854 karma · joined January 15, 2016
- Many people who did a PhD and didn't want to do research don't do a postdoc
- I would say, "40% have left so far'. Following the same cohort for a few years might yield even higher numbers (because as long you haven't made your mind about quitting research, you are still a postdoc and not accounted for leaving, even it's your 10th year...).
- You can look up architecture decision record (ADR). There are plenty of good resources on google.
- MR/MR contains a ref to the ticket, an explanation of the changes, proof of work or way to test the change.
- Know how to leverage your tools (git, editor)
- Be curious [other languages, tools (formal methods, fuzzing..), fields (web if you embedded, or the other way around), maths ... ]
Works fairly well.
Might even give them one issue to show you are serious and know what you are doing.
You brain, in response, build more of those. That's why coffee is less efficient over time, and why you feel more tired without coffee when you are used to drink some'
A proven conjecture is IMO better than an unproven one.
But usually, a new proof would shed new lights or build bridges between concepts that were before unrelated.
And in that sense, a proof not understandable by humans is disappointing, because it doesn't really fullfil the need to understand the reason behind why it's true.
If you believe in functionalism (~mental states are identified by what they do rather than by what they are made of), then current AI is not a dead end.
We wouldn't need to define intelligence, just make it big and efficient enough to replicate what's currently existing would be intelligence by that definition.
Then, this library can be used for instance (their word) - Display images from two cameras. - Co-register cameras and compute the transformation from one camera's space to another. - Select regions in images for training machine learning (ML) models. - Perform image segmentation using a pre-trained ML model. - Convert radiance images to reflectance by utilizing a reference panel. - Display spectral signatures for in-depth analysis.
Source: I worked on Parrot's code. And you can make your mind by looking at the other projects in Parrot dev group...
They are not a tool per se, but just a way to compile your project.
Unprecedented, so far !
Cool product, terrible boomer management.
I compile my own version of Infinity for reddit with my keys and it works just fine.
In 2015, Paris Airport Air Traffic Software were running on windows 3.1.
Is windows 3.1 outdated ? Yes. No longer a public facing product ? Yes.
Dead ? Absolutely not (at least back then, maybe it got decommissioned).
Around that time, a friend of mine was doing his master Thesis in an astronomy lab. He had to load fortran 77 routines from a 5" floppy disk that were doing critical computation (they were in the process of migrating). Same diagnosis: outdated, became a niche product, but not dead at all.
Lisp was a major programming language as it powered nearly all AI research for many many years and many of it's ideas have spread into languages we use today. For instance, it was the first language with a garbage collector !
It has many more or less direct descendents that are used
The fact it was created some 70 years ago, but still run nearly unchanged and has active users and project is an achievement.
HN itself is (was ?) Written in lisp if I'm not mistaken
There is this list https://common-lisp.net/lisp-companies
Racket/scheme are lisp, you'll have to include them as well.
My team work on a project with many "funny" names, including but not limited to, liboyster, libowl, several characters from Fort Boyard (https://en.m.wikipedia.org/wiki/Fort_Boyard_(game_show), patric the starfish....
It was awful to work with, please don't go that.