1,867 karma · joined July 1, 2016
I subscribe for a month, watch any new shows or movies I like, cancel and move on. All of my devices are 4K+HDR capable, I don't know why "up to 4 devices" is even relevant anymore if you can't share the account with anyone else. My family? Living elsewhere.
Now that there are dozens if not hundreds of streaming services offering exclusive content there is no paying customer growth possible anymore. What are they counting on? That consumers will pay for more than one subscription on different platforms? While they essentially only watch one at a time? How stupid is that!
"Wait, you are saying: yesterday the value of money was 1.0, but today it is 0.99? When you query what is the value of money, please explain first what 'today' and 'yesterday' means please sir"
So, if I am getting this right: you cannot update an already trained model with updated data. Doing so will "confuse" the model with contradictions, without the ability to assess what information or concept is valid now.
The consequence would be to train new models with the updated training data each time, and discard the old ones. Which is not viable at all, to be honest.
This dynamic reminds me of unpopular or shady laws being passed in some parliaments in the middle of the night or during major sports events where the representative attendance and public attention is at its minimum to get enough majority votes. This "biggest heist in game history" follows the exact same dynamic.
That study reeks of corporate consulting and management. And it's out there now, research is going on about this.
But, in general, I wonder why old papers (that are basically a foundation to other theories) are not critically re-reviewed to see if they still hold to scrutiny today? Hmm.
I've read some comments here that I agree with; in a lot of cases the goals in dev are set to implement a solution to a problem either as fast as possible, or without a limited scope.
"We need a software that can fly rockets" - Yeah, but where to?
I studied computer science twice. Both times I have been taught to implement corner cases instead of working on clearly defined projects with scope and specification. "Write a small program that does X" yeah, but the solution space is unlimited, then, maybe? Can you elaborate? Put a scope on it? I mean, can do, may I order compute time on the campus super-computer?
Crap software was there ever since, and it got worse ever since.
Like that click-baity headline from the Telegraph, it has a better chance to get clicks phrasing it that way, instead of "Elephats without tusks are more likely to survive violent poachers", imho.