I'd say you can judge the quality of a marriage by how quickly they move on after their spouses passing. Some cases it's less than 6 months, which says something

I like the glossy finish of professionally developed photos over anything I can print at home. Not the same, I definitely don't have photobooks of printed digital images. But one hour photos and going to the pharmacy to develop things has long been abandoned I thought

But then I guess everything can be considered poisoned by previous work?

Maybe, we can allow training, but legally pursue people that use the models to generate copyrighted content? Either way I don't think AI using copyrighted work is going to stop, even if one country outlaws it, there will be some island somewhere with an internet connection hosting AI

I understand that he wants to minimize research into things that won't immediately provide value. But suggesting that biological beings have a monopoly on consciousness is short sighted. Elements are not conscious, cells are not conscious, but a large grouping of cells and suddenly intelligence emerges. Who is to say that a simulation of cells cannot do the same?

I still think the VM route is the better way forward as this would require being updated as linux changes. But I can see it becoming a small obscure system maintained by someone with too much spare time. Without the ability to use sockets though that reach out to the web I can't see too much use cases. Get SSH working and then you'll have something

Pretty sure you'll find something like 90% of the hires are from US universities, so they are slowly making exceptions to make it not effective. I wouldn't exclude existing people either

He is using Torch, anytime anyone says they wrote things by hand using python I'm suspicious. Truth is there is no reason not to use a machine learning framework, so you are always using something. (Whether that be PyTorch, Tensorflow, or something else) I find it interesting that the model definition is probably the important part (assuming the model is using standard transformers or existing techniques) Say someone else does the training, you just download the weights, it kind of makes the training process a chore. The code using a framework means the code is mostly throw away boilerplate. The only thing that matters is the results, and the model that generated them. I suppose the training data being available is important as well (as we see with ChatGPT) But then again, I think I would have moral qualms with what they are doing regarding copyright

Apple isn't a great platform for services, have a simple AI model running on a webcam, the webcam just stops returning images to opencv after a few hours, no reason or errors, a reboot fixes it. And I think it goes into some sort of sleep mode every now and again. Webpage disappears, go to login to the machine, then it starts working again without anything changed

I'm not sure I would describe Microsoft's direction after the azure cloud pivot as good. Seem to be in the intel style of idleness

I think you'll find that elite institutions are not really that much different than other state schools, just they accept fewer students and are "exclusive". I think there was an argument that you can flaunt that you were accepted to the school, as people fawn over the name, where the degree and GPA are secondary to the name of the school. Pretty sure it was the guys behind freakonomics. The admissions process tends to favor people with money, because they know their kids can coast on the name of their school alone and get into top positions

I would support gladiator humanoid robot games

I wouldn't say this is doom or gloom for CS and/or programming, just more a sign of the economic uncertainty due to inflation and interest rates. Kind of a wait and see approach. I would assume college hiring isn't completely stopped, good programmers are probably still getting offers. But someone with a degree on the resume and little else that shows they even like programming are probably pretty abundant

I assume this is similar to buying news stations, some sort of indirect form of narrative control? That or maybe they really want to fix the stats on FIFA

"Accidentally"