in a house made out of sushi."
Detection technology research papers consistently misunderstand how much more impact a false positive has compared to false negatives.
They often get equal weight.
Sometimes false positives are even ignored.
Proud to present EMOCA, our latest attempt to breathe some emotion into the in-the-wild face reconstruction arena. #CVPR2022
Code & model: https://t.co/tymybbrZF8
PDF: https://t.co/PXTZCHz3Fi
Video: https://t.co/wIXn0memWG
With @Michael_J_Black, @BolkartTimo
@tdietterich@amirbolous@sergeykarayev@emilymbender Humans are also far from perfect at these tasks so in my eyes it does not make sense to set the level to: ‘plays chess as good as Deep Blue’.
GPT-3 performs quite well at many tasks that require human reasoning or even creativity (in the sense we used these words before)
Fun conclusion of today's debugging:
With the most recent numpy, importing ddtrace causes floating-point arithmetics related warnings in scikit-learn. However, this happens only when celery module is installed.