To be able to do research, you have to spend ~$1500 amortized over 1-2 years. And AI hardware is improving much faster than Moore's law.https://twitter.com/NandoDF/status/782276800038727680 …
-
Hardware is the least of our worries. Having high quality open datasets in areas that matter is vastly more important issue.
A $1200 TitanX would have been one of the fastest supercomputers in 2005. Next, specialized ASICs will make 2016 GPU farms look like a joke.
-
-
When that happens, large companies will *still* have 10k times the resources you have. But *you* will have the resources they had yesterday.
-
That's what accessibility looks like. Complaining that you cannot compete on resources with the most resource-rich entities is tautological.
- Show replies
New conversation -
-
-
you could always just "train longer"
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
So you’d say that 64-bit throughput is overkill for research right now? (I favour Titan X just wondering about practicality)
-
everything is fp32 these days, moving to fp16.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.