aveni

Based on press, there are already some systems being deployed (Argonne National Labs, Pittsburgh Supercomputing Center).

Isn't this analysis orthogonal to pg's claim that the wealth tax is taking away X% of your startup equity?

Yes, if your startup is growing by 10% YoY your wealth is still going up even under the tax. However instead of you owning A% at the end, you only own <<A% because you've been selling equity every year to pay the government. Just look at your first plot between 1% and 3% tax. A delta of 2% tax results in losing 70% of your value.

Your post seems to suggest the idea "Since you're making so much anyway it's ok to take away most of it." Though of course there is a balance, I think this kind of trade would greatly reduce risk-taking and entrepreneurship.

You can use standard ML frameworks :) The code then goes through the Cerebras Graph Compiler to produce an executable that runs on the wafer.

More of the software stack was described at HotChips today, covered by AnandTech: https://www.anandtech.com/show/16006/hot-chips-2020-live-blo...

A lot of system-related questions were answered at HotChips last year: https://www.hotchips.org/hc31/HC31_1.13_Cerebras.SeanLie.v02...

And SC19: https://secureservercdn.net/198.12.145.239/a7b.fcb.myftpuplo...

Slides from their HotChips presentation: https://www.anandtech.com/show/16010/hot-chips-2020-live-blo...

GPT-3 is not that expensive. Estimating from the paper, to train the model, the GPU hardware costs were a few million dollars, and the electricity costs were probably under 100k. This is totally feasible for many companies today, especially if the hardware is a fixed cost and can be reused for training multiple models.

And as mentioned elsewhere, inference for a trained model is much, much cheaper.

I ride an electric scooter approx. 4 mi each way Caltrain <-> work everyday. I regularly maintain 15mph+ trip average.

Hi HN! We are a pair of students at MIT trying to measure how well humans can differentiate between real and (current state-of-the-art) GAN-generated faces, for a class project. We're concerned with GAN images' potential for fake news / ads, and we believe it would be good to know, empirically, how often people get fooled under different image exposure times.