News for hackers. Stuff that matters.
Re:Why GPT-3 Matters by KKKKkkkk1 on Monday July 20, 2020 @03:16PM (#23900941)
But it is slowing down. In computer vision, we had MNIST solved and waited for more than 2 decades of exponential growth in compute until ImageNet was solved. That 98% percent accuracy on ImageNet is nowhere near good enough for applications like self-driving cars. How many decades until we reach a 10^6 error rate? Keeping in mind that exponential growth in compute is over.
Re:Why GPT-3 Matters by the8472 on Monday July 20, 2020 @04:13PM (#23901416)
> Keeping in mind that exponential growth in compute is over.

DL is getting increasingly specialized hardware, there's plenty of growth there. Plus what we're seeing here is GPT scaling up without algorithmic changes. Algorithms are advancing too.