NYEPLAK

Quantifying the performance of the TPU our first machine learning chip By Norm Jouppi, Distinguished Hardware Engineer, Google Editors Note : Learn about our newly available Cloud TPUs. We�ve been using compute-intensive machine learning in our products for the past 15 years. We use it so much that we even designed an entirely new class of custom machine learning accelerator, the Tensor Processing Unit. Just how fast is the TPU, actually? Today, in conjunction with a TPU talk for a National Academy of Engineering meeting at the Computer History Museum in Silicon Valley, we�re releasing a study that shares new details on these custom chips, which have been running machine learning applications in our data centers since 2015. This first generation of TPUs targeted inference (the use of an already trained model, as opposed to the training phase of a model, which has somewhat different characteristics), and here are some of the results we�ve seen: On our production AI workloa...

Quantifying proteins in dried blood spots after decades of storage! ...and now for something completely different! I just find this one all-around interesting. First of all -- how did they find blood spots from newborns that were 40 years old?!? Is this a commonly acquired and stored in normal clinical practice? If so, what a potential resource! The paper is an analysis of the feasibility of large scale sample biobanks using blood spots dried on paper. If you were going to set up a biobank, this would be a cheap and easy way to do it. Finger prick, drop of blood on paper, store it. To study the stability of the proteins they use an immunoassay technology called Proximity Extension (PEA) and assess 92 proteins across samples going back as far as 50 years (I think I remember it saying 50 years in the paper somewhere) and look at different storage and acquisition techniques. A large number of the proteins appear to be genuinely unaffected by degradation over time. Another population of pr...