As Trump time comes to an end, a disturbing section of the left is thinking about reaching out to the hard right to try and make itself relevant with his angry, bigoted followers... [a short thread]

So it turns out that Google Chrome was making everything on my computer slow *even when it wasn’t running*, because it installs something called Keystone which is basically malware.

I made a website because this shouldn’t

Wired first reported on how bad Keystone was 11 years ago when they put it into Google Earth (they seem to put it in all their popular

The fact that Keystone hides itself in Activity Monitor is bizarre. (The only sign of it was excessive CPU usage of WindowServer which is a system process).

I don’t know if Google was doing something nefarious with Keystone, or a third party figured out how to (which Wired warned about). But either way, I’m not inclined to give Google-the-organization the benefit of the doubt (despite the many good people who work on Chrome)...

...since it's been a decade+ and this still hasn't been "fixed".

There is no reason for auto-update software to need to do what Chrome/Keystone was doing. It also has a long history of crashing Macs.

Important paper from Google on large batch optimization. They do impressively careful experiments measuring # iterations needed to achieve target validation error at various batch sizes. The main "surprise" is the lack of surprises. [thread]

https://t.co/7QIx5CFdfJ

The paper is a good example of lots of elements of good experimental design. They validate their metric by showing lots of variants give consistent results. They tune hyperparamters separately for each condition, check that optimum isn't at the endpoints, and measure sensitivity.

They have separate experiments where the hold fixed # iterations and # epochs, which (as they explain) measure very different things. They avoid confounds, such as batch norm's artificial dependence between batch size and regularization strength.

When the experiments are done carefully enough, the results are remarkably consistent between different datasets and architectures. Qualitatively, MNIST behaves just like ImageNet.

Importantly, they don't find any evidence for a "sharp/flat optima" effect whereby better optimization leads to worse final results. They have a good discussion of experimental artifacts/confounds in past papers where such effects were reported.

Here is a special thread introducing our series on “The Making of the Islamic World.” It offers 10 episodes about the transformation of Islamicate societies between the 7th and 17th centuries. You can find the series page here: https://t.co/7mL9eWmMmz

This is Chris Gratien posting. I created this series during Fall 2020 for my students as asynchronous material supported by readings and discussion. This is going to be a long thread, but it will be a great shortcut for those looking to make similar use of these materials.

There are two items worth pointing out: 1. This is an overview of the political, social, and cultural history of Islamicate societies. It does not deal with the origins of Islam or many theological questions. It does discuss Islam’s spread and the transformation of Islamic law.

2. This is the first part of a two-part sequence that takes the history of the Islamic world into the present. It emphasizes developments and themes that have particular relevance for later periods.

This is really a multivocal narrative of Islamic history. We feature over 20 scholars from almost as many institutions. Many of those who worked on it, myself included, study the early modern and modern era but have teaching obligations that take us further back in history.

So, I want to talk about how sexual harassment and fair pay are linked at Google (and beyond), because I think that's an angle that isn't being highlighted enough in the coverage of these walk-outs. I'm going to frame it largely around my personal experience there. Thread.

When I was sexually harassed by the director of the area I was working in, I was afraid to report it because I was worried that "getting him in trouble" would result in the subtle retaliation of missed leadership opportunities.

I wanted to continue working on the team I was on, because I'd gained a lot of very deep knowledge and expertise in that area, as well as reputation and camaraderie with the other folks working in that area. I didn't want to make the situation more "difficult."

To get promoted at Google, several need to happen: 1. you need opportunities for ownership and leadership above your current level (basically, opportunities to show you're working at the next level you're trying to get promoted to). The work you're "assigned" has a big impact.

2. You need glowing reviews from peers, *at or particularly above the level you're hoping to get promoted to.* Basically, you need people a lot more senior than you to say you're doing awesome work.