Well, the memory limit of a WebGPU process would be the limiting factor for training. In addition, the bandwidth between the nodes and the parameter server, if doing training in data-parallel fashion, is another limiting factor.
I wonder if WebGPU will reduce dependence on CUDA, esp as Tensorflow is being ported to WebGPU. With WebGPU's improved performance and utility and the fact that it runs on top of Vulcan, Metal and D3D with any GPU that has drivers for those, I wonder if DL folks will find it more tempting to use TFJS/WebGPU via Electron or the browser and just be done with CUDA (i.e. break or soften NVIDIA's monopoly)
<<
we hurt members of our LGBTQ+ community when they felt they couldn’t participate authentically and we didn’t respond quickly or strongly enough in supporting them. Worse, through our handling of this situation, we made them a target for harassment as people debated their right to express themselves and be addressed according to how they identify.
I am responsible for that, and I am deeply sorry. We absolutely support the LGBTQ+ community.
>>
GPT-2 could have written a better and more genuine apology.
First of all, you don't just use the new PR formula of naming the wrong, saying you take responsibility and then saying you're sorry. You need to dig deep and identify why you committed the wrong in the first place (homophobia, sexism, mean heartedness?) and commit to therapy and reforming.
hey guys, I want to take this opportunity to ask if anyone here was born in the early 70s in Baghdad and went to Baghdad College high school (https://en.wikipedia.org/wiki/Baghdad_College) in the early to late 80s. We left in 87 to London, UK then Boston, US after that, just before the end of the war with Iran, and I'd like to reconnect with any classmates/old friends who may also be into tech... I'm sure everyone is all over the world by now, in different places. Thank you, and take care wherever you are.
No, and they have a large attack surface thru various techniques, so not a good choice for anything but keeping touch with friends and relatives (assuming no mass usage occurs)
I see the following models of monetizing content on the web:
1. Gift economy (you give it away, take likes)
2. Goods and Service economy (you charge for it, sufficiently to make a living)
3. Advertising economy (many people want it but not many willing to pay for it)
4. Tipping (not a viable economic model, unless it's made compulsory and a minimum is set in which case it's not tipping)
1 & 2 have been with us since the dawn of human civilization.
3 is a modern invention and to many it is a failed experiment.