Turing-NLG: A 17B-parameter language model

Re:Turing-NLG: A 17B-parameter language model by 01100011 on Monday February 10, 2020 @03:21PM (#22292843)

How long until the language models stabilize enough that we can bake them into a low-cost, low-power chip for edge uses?

[ Parent | on HN ]

Re:Turing-NLG: A 17B-parameter language model by slashcom on Monday February 10, 2020 @06:55PM (#22294701)

There’s lots of work on distillation, smaller models, approximations, etc. People already have simpler forms of these running on smartphones. Models seem to be growing faster than we can make them small though :D

[ Parent | on HN ]

Re:Turing-NLG: A 17B-parameter language model by saurkt on Monday February 10, 2020 @08:11PM (#22295108)

Yes, these things keep us up at night as well :-).

[ Parent | on HN ]

Re:Turing-NLG: A 17B-parameter language model by foota on Monday February 10, 2020 @03:28PM (#22292925)

I think this is largely unnecessary, can't things like TPUs handle the inference?

[ Parent | on HN ]

Re:Turing-NLG: A 17B-parameter language model by struxure on Monday February 10, 2020 @03:49PM (#22293132)

Using TPUs is expensive. Also some applications may need small latency

[ Parent | on HN ]

Re:Turing-NLG: A 17B-parameter language model by the8472 on Monday February 10, 2020 @04:06PM (#22293339)

Putting all your speech/text onto cloud machines runs counter to e2e encrypted messaging.

[ Parent | on HN ]

Re:Turing-NLG: A 17B-parameter language model by foota on Monday February 10, 2020 @11:54PM (#22296170)

I think you can get hardware like a TPU for consumer products?

[ Parent | on HN ]