Viewing profile — oldesthacker

This is what Tomas Mikolov said on Facebook: > I wanted to popularize neural language models by improving Google Translate. I did start collaboration with Franz Och and his team, d…

The results of Table 3 are not really exciting. Could this change with 100 times more data? The key novelty in the specific context of this particular application is the quantized …

Let's take an example: his "unnormalised linear Transformer," a neural network with "self-attention" published in 1992 under another name. It wasn't just an idea, it was implemente…

> Isn't that exercising political capital? These are very good points. > ... the worst of many abusers of the citation system. Is this ad hominem argument by user "lacker" meant to…

Thanks! Of course, the silverbacks! No offense.

The machine learning field as a whole has a huge credit assignment problem. This post seems to encourage other ML researchers to come out with their own priority disputes. Tomas Mi…

How new is this insight? The failure of teams of highly capable individuals is often due to egocentric silver-haired gorillas in the room: >They spent excessive time in abortive or…

> It was all about the economics, all the way back from early 90s. We cant do x node because it would be too expensive to sustain it at a node every two years. Totally agree. > AI …

Interesting bit about Samsung’s secret sauce: Samsung went even smaller than Intel, showing results for 48-nm and 45-nm contacted poly pitch (CPP), compared to Intel’s 60 nm, thoug…