I firmly believe it will slow down significantly. My prediction for the future is that there will be a much bigger focus on a few “base” models that will be tweaked slightly for different roles, rather than “from the ground up” retraining like we see now. The industry is already starting to move in that direction.
While I agree in principle, one thing I’d like to clarify is that TRAINING is super energy intensive, once the network is trained, it’s more or less static. Actually using the network isn’t dramatically more energy than any other indexed database lookup.
Actually, Windows does allow you to use an alternate “compositor”— a feature which is used quite frequently in the industrial/embedded space. Windows calls them “custom shells”. The default is Explorer, but it can be set to any executable.
https://learn.microsoft.com/en-us/windows/iot/iot-enterprise/customize/shell-launcher
Nope, it’s either inspecting the TTL of packets coming from your phone (unless you have a VERY custom setup, the TTL from devices other than your phone will be very different), or it’s deep packet inspection. I tried to trick t-mobile last year into giving me home internet on a phone sim, so I did a whack ton of research.
How is it scammy or gimmicky to put ads on a product you don’t pay for? I guess I’m just missing the disconnect here. This service costs Nvidia money to run.
You don’t need to exercise your stock options to access their value. It’s common practice to take loans out against their value, which allows you to access your money effectively tax free by instead paying interest against the loan. This is (again) a fairly commonplace practice used to make collecting tax difficult, and allow them to make the argument to regulators that they aren’t actually being paid that much, it’s totally just options they would never sell off. That’s why C suite has such a “burn everything to the ground, as long as our stock price goes up” mentality, because if it doesn’t, they have to start worrying about interest on their loans— because they have fairly low liquidity (percentage wise).
Base pay is a meaningless figure in tech. Total comp is all that matters. The true high rollers intentionally keep their base pay low to dodge taxes. This doesn’t mean they aren’t still getting and using this money, it’s just an intentional straw man to allow them to make this exact argument.
You can attach a fake one in software via XVFB (X Virtual Frame Buffer). It’s a little involved if you aren’t familiar with X, but it only took me an hour or so to get setup. Then you don’t need any hardware at all, and can set whatever resolution you’d like.
If that were solely true, there would be a lot more competition in the field right now. Amazon, (and to a much lesser extent the other 2 big names, GCP and Azure) are so massive not because they have a lot of power (plenty of other companies like digital ocean or OVM have plenty of scaling power too)— but because the integrations between their products are so seamless. Most of that functionality has a foundation in FOSS software that they’ve built on top of.
Indexing and lookups on datasets as big as companies like Google and Amazon are running also take trillions of operations to complete, especially when you take into account the constant reindexing that needs to be done. In some cases, encoding data into a neural network is actually cheaper than storing the data itself. You can see this in practice with gaussian splatting point cloud capture, where they are training networks to guide points in the cloud at runtime, rather than storing the position of trillions of points over time.